Method for detecting running state of electrolytic cell

Abstract

The invention relates to the technical field of running state detection, and particularly discloses an electrolytic cell running state detection method which comprises the following steps: after determining a preset oxygen-in-hydrogen parameter and a preset hydrogen-in-hydrogen parameter, acquiring time queue data of a real-time oxygen-in-oxygen parameter and a real-time hydrogen-in-oxygen parameter in a sampling period; and performing fine-grained time sequence interactive coding on the real-time hydrogen-in-oxygen parameter and the real-time oxygen-in-hydrogen parameter by using a data processing technology based on deep learning to capture a time sequence cooperative change mode between the parameters, and further, determining the real-time hydrogen-in-oxygen parameter and the real-time oxygen-in-hydrogen parameter based on the characteristic offset of the actual time sequence cooperative change mode between the parameters relative to the preset parameter. Therefore, whether the running state of the electrolytic cell is abnormal or not is intelligently identified. In this way, the accuracy and efficiency of detection of the running state of the electrolytic cell can be effectively improved, potential problems can be found in time, corresponding measures can be taken, and safety and stability of the production process are guaranteed.

Description

Technical Field

[0001] This application relates to the technical field of operating state detection, and more specifically, to a method for detecting the operating state of an electrolytic cell. Background Art

[0002] In modern industrial production, electrolytic cells are widely used in multiple fields such as metal refining, chemical product manufacturing, and hydrogen production. The working principle of an electrolytic cell is to convert electrical energy into chemical energy through electricity to achieve oxidation-reduction reactions of substances. However, due to its complex and harsh working environment, such as high temperature, high pressure, and strong current, the electrolytic cell is prone to being affected by various factors during operation, resulting in unstable operating states and even potential safety accidents.

[0003] To ensure the safe and stable operation of electrolytic cells, real-time monitoring and state detection are particularly important. Traditional methods for detecting the operating state of electrolytic cells often rely on simple threshold judgments or the monitoring of a single parameter. This approach has a high false alarm rate and missed alarm rate, and cannot effectively detect potential safety hazards and operating abnormalities in a timely manner.

[0004] With the development of industrial automation and intelligence, introducing intelligent control theories and technologies has become an important means to improve the operation management level of electrolytic cells. Fuzzy logic controllers are widely used in various process control systems, including the detection of the operating state of electrolytic cells, because they can handle uncertainties and complexities. For example, the invention patent with the publication number CN117538391A proposes a method for detecting the operating state of an electrolytic cell. It adjusts the preset values of hydrogen in oxygen and oxygen in hydrogen according to the actual power supply parameters and the parameters of the pressure transmitter, and uses a fuzzy controller to judge the state of the electrolytic cell based on the deviation and the rate of change of the deviation between the actual collected values of hydrogen in oxygen and oxygen in hydrogen and the preset values within one collection period, improving the safety of the electrolytic cell and avoiding the occurrence of safety accidents.

[0005] However, this method still has certain limitations. On the one hand, the design of the fuzzy controller depends on expert experience and historical data. The parameter adjustment and optimization process is relatively complex and it is difficult to fully adapt to various complex changes during the operation of the electrolytic cell. On the other hand, the analysis and processing method of the collected parameters in the above solution is relatively simple. It only uses the deviation and the rate of change of the deviation between the actual collected values and the preset values for state judgment, ignoring the deep-seated temporal relationship between the parameters, which limits the comprehensive understanding of the complex physical and chemical processes inside the electrolytic cell and may lead to misjudgment or missed judgment.

[0006] Therefore, an optimized method for detecting the operating state of an electrolytic cell is expected. Summary of the Invention

[0007] The present application provides a method for detecting the operating state of an electrolytic cell, which can effectively improve the accuracy and efficiency of detecting the operating state of the electrolytic cell, help to timely discover potential problems and take corresponding measures, and ensure the safety and stability of the production process.

[0008] In a first aspect, a method for detecting the operating state of an electrolytic cell is provided, including:

[0009] Determine a preset hydrogen-in-oxygen parameter and a preset oxygen-in-hydrogen parameter;

[0010] Obtain a time queue of real-time hydrogen-in-oxygen parameters and a time queue of real-time oxygen-in-hydrogen parameters within a sampling period;

[0011] Extract time-series pattern features from the time queue of the real-time hydrogen-in-oxygen parameters and the time queue of the real-time oxygen-in-hydrogen parameters respectively to obtain a real-time hydrogen-in-oxygen parameter time-series pattern feature coding vector and a real-time oxygen-in-hydrogen parameter time-series pattern feature coding vector;

[0012] Perform fine-grained interaction based on external knowledge enhancement on the real-time hydrogen-in-oxygen parameter time-series pattern feature coding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature coding vector to obtain a real-time parameter time-series pattern collaborative interaction coding vector;

[0013] Perform semantic embedding correlation coding on the preset hydrogen-in-oxygen parameter and the preset oxygen-in-hydrogen parameter to obtain a semantic embedding correlation matrix between the preset parameters;

[0014] Based on the feature offset between the real-time parameter time-series pattern collaborative interaction coding vector and the semantic embedding correlation matrix between the preset parameters, determine whether there is an abnormality in the operating state of the electrolytic cell.

