Fault prediction method and device of electrolytic hydrogen production system, electronic equipment and storage medium

Abstract

The invention discloses a fault prediction method and device for an electrolytic hydrogen production system, electronic equipment and a storage medium. The method comprises the steps that real-time operation data of the electrolytic hydrogen production system are obtained through a sensor in the electrolytic hydrogen production system; determining key physical parameter data of the electrolytic hydrogen production system through a preset physical model based on the real-time operation data; based on the key physical parameter data and fault associated data in the real-time operation data, a pre-trained CNN-LSTM model is utilized to output a fault prediction result; and when the fault prediction result satisfies a preset condition, sending out alarm information. According to the technical scheme of the invention, active prevention of faults can be effectively realized.

Description

Technical Field

[0001] The present invention relates to the technical field of electrolytic hydrogen production, and in particular to a fault prediction method, device, electronic equipment and storage medium for an electrolytic hydrogen production system. Background Art

[0002] At present, traditional methods for health monitoring of hydrogen electrolysis systems mainly include regular inspections and daily maintenance plans. These methods have some limitations: 1. Most faults are not discovered until they occur, so maintenance is a passive approach, and it is difficult to proactively discover problems and prevent faults; 2. Frequent manual inspections and maintenance may consume a lot of resources and costs.

[0003] Therefore, improvements are needed to at least partially solve the above problems. Summary of the invention

[0004] A series of simplified concepts are introduced in the summary of the invention, which will be further described in detail in the specific embodiments. The summary of the invention does not mean to attempt to define the key features and essential technical features of the technical solution claimed for protection, nor does it mean to attempt to determine the scope of protection of the technical solution claimed for protection.

[0005] In order to at least partially solve the above technical problems, according to a first aspect of the present invention, a fault prediction method for a hydrogen production system by electrolysis is provided, which comprises:

[0006] Acquiring real-time operation data of the hydrogen production system by means of sensors in the hydrogen production system by electrolysis;

[0007] Based on the real-time operation data, key physical parameter data of the electrolysis hydrogen production system is determined by a preset physical model;

[0008] Based on the key physical parameter data and the fault-related data in the real-time operation data, a pre-trained CNN-LSTM model is used to output a fault prediction result;

[0009] When the fault prediction result meets the preset conditions, an alarm message is issued.

[0010] Optionally, the physical model includes an electrochemical model and a thermodynamic model.

[0011] Optionally, the key physical parameter data include energy conversion efficiency data of an electrolyzer in the electrolytic hydrogen production system, electrolyte concentration change data, and local temperature gradient data in the electrolyzer.

[0012] Optionally, the fault-related data in the real-time operation data is determined in the following manner:

[0013] Cleaning the historical operation data of the electrolysis hydrogen production system;

[0014] Performing feature association analysis, feature extraction and feature selection on the cleaned historical operation data to determine initial fault association data;

[0015] Using the initial fault association data for training and verification of the CNN-LSTM model to adjust and optimize the initial fault association data to obtain the final fault association data;

[0016] The historical operation data includes operation data acquired by sensors in the hydrogen production system by electrolysis before and after the fault occurs.

[0017] Optionally, performing feature association analysis on the cleaned historical operation data includes:

[0018] Using the Pearson correlation coefficient to evaluate the correlation between the historical operating data and the occurrence of faults; and / or,

[0019] The historical operating data associated with the fault is identified by a dimensionality reduction method.

[0020] Optionally, the feature extraction includes wavelet transform and / or time-frequency analysis;

[0021] Optionally, the feature selection comprises recursive feature elimination and / or tree-based feature selection.

[0022] According to a second aspect of the present invention, a fault prediction device for an electrolytic hydrogen production system is provided, comprising:

[0023] An acquisition unit, used for acquiring real-time operation data of the hydrogen production system by electrolysis through sensors in the hydrogen production system by electrolysis;

[0024] A determination unit, configured to determine key physical parameter data of the electrolysis hydrogen production system through a preset physical model based on the real-time operation data;

[0025] An output unit, configured to output a fault prediction result using a pre-trained CNN-LSTM model based on the key physical parameter data and the fault-related data in the real-time operation data;

[0026] The alarm unit is used to issue an alarm message when the fault prediction result meets the preset conditions.

[0027] According to a third aspect of the present invention, there is provided an electronic device, comprising:

[0028] a memory having a computer program stored thereon;

[0029] A processor is used to execute the computer program in the memory to implement the steps of the method as described above.

[0030] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the method described above when executed by a processor.

