Safety management method and system for hydrogen energy storage system

Abstract

The invention relates to a safety management method and system for a hydrogen energy storage system, and relates to the technical field of energy storage systems, and the method comprises the steps: obtaining the operation state data of the hydrogen energy storage system; extracting global features from the operation state data; extracting energy propagation characteristics from the operation state data based on an energy transmission path, extracting hydrogen propagation characteristics from the operation state data based on a hydrogen transmission path, and fusing the energy propagation characteristics and the hydrogen propagation characteristics to obtain local characteristics; fusing the local features and the global features to obtain enhanced global features; inputting the enhanced global features and the local features into a preset risk assessment model to obtain a risk assessment result; and generating alarm information based on the risk assessment result. According to the method, the local features are introduced based on two core physical processes of energy propagation and hydrogen propagation of the hydrogen energy storage system to enhance the global features, so that the risk assessment precision and the response speed are improved.

Description

Technical Field

[0001] The present invention relates to the field of energy storage technology, and in particular to a safety management method and system for a hydrogen energy storage system. Background Art

[0002] Hydrogen Energy Storage System (HESS) is the core technology for the large-scale application of renewable energy. It realizes the efficient conversion of electric energy-hydrogen energy-electric energy through water electrolysis, high-pressure hydrogen storage and fuel cell power generation. Its core equipment includes electrolyzers, compressors, hydrogen storage tanks and fuel cells, involving multi-physical field coupling processes such as power conversion and hydrogen transportation. However, the flammable and explosive characteristics of hydrogen energy, the complex interactions between equipment and the strong volatility of renewable energy input make the safe operation of the system face severe challenges.

[0003] In recent years, deep learning technology has been gradually applied to the management of hydrogen energy storage systems. Long short-term memory networks, convolutional neural networks and other networks are used to conduct risk assessments on equipment sensor data, and then risk warnings are issued based on abnormal monitoring results to ensure equipment safety. The safety management of hydrogen energy storage systems mainly includes risk assessment and fault location. Traditional risk assessment collects the operating status data of each device separately, evaluates the health status of the device based on these operating status data, and evaluates the risk level of the system based on the weight of the device and the health status of the device. This method performs weighted averaging of the health status, resulting in loss of detailed features, low risk assessment accuracy, delayed response, and other problems, which cannot meet the safety management needs of hydrogen energy storage systems. At the same time, most existing models are purely data-driven, and do not integrate the physical laws of the energy transmission path and hydrogen diffusion process of the hydrogen energy storage system, resulting in poor model interpretability, low accuracy, and delayed response, which cannot meet the safety management needs of hydrogen energy storage systems. Summary of the invention

[0004] In order to overcome the above technical problems existing in the prior art, the present invention provides a safety management method and system for a hydrogen energy storage system.

[0005] On the one hand, a safety management method for a hydrogen energy storage system is provided, comprising: Acquiring operating status data of the hydrogen energy storage system; extracting global features from the operating status data; Extracting energy propagation features from the operating state data based on the energy transmission path, extracting hydrogen propagation features from the operating state data based on the hydrogen transmission path, and fusing the energy propagation features and the hydrogen propagation features to obtain local features; The local feature and the global feature are fused to obtain an enhanced global feature; Inputting the enhanced global features and the local features into a preset risk assessment model to obtain a risk assessment result; Generate alarm information based on risk assessment results.

[0006] Preferably, extracting global features from the operating status data includes: extracting the operation mode and the importance of the equipment from the operation status data; Determining a device weight based on the operation mode and the device importance; The operating status data is fused based on the device weight to obtain a global feature.

[0007] Preferably, extracting energy propagation characteristics from the operating status data based on the energy transmission path includes: Based on the energy transmission process of the hydrogen energy storage system, construct an energy transmission path; extracting current data and voltage data of nodes of the energy transmission path from the operating status data based on the energy transmission path; Performing feature reconstruction on the current data and the voltage data to obtain an energy propagation matrix; The energy propagation matrix is ​​convolved to obtain energy propagation features.

[0008] Preferably, the current data and the voltage data are subjected to feature reconstruction to obtain an energy propagation matrix, including: Performing FFT analysis on the current data, and determining the harmonic distortion rate according to the FFT analysis result; Performing wavelet decomposition on the current data and the voltage data respectively, and determining the current spectrum energy proportion and the current spectrum energy proportion according to the wavelet decomposition results; The node weights of the nodes of each energy transmission path are determined according to the energy transmission path, and an energy propagation matrix is ​​constructed based on the node weights, the harmonic distortion rate, the current spectrum energy proportion and the current spectrum energy.