[0015] A method for detecting the operating state of an electrolytic cell provided by the present application, after determining the preset hydrogen-in-oxygen parameter and the preset oxygen-in-hydrogen parameter, collects the time queue data of the real-time hydrogen-in-oxygen parameter and the real-time oxygen-in-hydrogen parameter within a sampling period, and uses deep learning-based data processing technology to perform fine-grained time-series interaction coding on the real-time hydrogen-in-oxygen parameter and the real-time oxygen-in-hydrogen parameter to capture the time-series co-variation pattern between the parameters. Furthermore, based on the feature offset of the actual time-series co-variation pattern between the parameters relative to the preset parameters, it is used to intelligently identify whether there is an abnormality in the operating state of the electrolytic cell. In this way, the accuracy and efficiency of detecting the operating state of the electrolytic cell can be effectively improved, which helps to timely discover potential problems and take corresponding measures, and ensure the safety and stability of the production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application will be briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present application and do not limit the present application.

[0017] Figure 1 Schematic flowchart of the electrolytic cell operating state detection method according to an embodiment of the present application.

[0018] Figure 2 Schematic diagram of data flow of the electrolytic cell operating state detection method according to an embodiment of the present application.

[0019] Figure 3 Schematic flowchart of step S6 in the electrolytic cell operating state detection method according to an embodiment of the present application.

[0020] Figure 4 Schematic flowchart of step S62 in the electrolytic cell operating state detection method according to an embodiment of the present application. Detailed implementation manners

[0021] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall also fall within the scope of protection of the present application.

[0022] In view of the above technical problems, the technical concept of the present application is as follows: after determining the preset hydrogen-in-oxygen parameter and the preset oxygen-in-hydrogen parameter, collect the time queue data of the real-time hydrogen-in-oxygen parameter and the real-time oxygen-in-hydrogen parameter within a sampling period, and use the data processing technology based on deep learning to perform fine-grained temporal interaction encoding on the real-time hydrogen-in-oxygen parameter and the real-time oxygen-in-hydrogen parameter to capture the temporal co-variation pattern between the parameters. Furthermore, based on the feature offset of the actual temporal co-variation pattern between the parameters relative to the preset parameters, the operating state of the electrolytic cell is intelligently identified whether there is an abnormality. In this way, the accuracy and efficiency of the electrolytic cell operating state detection can be effectively improved, which helps to timely discover potential problems and take corresponding measures to ensure the safety and stability of the production process.

[0023] Such as Figure 1 and Figure 2As shown, the method for detecting the operating state of the electrolytic cell includes: S1, determining a preset hydrogen-in-oxygen parameter and a preset oxygen-in-hydrogen parameter; S2, obtaining a time queue of real-time hydrogen-in-oxygen parameters and a time queue of real-time oxygen-in-hydrogen parameters within a sampling period; S3, respectively performing time series pattern feature extraction on the time queue of real-time hydrogen-in-oxygen parameters and the time queue of real-time oxygen-in-hydrogen parameters to obtain a real-time hydrogen-in-oxygen parameter time series pattern feature coding vector and a real-time oxygen-in-hydrogen parameter time series pattern feature coding vector; S4, performing fine-grained interaction based on external knowledge enhancement on the real-time hydrogen-in-oxygen parameter time series pattern feature coding vector and the real-time oxygen-in-hydrogen parameter time series pattern feature coding vector to obtain a real-time parameter time series pattern collaborative interaction coding vector; S5, performing semantic embedding correlation coding on the preset hydrogen-in-oxygen parameter and the preset oxygen-in-hydrogen parameter to obtain a semantic embedding correlation matrix between preset parameters; S6, determining whether there is an abnormality in the operating state of the electrolytic cell based on the feature offset between the real-time parameter time series pattern collaborative interaction coding vector and the semantic embedding correlation matrix between preset parameters.

[0024] Exemplarily, in step S1, a preset hydrogen-in-oxygen parameter and a preset oxygen-in-hydrogen parameter are determined. It should be understood that during the operation of the electrolytic cell, the ratios of hydrogen-in-oxygen (H2 in O2) and oxygen-in-hydrogen (O2 in H2) are two key indicators, reflecting the electrolysis efficiency, product quality, and the safety state of the electrolytic cell system. The preset hydrogen-in-oxygen parameter and preset oxygen-in-hydrogen parameter are usually set based on theoretical calculations, experimental data, or historical experience, representing the parameter values that the electrolytic cell should reach under ideal or stable conditions, providing a normal operating reference value for detecting the operating state of the electrolytic cell.

[0025] Exemplarily, in step S2, a time queue of real-time hydrogen-in-oxygen parameters and a time queue of real-time oxygen-in-hydrogen parameters within a sampling period are obtained. It should be understood that in practical applications, due to factors such as electrolytic cell aging, external interference, and material consumption, the actual hydrogen-in-oxygen parameter and oxygen-in-hydrogen parameter usually deviate from the preset values, which may be normal fluctuations or signals of abnormal operation of the electrolytic cell. Therefore, this application further obtains a time queue of real-time hydrogen-in-oxygen parameters and a time queue of real-time oxygen-in-hydrogen parameters within a sampling period, and evaluates whether there is an abnormality in the current operating state of the electrolytic cell by performing time series comparison analysis on the hydrogen-in-oxygen parameter and oxygen-in-hydrogen parameter in the actual operation process of the electrolytic cell with the preset values.

[0026] Exemplarily, in step S3, temporal pattern feature extraction is respectively performed on the time queue of the real-time hydrogen-in-oxygen parameter and the time queue of the real-time oxygen-in-hydrogen parameter to obtain a real-time hydrogen-in-oxygen parameter temporal pattern feature encoding vector and a real-time oxygen-in-hydrogen parameter temporal pattern feature encoding vector. It should be understood that considering the dynamic changes in hydrogen-in-oxygen and oxygen-in-hydrogen generated during the operation of the electrolytic cell over time, therefore, in order to fully capture the temporal dynamic change patterns of the real-time hydrogen-in-oxygen parameter and the real-time oxygen-in-hydrogen parameter, the present application uses a bidirectional LSTM model with excellent performance in time series analysis as a temporal pattern feature extractor to respectively perform bidirectional temporal encoding on the time queue of the real-time hydrogen-in-oxygen parameter and the time queue of the real-time oxygen-in-hydrogen parameter, so as to utilize the bidirectional information transmission mechanism of the bidirectional LSTM model and consider the long-distance dependence information of the forward and reverse time series simultaneously, thereby more accurately capturing the dynamic trend and temporal pattern of the parameter changing with time, generating a real-time hydrogen-in-oxygen parameter temporal pattern feature encoding vector and a real-time oxygen-in-hydrogen parameter temporal pattern feature encoding vector, and providing strong support for the accurate evaluation of the operating state of the electrolytic cell.