[0031] According to the fault prediction method, device, electronic device and storage medium of the electrolytic hydrogen production system of the present application, based on the real-time operation data obtained by the sensor, the key physical parameter data of the electrolytic hydrogen production system is determined by a preset physical model, and based on the key physical parameter data and the fault-related data in the real-time operation data, a pre-trained CNN-LSTM model is used to perform fault prediction, wherein the introduction of the key physical parameter data and the fault-related data can significantly improve the accuracy of fault prediction. Thus, the technical solution of the present application can accurately predict the fault and issue an alarm before the fault occurs based on the real-time operation data, thereby realizing active prevention and early processing of the fault, improving the safety of the operation of the electrolytic hydrogen production system, and at the same time, effectively reducing the frequency of manual inspection and maintenance and reducing costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to make the advantages of the present invention more easily understood, the present invention briefly described above will be described in more detail by referring to the specific embodiments shown in the accompanying drawings. It can be understood that these drawings only depict typical embodiments of the present invention and therefore should not be considered as limiting the scope of protection thereof, and the present invention is described and explained with additional characteristics and details through the accompanying drawings.

[0033] Figure 1 A schematic flow chart of a method for predicting a fault of a hydrogen production system by electrolysis according to an embodiment of the present application;

[0034] Figure 2 is a schematic structural block diagram of a fault prediction device for a hydrogen production system by electrolysis according to an embodiment of the present application;

[0035] Figure 3 FIG. 4 is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0036] In the following description, a large number of specific details are provided to provide a more thorough understanding of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without one or more of these details. In other examples, in order to avoid confusion with the embodiments of the present invention, some technical features well known in the art are not described.

[0037] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be noted that the terms "upper", "lower" and similar expressions used herein are only for the purpose of illustration and are not intended to be limiting.

[0038] Herein, ordinal numbers such as “first” and “second” cited in this application are merely identifications and do not have any other meanings, such as a specific order or the like.

[0039] In order to fully understand the embodiments of the present invention, detailed structures will be presented in the following description. Obviously, the implementation of the embodiments of the present invention is not limited to the specific details familiar to those skilled in the art. The preferred embodiments of the present invention are described in detail below, but in addition to these detailed descriptions, the present invention may also have other embodiments.

[0040] See attached Figure 1 A fault prediction method for a hydrogen production system by electrolysis according to an embodiment of the present application is schematically described. The fault prediction method includes:

[0041] S10: obtaining real-time operation data of the electrolysis hydrogen production system through sensors in the electrolysis hydrogen production system;

[0042] S20: Based on the real-time operation data, determine the key physical parameter data of the electrolysis hydrogen production system through a preset physical model;

[0043] S30: Based on the key physical parameter data and the fault-related data in the real-time operation data, the fault prediction results are output using the pre-trained CNN-LSTM (convolutional neural network-long short-term memory network) model;

[0044] S40: When the fault prediction result meets the preset conditions, an alarm message is issued.

[0045] According to the fault prediction method of the electrolysis hydrogen production system of the present application, based on the real-time operation data obtained by the sensor, the key physical parameter data of the electrolysis hydrogen production system is determined by a preset physical model, and based on the key physical parameter data and the fault-related data in the real-time operation data, fault prediction is performed using a pre-trained CNN-LSTM model.

[0046] Among them, the physical model can provide key physical parameter data directly related to the operation of the system. Compared with the key physical parameter data directly obtained by sensors, it can more sensitively reflect the changes in the system state, provide more background information for the model, and help identify complex faults. Fault-related data is data in real-time operation data that is highly correlated with the occurrence of faults. It is of great significance to improve the accuracy of model prediction. Inputting fault-related data in real-time operation data rather than all data into the CNN-LSTM model can also effectively reduce the amount of calculation and calculation time. The introduction of key physical parameter data and fault-related data can significantly improve the accuracy of fault prediction.

[0047] Therefore, the technical solution of the present application can accurately predict faults and issue an alarm before the fault occurs based on real-time operating data, realize active prevention and early processing of faults, improve the safety of operation of the electrolysis hydrogen production system, and at the same time, effectively reduce the frequency of manual inspection and maintenance and reduce costs.

[0048] In this embodiment, the electrolytic hydrogen production system is an alkaline solution electrolytic hydrogen production system, which mainly includes one or more electrolytic cells. The electrolytic cells electrolyze water to produce hydrogen in an alkaline electrolyte environment. The electrolyte is generally a 30% mass concentration KOH solution or a 26% mass concentration NaOH solution.