[0009] Preferably, extracting hydrogen propagation characteristics from the operating status data based on the hydrogen transmission path includes: Establishing a hydrogen transmission path based on the transmission process of hydrogen in the hydrogen energy storage system; extracting a pressure signal and a flow signal from the operating status data based on the hydrogen transmission path; Performing feature reconstruction on the flow signal and the pressure signal to obtain a hydrogen propagation matrix; The hydrogen propagation matrix is ​​convolved to obtain hydrogen propagation features.

[0010] Preferably, the flow signal and the pressure signal are subjected to feature reconstruction to obtain a hydrogen propagation matrix, including: determining a flow rate fluctuation rate according to the flow signal, and determining a pressure gradient according to the pressure signal; Inputting the flow signal and the pressure signal into a preset physical information neural network model to obtain a predicted flow fluctuation rate and a predicted pressure gradient; A hydrogen propagation matrix is ​​generated according to the flow fluctuation rate, the pressure gradient, the predicted flow fluctuation rate and the predicted pressure gradient.

[0011] Preferably, the energy propagation feature and the hydrogen propagation feature are merged to obtain a local feature, further comprising: When the system device is in cluster mode, calculating a load balancing index between devices according to the load rate; generating a load balancing feature based on the inter-device load balancing index; The inter-device load balancing feature, the energy propagation feature and the hydrogen propagation feature are integrated to obtain a local feature.

[0012] In a second aspect, a safety management system for a hydrogen energy storage system is provided, the system comprising: An acquisition module, used to acquire operating status data of the hydrogen energy storage system; A global feature extraction module, used to extract global features from the operating status data; A local feature extraction module, configured to extract energy propagation features from the operating state data based on an energy transmission path, extract hydrogen propagation features from the operating state data based on a hydrogen transmission path, and fuse the energy propagation features with the hydrogen propagation features to obtain local features; A global feature enhancement module, used for fusing the local feature with the global feature to obtain an enhanced global feature; A risk assessment module, used for inputting the enhanced global features and the local features into a preset risk assessment model to obtain a risk assessment result; The alarm module is used to generate alarm information based on risk assessment results.

[0013] Through the technical solution provided by the present invention, the present invention has at least the following technical effects: Based on the two core physical processes of energy propagation and hydrogen propagation of the hydrogen energy storage system, the present invention introduces local features to enhance the global features, increases the detailed features of key equipment, significantly improves the risk assessment accuracy and response speed, and can meet the safety management needs of the hydrogen energy storage system.

[0014] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0016] Figure 1 is a flow chart of a safety management method for a hydrogen energy storage system provided by an embodiment of the present invention;

[0017] Figure 2 It is a system block diagram of a safety management system of a hydrogen energy storage system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The specific implementation of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.

[0019] The terms "system" and "network" in the embodiments of the present invention can be used interchangeably. "Multiple" means two or more than two. In view of this, "multiple" can also be understood as "at least two" in the embodiments of the present invention. "And / or" describes the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / ", unless otherwise specified, generally indicates that the previous and subsequent associated objects are in an "or" relationship. In addition, it should be understood that in the description of the embodiments of the present invention, the words "first", "second", etc. are only used to distinguish the purpose of description, and cannot be understood as indicating or implying relative importance, nor can they be understood as indicating or implying order.

[0020] Based on the above, see Figure 1 , an embodiment of the present invention provides a safety management method for a hydrogen energy storage system, the method comprising: Step 1, obtaining operating status data of the hydrogen energy storage system; The equipment in the hydrogen energy storage system includes electrolyzers, compressors, fuel cells, pipelines, valves, power distribution cabinets, cooling systems, etc. These devices are deployed with a number of sensors that can obtain the corresponding operating status data. For example, the obtained electrolyzer operating status data includes input current, input voltage, temperature, pressure, input current, hydrogen concentration, etc., and the obtained compressor operating status data includes input current, input voltage, vibration, oil pressure, speed, flow, etc. In one possible implementation, the data collected by all sensors and the time information and power recorded by each device are used as operating status data for subsequent risk assessment. Although this method can realize the hydrogen energy storage system, a hydrogen energy storage system is equipped with a large number of sensors and many redundant designs, which will lead to a sharp increase in the amount of data, thereby affecting the efficiency of risk assessment and the response speed of safety management.