[0027] That is, performing temporal pattern feature extraction on the time queue of the real-time hydrogen-in-oxygen parameter and the time queue of the real-time oxygen-in-hydrogen parameter respectively to obtain a real-time hydrogen-in-oxygen parameter temporal pattern feature encoding vector and a real-time oxygen-in-hydrogen parameter temporal pattern feature encoding vector includes: respectively inputting the time queue of the real-time hydrogen-in-oxygen parameter and the time queue of the real-time oxygen-in-hydrogen parameter into a temporal pattern feature extractor based on a bidirectional LSTM model to obtain the real-time hydrogen-in-oxygen parameter temporal pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter temporal pattern feature encoding vector.

[0028] Exemplarily, in step S4, a fine-grained interaction based on external knowledge enhancement is performed on the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector to obtain a real-time parameter time-series pattern collaborative interaction encoding vector. It should be understood that in the hydrogen-oxygen reaction, hydrogen and oxygen are reactants, and the two can undergo a chemical reaction to form water under certain conditions. Therefore, in the hydrogen-oxygen reaction system, the changes in the oxygen-in-hydrogen parameters and the hydrogen-in-oxygen parameters affect each other. For example, when the oxygen content in hydrogen increases, it may promote the occurrence of the hydrogen-oxygen reaction, resulting in a decrease in the hydrogen content in oxygen. That is to say, there is a certain correlation between the oxygen-in-hydrogen parameters and the hydrogen-in-oxygen parameters. Based on this, in order to effectively capture the co-variation pattern of the real-time hydrogen-in-oxygen parameters and the real-time oxygen-in-hydrogen parameters in the time dimension, the present application further performs a time-series fine-grained interaction analysis on the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector. In particular, in order to further enhance the time-series interaction effect of the real-time hydrogen-in-oxygen parameters and the real-time oxygen-in-hydrogen parameters, the present application proposes a fine-grained interaction method based on external knowledge enhancement, which introduces external knowledge to guide the interaction process between parameters, so as to enhance the model's understanding and modeling ability of the complex relationship between parameters, thereby further improving the detection accuracy of the electrolytic cell operating state.

[0029] In one embodiment, as Figure 3 shown, performing a fine-grained interaction based on external knowledge enhancement on the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector to obtain a real-time parameter time-series pattern collaborative interaction encoding vector includes: S41, based on external knowledge, performing feature interaction attention optimization on the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector to obtain an externally knowledge-optimized real-time parameter time-series fine-grained interaction feature matrix; S42, based on the externally knowledge-optimized real-time parameter time-series fine-grained interaction feature matrix, performing feature modulation optimization on the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector respectively to obtain an optimized real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and an optimized real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector; S43, performing a position-by-position time-series collaborative interaction encoding on the optimized real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the optimized real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector to obtain the real-time parameter time-series pattern collaborative interaction encoding vector.

[0030] In one embodiment, in step S41, based on external knowledge, feature interaction attention optimization is performed on the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector to obtain an externally knowledge-optimized real-time parameter inter-temporal fine-grained interaction feature matrix, including: inputting the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector into a fine-grained feature interaction network to obtain a real-time parameter inter-temporal fine-grained interaction feature matrix; and inputting the real-time parameter inter-temporal fine-grained interaction feature matrix into an attention unit based on external knowledge to obtain the externally knowledge-optimized real-time parameter inter-temporal fine-grained interaction feature matrix. Specifically, this process can be expressed by the formula:

[0031]

[0032]

[0033] where V1 represents the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector, V2 represents the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector, (·) T represents the transpose of a vector, represents matrix multiplication operation, M p represents the real-time parameter inter-temporal fine-grained interaction feature matrix, M k and M v represent the learnable memory parameter matrices of the attention unit based on external knowledge, norm(·) represents a normalization function, and M y represents the externally knowledge-optimized real-time parameter inter-temporal fine-grained interaction feature matrix.

[0034] It should be understood that during the operation of the electrolytic cell, the ratios of hydrogen in oxygen (H2 in O2) and oxygen in hydrogen (O2 in H2) change dynamically over time. This change not only reflects the electrolysis efficiency and product quality but also relates to the safety state of the system. The fine-grained feature interaction network is used to process the time-series data of these two parameters, namely the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector. Through the fine-grained feature interaction network, the interaction between the two parameters at different time points can be deeply explored, and the potential complex relationships between them can be identified. For example, in some cases, if the oxygen content in hydrogen increases, it may accelerate the hydrogen-oxygen reaction, thereby affecting the change in the hydrogen content in oxygen. This process can help the present application understand the physical and chemical processes occurring inside the electrolytic cell and provide a basis for subsequent state assessment.