[0049] A plurality of sensors are provided in the hydrogen production system by electrolysis to obtain real-time operation data of the hydrogen production system by electrolysis, that is, real-time operation data of the electrolyzer. Exemplarily, in step S10, the real-time operation data obtained by the sensors in the hydrogen production system by electrolysis may include the temperature, pressure, voltage, current, medium flow, online component analysis instrument parameters, and hydrogen production of the electrolyzer. It should be noted that the real-time operation data is data directly obtained by sensors (such as temperature sensors, pressure sensors, etc.).

[0050] In step S20, the preset physical model may include an electrochemical model and a thermodynamic model. The electrochemical model is used to describe the electrochemical reaction process in the electrolyzer, including the reaction of the anode and cathode, the ion migration in the electrolyte, polarization phenomena, etc. The thermodynamic model is used to describe the heat transfer and temperature distribution in the device, including heat flow, temperature gradient, heat dissipation efficiency, etc. Based on real-time operation data, the key physical parameter data of the electrolysis hydrogen production system can be determined by the electrochemical model and the thermodynamic model. These key physical parameter data are highly correlated with the operation and failure of the electrolyzer, and these key physical parameter data are difficult to obtain directly through sensors. Exemplarily, the key physical parameter data may include electrolyte conductivity data, electrolyzer efficiency data, electrolyte concentration change data, local temperature gradient data, electrolyzer heat loss data, etc.

[0051] In this embodiment, the key physical parameter data may include the energy conversion efficiency (η) data of the electrolyzer in the electrolytic hydrogen production system, the electrolyte concentration change (ΔC) data (i.e., the change data of the ion concentration in the electrolyte), and the local temperature gradient in the electrolyzer. Data (that is, the temperature change rate at different locations inside the electrolyzer), etc. Exemplarily, the electrolyzer efficiency can be calculated based on the voltage, current and hydrogen production data in the real-time operation data, etc., through the above-mentioned electrochemical model; the electrolyte concentration change data is based on the current density and reaction time during the electrolysis process, estimated by the above-mentioned electrochemical model; the local temperature gradient can be calculated based on the temperature in the real-time operation data through the above-mentioned thermodynamic model. It should be noted that the specific calculation process of the above-mentioned key physical parameter data is known to computer personnel in this field, and is not an improvement point of the present application, and will not be described in detail here.

[0052] In some other embodiments, the physical model in step S20 may also be a physical model known to those skilled in the art, such as a fluid mechanics model, or a physical model established by those skilled in the art based on the operation of the electrolyzer and objective physical laws. As long as the real-time operation data obtained by the sensor can be used, the key physical parameter data highly related to the occurrence of the fault can be obtained through the physical model.

[0053] In step S30, the fault-related data in the real-time operation data can be determined by a custom feature selection method. The purpose of the custom feature selection is to find the feature data that is most relevant to the fault by analyzing the historical data. These feature data may be complex and multi-dimensional, and it is difficult to capture them by standard methods. Through custom feature selection, the most predictive data can be extracted to further improve the accuracy of the model. It should be noted that the fault-related data can be part of the original data in the real-time operation data, or it can be features extracted from the real-time operation data by various feature extraction methods known to those skilled in the art. In some other embodiments, the fault-related data can be the data that is most relevant to the fault that is directly determined by those skilled in the art from the real-time operation data based on experience, big data analysis, etc. For example, part of the original data in the fault-related data determined based on experience, big data analysis, etc. is used as the fault-related data, and / or, the feature data extracted from the fault-related data determined based on experience, big data analysis, etc. is used as the fault-related data.

[0054] In this embodiment, the fault-related data in the real-time operation data is determined in the following manner:

[0055] First, the historical operation data of the electrolytic hydrogen production system is cleaned. The historical operation data refers to the data directly obtained by sensors during the operation of the electrolytic hydrogen production system (that is, historical real-time operation data), especially the historical data before and after the failure. The historical operation data can include time series sensor readings, etc. Data cleaning includes operations such as removing duplicate data, filling missing values, processing outliers, and data standardization to improve the accuracy and reliability of the data. Data cleaning can eliminate data errors and noise and improve the accuracy of analysis and modeling.

[0056] Then, feature association analysis, feature extraction and feature selection are performed on the cleaned historical operation data to determine the initial fault association data.