[0021] Therefore, the embodiment of the present invention chooses to obtain all sensor data on key equipment and obtain part of the sensor data on non-key equipment, wherein the key equipment includes electrolyzers, compressors, fuel cells and cooling systems, and the non-key equipment includes pipelines, valves, distribution cabinets, etc. Selecting to obtain part of the sensor data based on the distance between the sensors on the non-key equipment and the key equipment can not only reduce the amount of data, but also evaluate the overall system operation.

[0022] Furthermore, the operating status data is time series data, and the acquisition process is as follows: obtain the original operating status data with time information, perform time series alignment according to the time information, and perform noise reduction on the aligned data, apply 50Hz notch filtering and 1kHz low-pass filtering to the electrical signals in the data, use wavelet threshold denoising on the mechanical vibration signals of the data, fill missing values ​​in the data after noise reduction, use forward filling for short-term missing values, and fill long-term missing values ​​based on the LSTM prediction model.

[0023] After obtaining multi-source monitoring data, these data need to be analyzed to extract the data features.

[0024] Step 2: extracting global features from the running status data.

[0025] In the embodiment of the present invention, the global feature fuses the operation status data of each device at the device level to obtain the device feature, and then fuses these device features based on the device weight to obtain the global feature.

[0026] In one possible implementation, the ratio of the historical failure frequency of a single device to the average failure frequency of all devices in the hydrogen energy storage system is calculated, which can well reflect the reliability of the device. The ratio of the maintenance cost to the replacement cost of a single device is calculated, which can well reflect the economic efficiency of the device. Then, the importance of the device is comprehensively calculated based on the two ratios, and the device characteristics are fused according to the device weights to obtain global characteristics.

[0027] However, in actual applications, the hydrogen energy storage system has different operating modes, such as full-load hydrogen production mode, energy storage priority mode, emergency standby mode, etc. Key equipment such as electrolyzers, compressors, hydrogen storage tanks, and fuel cells have different importance. For example, in full-load hydrogen production mode, the system is fully engaged in hydrogen production. As the core equipment for electrolyzing water into hydrogen and oxygen, the electrolyzer has an extremely high load. Its operating state directly determines the efficiency and output of hydrogen production, and plays a key role in the hydrogen production process of the entire system. The compressor is responsible for compressing the generated hydrogen to a suitable pressure for storage or transportation. Its load is also at a high level. If the compressor operates abnormally, it will affect the storage and subsequent use of hydrogen. In contrast, the hydrogen storage tank is mainly used to receive and store hydrogen at this time. Although it is also important, its importance is slightly weaker in the current mode compared to the key equipment in the hydrogen production link. Therefore, the present invention considers the operating mode and further adjusts the equipment weight.

[0028] In an embodiment of the present invention, extracting global features from the operating status data includes: extracting the operating mode and the importance of the equipment from the operating status data; determining the equipment weight based on the operating mode and the equipment importance; and fusing the operating status data based on the equipment weight to obtain the global features.

[0029] Specifically, the operation mode can be determined according to the actual operation power of each device during the adoption cycle, and the load rate of the system device can be determined according to the ratio of the actual operation power to the rated operation power. When the load rates are all within the normal range, the current operation mode is determined according to the load rate and the preset load rate operation mode relationship mapping table, and then the device weight is adjusted according to the preset operation mode device importance adjustment rule to obtain the adjusted device weight. Of course, the operation mode can also be determined according to actual needs, which will not be repeated in the embodiments of the present invention.

[0030] The embodiment of the present invention can dynamically adjust the device weight according to the different operation modes of the hydrogen energy storage system.

[0031] After extracting the global features, the risk situation of the system can be evaluated based on the global features. The global features are extracted based on the weighted fusion of the operating status data at the device level. Therefore, it can capture the overall trend and pattern of the data, which helps to fully understand the overall state of the hydrogen energy storage system. However, since the global features are extracted from the entire data set and the data are processed in the same way, the detailed features of the global features are lost and are relatively insensitive to the anomalies or noise of some individual data points. In one possible implementation, fault statistics are introduced, a fault topology map is generated based on the fault statistics, and the fault topology map is fused with the global features to enhance the global features. Although this method can realize the risk transmission assessment across devices, it relies on the statistical correlation of historical faults. In the face of sudden unknown faults, for example, a hydrogen power station caused an instantaneous surge in the power grid due to lightning strikes, and the harmonic distortion rate of the electrolyzer input current soared from 5% to 40% in 0.5 seconds, and the harmonics were transmitted to the fuel cell, causing the inverter to burn out. If there is no similar scene in the historical data, it cannot be identified. In addition, this method requires a large number of fault samples. Even if there are similar scenes in the historical data, the amount of data in the historical data may not meet the demand, that is, this method has problems such as difficulty in data collection, complex model construction and training process. In another possible implementation, all operating status data are fused at the sensor level to learn the correlation between sensor data between different devices, so as to realize the recognition of detailed features. However, this method introduces a large amount of data, generates more noise, and is also very complicated in the subsequent model training process. Therefore, it is necessary to provide a local feature acquisition method that does not rely on historical statistical faults and a large amount of data to enhance global features. Since the hydrogen energy storage system mainly involves two core physical processes: energy conversion and hydrogen transmission, the data in these two core physical processes can cover the system operation. Therefore, an embodiment of the present invention proposes a local feature extraction method that takes energy propagation and hydrogen transmission into consideration.