[0035] Next, introducing external knowledge (such as electrolyzer operation data, chemical reaction principles, etc.) as guidance can enable the model to better understand the correlations and causal relationships between parameters. The attention unit based on external knowledge can be regarded as a mechanism that allows the model to adjust its attention to the interactions between parameters at different time points according to known professional knowledge. When the present application sends the real-time parameter temporal fine-grained interaction feature matrix into the attention unit based on external knowledge, this unit will use pre-trained parameters (optimized by external knowledge) to weight the elements in the feature matrix. This means that the model will pay more attention to those interaction patterns that are considered more important or indicative; while those that are less relevant or noise components will be appropriately weakened. This external knowledge-based attention optimization makes the finally obtained external knowledge-optimized real-time parameter temporal fine-grained interaction feature matrix more focused on the key information that truly reflects the changes in the electrolyzer operation state. This not only improves the sensitivity of the model to abnormal situations but also helps to reduce the false alarm rate and missed alarm rate, ensuring the accuracy and reliability of the electrolyzer operation state monitoring.

[0036] That is, by performing fine-grained feature interactions on the real-time hydrogen-in-oxygen parameter temporal pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter temporal pattern feature encoding vector, the temporal correlation between the two is understood, the fine-grained interaction information between the two is captured, and a real-time parameter temporal fine-grained interaction feature matrix is generated. Subsequently, the generated real-time parameter temporal fine-grained interaction feature matrix is sent into the attention unit based on external knowledge, and external knowledge is used to further optimize the real-time parameter temporal fine-grained interaction feature matrix, thereby enhancing the model's ability to recognize and understand the temporal correlation relationship between parameters. In a specific implementation, electrolyzer operation data and chemical reaction principles are used as external knowledge to train the parameter matrix of the attention unit, guiding the attention unit to perform weighted optimization on the real-time parameter temporal fine-grained interaction feature matrix, so as to identify and strengthen the important temporal interaction patterns between parameters.

[0037] In one embodiment, in step S42, based on the external knowledge, the fine-grained interaction feature matrix between real-time parameters is optimized, and the feature modulation optimization is respectively performed on the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector to obtain the optimized real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the optimized real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector, including: performing a linear transformation on the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector to obtain a first query feature vector and a first value feature vector, and using the fine-grained interaction feature matrix between real-time parameters optimized by the external knowledge as a key matrix, and inputting the first query feature vector, the first value feature vector, and the key matrix into the fine-grained modulation module based on the Transformer structure to obtain the optimized real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector; performing a linear transformation on the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector to obtain a second query feature vector and a second value feature vector, and using the fine-grained interaction feature matrix between real-time parameters optimized by the external knowledge as a key matrix, and inputting the second query feature vector, the second value feature vector, and the key matrix into the fine-grained modulation module based on the Transformer structure to obtain the optimized real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector. Specifically, this process can be expressed by the formula as follows:

[0038]

[0039] Among them, W 1q and W 1v respectively represent the first query embedding matrix and the first value embedding matrix, V 1q and V 1v respectively represent the first query feature vector and the first value feature vector, W 2q and W 2v respectively represent the second query embedding matrix and the second value embedding matrix, V 2q and V 2v respectively represent the second query feature vector and the second value feature vector, b 1q 、b 1v 、b 2q and b 2v respectively represent different bias terms, softmax(·) represents the normalized exponential function, d represents the feature scale value of the fine-grained feature interaction matrix optimized by the external knowledge, and V′1 and V′2 respectively represent the optimized real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the optimized real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector.

[0040] It should be understood that when processing the time - series pattern feature encoding vector of hydrogen in oxygen in real - time, the present application first performs a linear transformation on it to generate a first query feature vector and a first value feature vector. At the same time, the externally - known - optimized time - series fine - grained interaction feature matrix between real - time parameters obtained previously is used as the key matrix. Then, these three elements - the first query feature vector, the first value feature vector, and the key matrix - are input into the fine - grained modulation module based on the Transformer structure together. Through this mechanism, the module can utilize the self - attention mechanism to achieve information exchange and integration between internal features and external knowledge, ensuring that the time - series pattern feature encoding vector of hydrogen in oxygen in real - time can benefit from external knowledge, thereby improving the quality of its feature representation. Finally, this process outputs the optimized time - series pattern feature encoding vector of hydrogen in oxygen in real - time. For the time - series pattern feature encoding vector of oxygen in hydrogen in real - time, the present application adopts a similar processing method: also performing a linear transformation to obtain a second query feature vector and a second value feature vector, and using the same externally - known - optimized time - series fine - grained interaction feature matrix between real - time parameters as the key matrix. Subsequently, these elements are also fed into the fine - grained modulation module based on the Transformer structure. After the same information exchange and integration process, the optimized time - series pattern feature encoding vector of oxygen in hydrogen in real - time is obtained. Throughout the process, the fine - grained modulation module based on the Transformer structure plays a crucial role. It allows the present application to more precisely capture the dynamic trends of the hydrogen - in - oxygen parameters and oxygen - in - hydrogen parameters in real - time over time and their mutual influences. For example, during the operation of the electrolytic cell, if the oxygen content in hydrogen changes, it may cause a corresponding adjustment in the hydrogen content in oxygen. By introducing external knowledge (such as electrolytic cell operation data and chemical reaction principles), the present application can understand the reasons behind such changes at a deeper level and accurately evaluate whether these changes indicate potential problems or abnormal situations. In addition, this method also enhances the model's ability to understand the complex relationships between parameters, enabling the present application to more effectively identify the key signals that may indicate unstable operation of the electrolytic cell, thereby helping operators take timely measures to ensure the safety and stability of the production process.