[0057] Specifically, feature association analysis is used to analyze and determine historical operation data that has a high correlation with the occurrence of a fault. Exemplarily, performing feature association analysis on the cleaned historical operation data may include: using the Pearson correlation coefficient to evaluate the correlation between the historical operation data and the occurrence of a fault, with the historical operation data that has a high correlation with the occurrence of the fault, and the historical operation data with a high correlation may be the key to fault prediction. Performing feature association analysis on the cleaned historical operation data may also include: identifying the historical operation data associated with the fault by a dimensionality reduction method. The dimensionality reduction method may include principal component analysis (PCA), factor analysis (FA), etc.

[0058] Feature extraction is used to extract features associated with faults from historical operating data, providing strong support for model training. Exemplarily, feature extraction includes wavelet transform and / or time-frequency analysis. Exemplarily, historical operating data that is determined to have a high correlation with the occurrence of faults through feature association analysis can be subjected to wavelet transform and / or time-frequency analysis to extract features associated with faults. Wavelet transform decomposes non-stationary data into data sequences, i.e., wavelet coefficients, that express information at different levels and frequency bands, thereby achieving feature extraction of data. Wavelet transform can capture subtle changes and transient features in time series. Wavelet transform can decompose data into components of different frequencies, thereby capturing more delicate fault precursor signals. Time-frequency analysis can be to extract features within specific frequencies or time periods in historical operating data through short-time Fourier transform (STFT), which may be related to specific fault types of hydrogen electrolysis systems.

[0059] Feature selection is used to select the data (features) in the historical operation data that contribute most to fault prediction, and exclude those data (features) that do not contribute or. Specifically, feature selection may include recursive feature elimination (RFE) and / or tree-based feature selection. For recursive feature elimination, a machine learning model (support vector machine) can be used to recursively remove data (features) in the historical operation data that have little impact on the model, and retain the data (features) that contribute most to fault prediction. For tree-based feature selection, extreme random trees (Extra Trees) can be used to sort the data (features) in the historical operation data according to their importance and select the most representative data (features). Exemplarily, feature selection can be performed on features extracted by wavelet transform and / or time-frequency analysis to obtain the features that contribute most to fault prediction, that is, the initial fault association data.

[0060] Exemplarily, feature association analysis, feature extraction, and feature selection can be performed sequentially, that is, feature association analysis is performed first to determine historical operation data with a high correlation with the occurrence of the fault (the correlation meets a preset target), and then feature extraction is performed on the historical operation data with a high correlation with the occurrence of the fault determined by the feature association analysis, and then feature selection is performed on the extracted features to obtain initial fault association data. In some embodiments, feature association analysis, feature extraction, and feature selection can be performed in other orders, or in parallel (feature association analysis, feature extraction, and feature selection are performed on the cleaned historical operation data, respectively), to obtain initial fault association data.

[0061] Finally, the initial fault-associated data is used for the training and verification of the CNN-LSTM model to adjust and optimize the initial fault-associated data to obtain the final fault-associated data. Exemplarily, the initial fault-associated data can be used for the training and verification of the CNN-LSTM model to observe the effect of the selection of the initial fault-associated data on the model prediction performance. According to the model performance, the data with a small improvement effect on the model prediction performance can be excluded from the initial fault-associated data to obtain the final fault-associated data. In some embodiments, according to the model performance, the specific methods of feature association analysis, feature extraction and feature selection can be adjusted and optimized, and the initial fault-associated data with a better improvement effect on the model prediction performance can be reselected and used as the final fault-associated data. The final fault-associated data is the feature data most relevant to the fault in the historical operation data, which corresponds to the feature data most relevant to the fault in the real-time operation data. Thus, the fault-associated data can be determined based on this after the real-time operation data is obtained.

[0062] In step S30, the key physical parameter data and the fault-related data in the real-time operation data are input into a pre-trained CNN-LSTM model, and the fault prediction result is output through the CNN-LSTM model.

[0063] The CNN-LSTM model includes CNN (convolutional neural network) for spatial feature extraction and LSTM (long short-term memory network) for temporal feature analysis. CNN is extremely effective in extracting spatial and temporal features from multidimensional data. When predicting faults in electrolytic hydrogen production systems, CNN can analyze input data and detect patterns that may indicate a fault or an impending fault. LSTM networks are specifically designed to handle sequence and time correlations, so they are very suitable for analyzing time series data extracted and processed by CNN.

[0064] In this embodiment, the key steps of the CNN-LSTM model are as follows:

[0065] A.CNN extracts features:

[0066] The CNN processes the input data (key physical parameter data and fault association data) in overlapping windows, extracts spatial features and creates feature maps that capture the important aspects of each window.