[0041] That is, the time-series fine-grained interaction feature matrix between real-time parameters optimized with external knowledge is used as the key matrix. At the same time, a first query feature vector and a first value feature vector are constructed based on the time-series pattern feature encoding vector of the real-time hydrogen-in-oxygen parameter, and a second query feature vector and a second value feature vector are constructed based on the time-series pattern feature encoding vector of the real-time oxygen-in-hydrogen parameter. The self-attention mechanism of the Transformer structure is used to achieve information exchange and integration between internal features and external knowledge, so as to ensure that the time-series pattern feature encoding vector of the real-time hydrogen-in-oxygen parameter and the time-series pattern feature encoding vector of the real-time oxygen-in-hydrogen parameter can benefit from external knowledge and improve the quality of their feature expressions.

[0042] In one embodiment, in step S43, the optimized time-series pattern feature encoding vector of the real-time hydrogen-in-oxygen parameter and the optimized time-series pattern feature encoding vector of the real-time oxygen-in-hydrogen parameter are subjected to per-position time-series collaborative interaction encoding to obtain the time-series pattern collaborative interaction encoding vector between the real-time parameters, including: calculating the point-by-position division between the optimized time-series pattern feature encoding vector of the real-time hydrogen-in-oxygen parameter and the optimized time-series pattern feature encoding vector of the real-time oxygen-in-hydrogen parameter to obtain the time-series pattern collaborative interaction encoding vector between the real-time parameters. Specifically, this process can be expressed by the formula:

[0043]

[0044] where V′1 and V′2 respectively represent the optimized time-series pattern feature encoding vector of the real-time hydrogen-in-oxygen parameter and the optimized time-series pattern feature encoding vector of the real-time oxygen-in-hydrogen parameter, and V i represents the time-series pattern collaborative interaction encoding vector between the real-time parameters.

[0045] It should be understood that by performing point-by-position division on the optimized time-series pattern feature encoding vector of the real-time hydrogen-in-oxygen parameter and the optimized time-series pattern feature encoding vector of the real-time oxygen-in-hydrogen parameter, the present application can capture the relative change ratio of these two parameters at each time point. This operation effectively reflects the interaction intensity and pattern between the two gas components at different operating stages of the electrolytic cell. For example, during the electrolysis process, if the hydrogen content in oxygen at a certain moment changes significantly relative to its normal level, and the oxygen content in hydrogen at the same moment is also adjusted accordingly, then through the point division operation, the present application can quantify this change ratio relationship and add it as a new feature to the time-series pattern collaborative interaction encoding vector between the real-time parameters.

[0046] The real-time parameter temporal pattern collaborative interaction coding vector not only retains the time series information of the original parameters but also introduces the dimension of the relative change between two parameters, which provides richer context for subsequent analysis. Specifically, this coding vector can help identify critical moments or trends that have a significant impact on the operating state of the electrolytic cell. For example, an abnormally high or low interaction ratio may indicate changes in system efficiency, potential safety risks, or early signs of equipment failure.

[0047] Exemplarily, in step S5, semantic embedding correlation coding is performed on the preset hydrogen-in-oxygen parameter and the preset oxygen-in-hydrogen parameter to obtain a semantic embedding correlation matrix between the preset parameters. It should be understood that in order to realize the identification and detection of the operating state of the electrolytic cell, it is necessary to further perform a comparative analysis on the real-time operating parameter data and the preset parameter data. Based on this, first, it is necessary to perform a feature space transformation on the preset hydrogen-in-oxygen parameter and the preset oxygen-in-hydrogen parameter, and at the same time establish a semantic association representation between the preset hydrogen-in-oxygen parameter and the preset oxygen-in-hydrogen parameter, so as to perform a comparative analysis with the real-time parameter temporal pattern collaborative interaction coding vector. Therefore, in this application, the preset hydrogen-in-oxygen parameter and the preset oxygen-in-hydrogen parameter are first mapped to a unified low-dimensional embedding feature space through an embedding coding technique respectively to achieve the alignment of the feature dimensions, and a preset hydrogen-in-oxygen parameter embedding coding vector and a preset oxygen-in-hydrogen parameter embedding coding vector are generated. Then, by calculating the fine-grained correlation matrix between the preset hydrogen-in-oxygen parameter embedding coding vector and the preset oxygen-in-hydrogen parameter embedding coding vector, a semantic association representation between the preset parameters is constructed, and a semantic embedding correlation matrix between the preset parameters is generated. In a specific example of this application, the semantic embedding correlation matrix between the preset parameters is obtained by calculating the product of the preset hydrogen-in-oxygen parameter embedding coding vector and the transposed vector of the preset oxygen-in-hydrogen parameter embedding coding vector.

[0048] In one embodiment, performing semantic embedding correlation coding on the preset hydrogen-in-oxygen parameter and the preset oxygen-in-hydrogen parameter to obtain a semantic embedding correlation matrix between the preset parameters includes: after respectively low-dimensionally embedding and coding the preset hydrogen-in-oxygen parameter and the preset oxygen-in-hydrogen parameter into a preset hydrogen-in-oxygen parameter embedding coding vector and a preset oxygen-in-hydrogen parameter embedding coding vector, calculating the fine-grained correlation matrix between the preset hydrogen-in-oxygen parameter embedding coding vector and the preset oxygen-in-hydrogen parameter embedding coding vector to obtain the semantic embedding correlation matrix between the preset parameters.

[0049] Exemplarily, in step S6, based on the feature offset between the temporal pattern collaborative interaction coding vector among the real-time parameters and the semantic embedding correlation matrix among the preset parameters, it is determined whether there is an abnormality in the operating state of the electrolytic cell. It should be understood that the feature offset metric allows the present application to quantify the degree of change of the real-time parameters relative to the preset parameters. Such a change may be a normal fluctuation or a signal of abnormal operation. For example, if the hydrogen content in oxygen suddenly increases significantly within a certain period of time, and the oxygen content in hydrogen also changes abnormally accordingly, this may indicate that an undesired chemical reaction or other problems have occurred inside the electrolytic cell. By calculating the feature offset between the two, the present application can timely capture such abnormal changes, and then identify potential problems or abnormal situations. The physical and chemical processes occurring inside the electrolytic cell are very complex, and the change of a single parameter often cannot comprehensively reflect the true state of the system. By constructing the temporal pattern collaborative interaction coding vector among the real-time parameters, the present application not only considers the change of a single parameter, but also considers the interaction and co-variation patterns among different parameters. This multi-dimensional perspective helps to more deeply understand the dynamic behavior during the operation of the electrolytic cell, and by comparing with the semantic embedding correlation matrix among the preset parameters, it can more accurately determine which changes are normal and which may be abnormal. Finally, by using the method based on feature offset to determine whether there is an abnormality in the operating state of the electrolytic cell, the present application realizes the intelligent monitoring of the operating state of the electrolytic cell. This method can not only help operators timely discover potential safety hazards and abnormal operations, but also guide them to take corresponding preventive measures or corrective actions to ensure the safety and stability of the production process. In summary, this method effectively improves the accuracy of detecting the operating state of the electrolytic cell and provides strong support for industrial production.

[0050] In one embodiment, as Figure 4 shown, based on the feature offset between the temporal pattern collaborative interaction coding vector among the real-time parameters and the semantic embedding correlation matrix among the preset parameters, determining whether there is an abnormality in the operating state of the electrolytic cell includes: S61, inputting the temporal pattern collaborative interaction coding vector among the real-time parameters and the semantic embedding correlation matrix among the preset parameters into a parameter temporal offset metric network based on a transition matrix to obtain a real-time parameter - preset parameter temporal offset transition coding vector; S62, inputting the real-time parameter - preset parameter temporal offset transition coding vector into a state recognizer based on a classifier to obtain a recognition result, and the recognition result is used to indicate whether there is an abnormality in the operating state of the electrolytic cell.

[0051] In one embodiment, in step S61, inputting the real-time parameter inter-temporal pattern collaborative interaction coding vector and the preset parameter semantic embedding correlation matrix into the parameter temporal offset metric network based on the transition matrix to obtain the real-time parameter - preset parameter temporal offset transition coding vector, includes: calculating the product of the real-time parameter inter-temporal pattern collaborative interaction coding vector and the inverse matrix of the preset parameter semantic embedding correlation matrix to obtain the real-time parameter - preset parameter temporal offset transition coding vector. That is, by performing feature transfer calculation on the real-time parameter inter-temporal pattern collaborative interaction coding vector and the preset parameter semantic embedding correlation matrix, to quantify the state transition degree of the real-time parameter relative to the preset parameter, understand the temporal change path of the real-time parameter, so as to obtain the real-time parameter - preset parameter temporal offset transition coding vector, to reveal the deviation degree and change trend of the electrolytic cell operating state relative to the preset standard, and provide data support for the subsequent evaluation of the electrolytic cell operating state.

[0052] In step S62, inputting the real-time parameter - preset parameter temporal offset transition coding vector into the state recognizer based on the classifier to obtain the recognition result, where the recognition result is used to indicate whether there is an abnormality in the electrolytic cell operating state. Further, in a specific embodiment, the state recognizer based on the classifier uses a softmax classifier for classification. The softmax classifier is a machine learning model widely used in multi-class classification problems. It can map the input features to the probability distributions of each class, thereby helping the present application to judge whether the operating state of the electrolytic cell is normal or abnormal. Specifically, when the present application takes the real-time parameter - preset parameter temporal offset transition coding vector as the input and passes it to the softmax classifier, the classifier will calculate the scores or logits corresponding to each possible class (such as "normal" and "abnormal"). These scores reflect the likelihood of the input data belonging to each class. Subsequently, these raw scores are converted into a normalized probability distribution through the softmax function, where all probability values add up to 1. The advantage of this is that the output result has an intuitive probability interpretation, that is, it can clearly show how likely it is to be judged as normal or abnormal under a specific operating state. Finally, select the class corresponding to the larger probability as the recognition result.

[0053] Here, when the real-time parameter inter-temporal mode collaborative interaction coding vector and the preset parameter semantic embedding correlation matrix respectively represent the temporal semantic correlation features of the real-time hydrogen-in-oxygen parameter and the real-time oxygen-in-hydrogen parameter, and the low-dimensional embedding semantic correlation features of the preset hydrogen-in-oxygen parameter and the preset oxygen-in-hydrogen parameter, when they are input into the parameter temporal offset metric network based on the transfer matrix, the real-time parameter - preset parameter temporal offset transfer coding vector will also have significant temporal transfer feature richness due to inconsistent feature orders and feature scales, resulting in repeated unstructured probability mapping, thus affecting the accuracy of the recognition result obtained by the classifier-based state recognizer.

[0054] To address the above technical problems, in a specific example of the present application, before inputting the real-time parameter - preset parameter temporal offset transfer coding vector into the classifier-based state recognizer, the real-time parameter - preset parameter temporal offset transfer coding vector is subjected to structured modulation, where the structured modulation process includes the following steps:

[0055] Based on the absolute value magnitudes of the feature values at each position in the real-time parameter - preset parameter temporal offset transfer coding vector, the real-time parameter - preset parameter temporal offset transfer coding vector is subjected to normalization modulation to obtain a real-time parameter - preset parameter temporal offset transfer normalized coding vector;

[0056] Perform autocorrelation collaborative quantization analysis on the real-time parameter - preset parameter temporal offset transfer normalized coding vector to obtain a real-time parameter - preset parameter temporal offset transfer collaborative matrix, expressed as:

[0057]

[0058] where W c represents the pre-trained weight matrix, B c represents the pre-trained bias matrix, V order represents the real-time parameter - preset parameter temporal offset transfer normalized coding vector, T represents the transpose symbol, L represents the length of the real-time parameter - preset parameter temporal offset transfer normalized coding vector, ⊙ represents pointwise multiplication by position, represents matrix multiplication, M c represents the real-time parameter - preset parameter temporal offset transfer collaborative matrix.