[0067] B. Feature map flattening:

[0068] Before entering the LSTM, the output of the last pooling or convolutional layer of the CNN is flattened into a one-dimensional vector if it has not been flattened yet. This vector provides a compressed representation of the key features of the input window.

[0069] C. Sequential input preparation:

[0070] These flattened vectors are then sorted into time order. This order is crucial because it preserves the temporal relationship between consecutive measurements.

[0071] D.LSTM timing analysis:

[0072] The LSTM network receives these sequences and processes them to learn and identify long-term dependencies and patterns that indicate the health and operation of the system. Each LSTM unit uses gates to control the flow of information, effectively learning which information to keep and which to ignore, which is critical for maintaining long-term memories of important events, such as gradients that lead to potential failures.

[0073] D. Prediction output:

[0074] The output of LSTM can be configured to predict the future instantaneous state and classify the current state as normal or potential fault. That is, the fault prediction result output by the CNN-LSTM model can be normal or potential fault. In some embodiments, the fault prediction result output by LSTM can be configured as the probability of fault occurrence.

[0075] The CNN-LSTM model can be trained and verified using historical operating data of the electrolytic hydrogen production system. Exemplarily, the training environment can be a computing platform that can handle large data sets and intensive computing, such as a GPU or a cloud-based service. The k-fold cross-validation technique is used for model verification to ensure that the model can be well generalized in different operating scenarios. In addition, the model is evaluated using indicators such as accuracy, precision, recall, F1 score, and ROC curve to fine-tune parameters and select the best model. It should be noted that the specific training and verification methods of the CNN-LSTM model are well known to those skilled in the art and will not be described here. The technical solution of the present application mainly uses the CNN-LSTM model for fault prediction. CNN can extract complex features from input data, capture subtle patterns and anomalies, and LSTM can effectively simulate behaviors that change over time, identify long-term trends and short-term fluctuations, and the use of the CNN-LSTM model can effectively improve the accuracy and robustness of fault prediction.

[0076] In step S40, when the fault prediction result meets the preset conditions, for example, when the fault prediction result output by the CNN-LSTM model can be a potential fault, or when the probability of the fault output by the LSTM exceeds a set threshold, an alarm message can be issued through an alarm device (such as an audible or visual alarm, etc.) so that the staff can know and handle it in time. In some embodiments, a user interface that can display real-time operating data, fault prediction results, and alarm information can be developed. When the fault prediction result meets the preset conditions, the alarm information is displayed through the user interface so that the staff can monitor the system operation status and respond to the alarm quickly.

[0077] The present application also provides a fault prediction device for an electrolytic hydrogen production system, which comprises:

[0078] The acquisition unit is used to acquire real-time operation data of the hydrogen production system by electrolysis through sensors in the hydrogen production system by electrolysis.

[0079] The determination unit is used to determine key physical parameter data of the electrolysis hydrogen production system based on real-time operation data through a preset physical model.

[0080] The output unit is used to output the fault prediction results based on the key physical parameter data and the fault association data in the real-time operation data using the pre-trained CNN-LSTM model.

[0081] The alarm unit is used to issue an alarm message when the fault prediction result meets the preset conditions.

[0082] It can be understood that the contents in the embodiments of the fault prediction method for the above-mentioned hydrogen production by electrolysis system are applicable to the embodiments of the present device, the functions specifically implemented by the embodiments of the present device are the same as those in the embodiments of the fault prediction method for the above-mentioned hydrogen production by electrolysis system, and the beneficial effects achieved are also the same as the beneficial effects achieved by the embodiments of the fault prediction method for the above-mentioned hydrogen production by electrolysis system.

[0083] The present application also provides an electronic device, which includes: a memory and a processor, wherein a computer program is stored in the memory, and the processor is used to execute the computer program in the memory to implement the steps of the fault prediction method of the above-mentioned hydrogen production system by electrolysis.

[0084] See attached Figure 3 The structure of an electronic device according to a specific embodiment of the present application is schematically described. The electronic device includes:

[0085] The processor 301 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute a computer program to implement the technical solution provided in the embodiment of the present application;

[0086] The memory 302 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 302 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant computer program is stored in the memory 302, and the processor 301 calls and executes the fault prediction method of the electrolysis hydrogen production system of the embodiment of the present application;

[0087] Input / output interface 303, used to implement information input and output;

[0088] The communication interface 304 is used to realize the communication interaction between the device and other devices. The communication can be realized through a wired manner (such as USB, network cable, etc.) or a wireless manner (such as mobile network, WIFI, Bluetooth, etc.);

[0089] A bus 305 that transmits information between the various components of the device (e.g., the processor 301, the memory 302, the input / output interface 303, and the communication interface 304);

[0090] The processor 301 , the memory 302 , the input / output interface 303 and the communication interface 304 are connected to each other in communication within the device via the bus 305 .