[0059] Perform feature independent activation on the real-time parameter - preset parameter temporal offset transfer collaborative matrix to obtain a real-time parameter - preset parameter temporal offset transfer collaborative activation expression matrix, expressed as:

[0060] M j = Sigmoid(Mc )

[0061] Among them, Sigmoid represents the activation function, and M j represents the collaborative activation expression matrix for the time - sequence offset transfer between real - time parameters and preset parameters.

[0062] Calculate the spatial - structure eigen - element modulation weights of the time - sequence offset transfer coding vector between real - time parameters and preset parameters, and perform structure modulation on the collaborative activation expression matrix for the time - sequence offset transfer between real - time parameters and preset parameters based on the spatial - structure eigen - element modulation weights to obtain the collaborative structure adaptive modulation matrix for the time - sequence offset transfer between real - time parameters and preset parameters, expressed as:

[0063]

[0064] M a = M j * root

[0065] Among them, v1, v2, and v n represent the eigenvalues at the first, second, and n - th positions of the time - sequence offset transfer coding vector between real - time parameters and preset parameters, root represents the spatial - structure eigen - element modulation weights, * represents multiplying each eigenvalue of the matrix by the spatial - structure eigen - element modulation weights, and M a represents the collaborative structure adaptive modulation matrix for the time - sequence offset transfer between real - time parameters and preset parameters.

[0066] Using the time - sequence offset transfer coding vector between real - time parameters and preset parameters as the query vector, structurally map it to the collaborative structure adaptive modulation matrix for the time - sequence offset transfer between real - time parameters and preset parameters to obtain the optimized time - sequence offset transfer coding vector between real - time parameters and preset parameters, expressed as:

[0067]

[0068] Among them, V ’ represents the optimized time - sequence offset transfer coding vector between real - time parameters and preset parameters.

[0069] Accordingly, by mapping the normalized signal of the time series offset transfer encoding vector between real-time parameters and preset parameters into the neuron link architecture configured based on connection strength, a quantization analysis benchmark of the activation correlation of the neural pattern set in the cooperation matrix can be realized. Based on this, a neural activity space with independent activation channels can be constructed through the neural adaptation regulation mechanism, and a synaptic weight clipping strategy of the multi-layer neural network of neuron responses based on signal query can be adopted to avoid weakening the intrinsic fidelity of neural conduction due to noise interference. In this way, the invalid repetition in the probability mapping process caused by the unstructured characteristics of the time series offset transfer encoding vector between real-time parameters and preset parameters can be effectively avoided, and the accuracy of the recognition result obtained by the time series offset transfer encoding vector between real-time parameters and preset parameters through the classifier-based state recognizer can be improved.

[0070] In summary, the electrolytic cell operating state detection method according to the embodiments of the present application is clarified. After determining the preset hydrogen-in-oxygen parameter and the preset oxygen-in-hydrogen parameter, the time queue data of the real-time hydrogen-in-oxygen parameter and the real-time oxygen-in-hydrogen parameter within a sampling period is collected, and the fine-grained time series interaction encoding is performed on the real-time hydrogen-in-oxygen parameter and the real-time oxygen-in-hydrogen parameter by using the data processing technology based on deep learning to capture the time series co-variation pattern between the parameters. Furthermore, based on the characteristic offset of the actual time series co-variation pattern between the parameters relative to the preset parameters, the abnormal state of the electrolytic cell operation can be intelligently identified. In this way, the accuracy and efficiency of the electrolytic cell operating state detection can be effectively improved, which helps to timely discover potential problems and take corresponding measures to ensure the safety and stability of the production process.

[0071] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0072] It should be understood that the specific examples herein are only for helping those skilled in the art better understand the embodiments of the present application, rather than limiting the scope of the embodiments of the present application.

[0073] It should also be understood that in various embodiments of the present application, the magnitude of the sequence number of each process does not mean the order of execution, and the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0074] It should also be understood that the various embodiments described in this specification can be implemented either alone or in combination, and the embodiments of the present application do not limit this.

[0075] Unless otherwise specified, all technical and scientific terms used in the embodiments of the present application have the same meaning as commonly understood by those skilled in the technical field of the present application. The terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the scope of the present application. The term "and / or" used in the present application includes any and all combinations of one or more of the related listed items. The singular forms "a", "above-mentioned", and "the" used in the embodiments of the present application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0076] In several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0077] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0078] In addition, the functional units in the various embodiments of the present application can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0079] When the above-mentioned function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0080] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed by this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A method for detecting the operating state of an electrolytic cell, characterized in that, Including: Determine the preset hydrogen-in-oxygen parameter and the preset oxygen-in-hydrogen parameter; Obtain the time queue of the real-time hydrogen-in-oxygen parameter and the time queue of the real-time oxygen-in-hydrogen parameter within a sampling period; Extract the time-series pattern features from the time queue of the real-time hydrogen-in-oxygen parameter and the time queue of the real-time oxygen-in-hydrogen parameter respectively to obtain the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector; Perform fine-grained interaction based on external knowledge enhancement on the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector to obtain the real-time parameter time-series pattern collaborative interaction encoding vector; Perform semantic embedding correlation encoding on the preset hydrogen-in-oxygen parameter and the preset oxygen-in-hydrogen parameter to obtain the semantic embedding correlation matrix between the preset parameters; Determine whether there is an abnormality in the operation state of the electrolytic cell based on the feature offset between the real-time parameter time-series pattern collaborative interaction encoding vector and the semantic embedding correlation matrix between the preset parameters.

2. The method for detecting the operating state of an electrolytic cell according to claim 1, wherein, Extract the time-series pattern features from the time queue of the real-time hydrogen-in-oxygen parameter and the time queue of the real-time oxygen-in-hydrogen parameter respectively to obtain the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector, including: Input the time queue of the real-time hydrogen-in-oxygen parameter and the time queue of the real-time oxygen-in-hydrogen parameter into the time-series pattern feature extractor based on the bidirectional LSTM model respectively to obtain the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector.

3. The method for detecting the operating state of an electrolytic cell according to claim 2, characterized in that, Perform fine-grained interaction based on external knowledge enhancement on the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector to obtain the real-time parameter time-series pattern collaborative interaction encoding vector, including: Based on external knowledge, perform feature interaction attention optimization on the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector to obtain the external knowledge optimized real-time parameter time-series fine-grained interaction feature matrix; Based on the external knowledge optimized real-time parameter time-series fine-grained interaction feature matrix, perform feature modulation optimization on the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector respectively to obtain the optimized real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the optimized real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector; Perform per-position time-series collaborative interaction encoding on the optimized real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the optimized real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector to obtain the real-time parameter time-series pattern collaborative interaction encoding vector.

4. The electrolytic cell operating state detection method according to claim 3, characterized in that, Based on external knowledge, perform feature interaction attention optimization on the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector to obtain the external knowledge optimized real-time parameter time-series fine-grained interaction feature matrix, including: Input the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector into the fine-grained feature interaction network to obtain the real-time parameter inter-time-series fine-grained interaction feature matrix; Input the real-time parameter inter-time-series fine-grained interaction feature matrix into the external knowledge-based attention unit to obtain the external knowledge-optimized real-time parameter inter-time-series fine-grained interaction feature matrix.

5. The method for detecting the operating state of an electrolytic cell according to claim 4, wherein Based on the external knowledge-optimized real-time parameter inter-time-series fine-grained interaction feature matrix, perform feature modulation optimization on the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector respectively to obtain the optimized real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the optimized real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector, including: Perform a linear transformation on the real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector to obtain a first query feature vector and a first value feature vector, and use the external knowledge-optimized real-time parameter inter-time-series fine-grained interaction feature matrix as the key matrix. Input the first query feature vector, the first value feature vector, and the key matrix into the fine-grained modulation module based on the Transformer structure to obtain the optimized real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector; Perform a linear transformation on the real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector to obtain a second query feature vector and a second value feature vector, and use the external knowledge-optimized real-time parameter inter-time-series fine-grained interaction feature matrix as the key matrix. Input the second query feature vector, the second value feature vector, and the key matrix into the fine-grained modulation module based on the Transformer structure to obtain the optimized real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector.

6. The method for detecting the operating state of an electrolytic cell according to claim 5, characterized in that, Perform per-position time-series collaborative interaction encoding on the optimized real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the optimized real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector to obtain the real-time parameter inter-time-series pattern collaborative interaction encoding vector, including: Calculate the element-wise division between the optimized real-time hydrogen-in-oxygen parameter time-series pattern feature encoding vector and the optimized real-time oxygen-in-hydrogen parameter time-series pattern feature encoding vector to obtain the real-time parameter inter-time-series pattern collaborative interaction encoding vector.

7. The method for detecting the operating state of an electrolytic cell according to claim 6, wherein, Perform semantic embedding correlation encoding on the preset hydrogen-in-oxygen parameter and the preset oxygen-in-hydrogen parameter to obtain the preset parameter inter-semantic embedding correlation matrix, including: After respectively low-dimensionally embedding and encoding the preset hydrogen-in-oxygen parameter and the preset oxygen-in-hydrogen parameter into a preset hydrogen-in-oxygen parameter embedding encoding vector and a preset oxygen-in-hydrogen parameter embedding encoding vector, calculate the fine-grained correlation matrix between the preset hydrogen-in-oxygen parameter embedding encoding vector and the preset oxygen-in-hydrogen parameter embedding encoding vector to obtain the preset parameter inter-semantic embedding correlation matrix.

8. The method for detecting the operating state of an electrolytic cell according to claim 7, wherein Based on the feature offset between the real-time parameter inter-time-series pattern collaborative interaction encoding vector and the preset parameter inter-semantic embedding correlation matrix, determine whether there is an abnormality in the operating state of the electrolytic cell, including: Input the temporal pattern collaborative interaction coding vector between the real-time parameters and the semantic embedding correlation matrix between the preset parameters into the parameter temporal offset metric network based on the transfer matrix to obtain the temporal offset transfer coding vector between the real-time parameters and the preset parameters; Input the temporal offset transfer coding vector between the real-time parameters and the preset parameters into the state recognizer based on the classifier to obtain the recognition result, which is used to indicate whether there is an abnormality in the operation state of the electrolytic cell.

9. The method for detecting the operating state of an electrolytic cell according to claim 8, wherein Input the temporal pattern collaborative interaction coding vector between the real-time parameters and the semantic embedding correlation matrix between the preset parameters into the parameter temporal offset metric network based on the transfer matrix to obtain the temporal offset transfer coding vector between the real-time parameters and the preset parameters, including: Calculate the product of the temporal pattern collaborative interaction coding vector between the real-time parameters and the inverse matrix of the semantic embedding correlation matrix between the preset parameters to obtain the temporal offset transfer coding vector between the real-time parameters and the preset parameters.

Images