[0091] An embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the fault prediction method for the above-mentioned hydrogen production system by electrolysis are implemented.

[0092] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0093] Although example embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above example embodiments are merely exemplary and are not intended to limit the scope of the present application to this. Those of ordinary skill in the art may make various changes and modifications therein without departing from the scope and spirit of the present application. All these changes and modifications are intended to be included within the scope of the present application as required by the appended claims.

[0094] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0095] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed.

[0096] In the description provided herein, a large number of specific details are described. However, it is understood that the embodiments of the present application can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0097] Similarly, it should be understood that in order to streamline the present application and help understand one or more of the various inventive aspects, in the description of the exemplary embodiments of the present application, the various features of the present application are sometimes grouped together into a single embodiment, figure, or description thereof. However, the method of the present application should not be interpreted as reflecting the following intention: the claimed application requires more features than the features clearly stated in each claim. More specifically, as reflected in the corresponding claims, the inventive point is that the corresponding technical problem can be solved with features less than all the features of a single disclosed embodiment. Therefore, the claims following the specific embodiment are hereby explicitly incorporated into the specific embodiment, wherein each claim itself serves as a separate embodiment of the present application.

[0098] It will be understood by those skilled in the art that, except for mutually exclusive features, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this specification may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature that provides the same, equivalent or similar purpose.

[0099] In addition, those skilled in the art will appreciate that, although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present application and form different embodiments. For example, in the claims, any one of the claimed embodiments can be used in any combination.

[0100] It should be noted that the above-mentioned embodiments illustrate rather than limit the invention and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims.

Claims

1. A fault prediction method for a hydrogen production system by electrolysis, characterized in that: include: Acquiring real-time operation data of the hydrogen production system by means of sensors in the hydrogen production system by electrolysis; Based on the real-time operation data, key physical parameter data of the electrolysis hydrogen production system is determined by a preset physical model; Based on the key physical parameter data and the fault-related data in the real-time operation data, a pre-trained CNN-LSTM model is used to output a fault prediction result; When the fault prediction result meets the preset conditions, an alarm message is issued.

2. The fault prediction method according to claim 1, characterized in that: The physical model includes an electrochemical model and a thermodynamic model.

3. The fault prediction method according to claim 2, characterized in that: The key physical parameter data include energy conversion efficiency data of the electrolyzer in the electrolytic hydrogen production system, electrolyte concentration change data and local temperature gradient data in the electrolyzer.

4. The fault prediction method according to claim 1, characterized in that: The fault-related data in the real-time operation data is determined in the following manner: Cleaning the historical operation data of the electrolysis hydrogen production system; Performing feature association analysis, feature extraction and feature selection on the cleaned historical operation data to determine initial fault association data; Using the initial fault association data for training and verification of the CNN-LSTM model to adjust and optimize the initial fault association data to obtain the final fault association data; The historical operation data includes operation data acquired by sensors in the hydrogen production system by electrolysis before and after the fault occurs.

5. The fault prediction method according to claim 4, characterized in that: Performing feature correlation analysis on the cleaned historical operation data includes: Using the Pearson correlation coefficient to evaluate the correlation between the historical operating data and the occurrence of faults; and / or, The historical operation data associated with the fault is identified by a dimensionality reduction method.

6. The fault prediction method according to claim 4, characterized in that: The feature extraction includes wavelet transform and / or time-frequency analysis.

7. The fault prediction method according to claim 4, characterized in that: The feature selection includes recursive feature elimination and / or tree-based feature selection.

8. A fault prediction device for an electrolytic hydrogen production system, characterized in that: include: An acquisition unit, used for acquiring real-time operation data of the hydrogen production system by electrolysis through sensors in the hydrogen production system by electrolysis; A determination unit, configured to determine key physical parameter data of the electrolysis hydrogen production system through a preset physical model based on the real-time operation data; An output unit, configured to output a fault prediction result using a pre-trained CNN-LSTM model based on the key physical parameter data and the fault-related data in the real-time operation data; The alarm unit is used to issue an alarm message when the fault prediction result meets the preset conditions.

9. An electronic device, characterized in that: include: a memory having a computer program stored thereon; A processor, configured to execute the computer program in the memory to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images