[0032] Step 3: extract energy propagation features from the operating status data based on the energy transmission path, extract hydrogen propagation features from the operating status data based on the hydrogen transmission path, and fuse the energy propagation features and the hydrogen propagation features to obtain local features.

[0033] In a hydrogen energy storage system, the process of electric energy transmission is that the electrolyzer converts electric energy into hydrogen energy, transmits it to the compressor drive motor, and generates electricity through fuel cells to feed back to the power grid. In one possible implementation, energy transmission characteristics are extracted from the operating status data based on the energy transmission path, including: extracting power parameters of the electrolyzer, compressor, and fuel cell from the operating status data, the power parameters include the current and voltage at the input end of each device and the current and voltage at the output end, constructing local features based on the extracted power parameters, and fusing the local features with the global features, thereby enhancing the risk perception ability in the global feature hydrogen energy storage physical process. Although this method can realize the extraction of energy transmission process characteristics, it does not analyze data such as current and voltage, resulting in insufficient understanding of the dynamic transmission characteristics of electric energy.

[0034] In an embodiment of the present invention, energy propagation characteristics are extracted from the operating status data based on the energy transmission path, including: extracting energy propagation characteristics from the operating status data based on the energy transmission path, including: constructing an energy transmission path based on the energy transmission process in the hydrogen energy storage system; extracting current data and voltage data of nodes of the energy transmission path from the operating status data based on the energy transmission path; reconstructing features of the current data and the voltage data to obtain an energy propagation matrix; and convolving the energy propagation matrix to obtain energy propagation characteristics.

[0035] In the electrolyzer→compressor→fuel cell path, the electrolyzer input end is the starting point for converting electrical energy into hydrogen energy. Its current and voltage directly affect the electrolysis efficiency and subsequent equipment safety. Monitoring here can capture key risks such as rectifier failure and power grid disturbance. The current and voltage content of the drive motor at the compressor drive end are directly related to the compressor efficiency and mechanical wear, etc. The current and voltage at the fuel cell output end determine the quality of electrical energy feedback. Therefore, the embodiment of the present invention extracts the current data and voltage data of the electrolyzer, compressor and fuel cell from the operating status data based on the energy transmission path.

[0036] In one possible way, based on the transmission process of electric energy in the hydrogen energy storage system, the electric energy transmission path constructed is electrolyzer→compressor→fuel cell, and the current data and voltage data of the electrolyzer input end, the current data and voltage data of the compressor drive motor input end, and the current data and voltage data of the fuel cell output end are extracted from the motion state data according to the energy transmission path, and the parameters such as the current data and the voltage data are calculated according to the current change rate and the voltage change rate, and the energy propagation matrix is ​​generated according to the current change rate and the voltage change rate. The energy propagation matrix generated by this method is a statistical feature, and no in-depth analysis of the current data and the voltage data is performed, resulting in insufficient details.

[0037] In one possible implementation, based on the transmission process of electric energy in the hydrogen energy storage system, an electric energy transmission path is constructed as electrolyzer→compressor→fuel cell. According to the energy transmission path, the current data and voltage data of the electrolyzer input end, the current data and voltage data of the compressor drive motor input end, and the current data and voltage data of the fuel cell output end are extracted from the motion state data. A 1024-point FFT is performed on the extracted current data, and the effective values ​​of the fundamental wave and harmonics are calculated to obtain the harmonic distortion rate of the electrolyzer, compressor and fuel cell. At the same time, the extracted voltage data is decomposed into three layers using the Daubechies4 wavelet basis to obtain multiple frequency bands, and the energy proportion of each frequency band is calculated. The energy path node position code is generated based on the energy transmission path, and the corresponding energy characteristic moment is generated based on the energy path node position code and the harmonic distortion rate energy proportion. The convolution network is used to convolve the energy propagation spectrum matrix to obtain the energy propagation characteristics.

[0038] Since the state of the electrolyzer is more directly affected by the current, for example, current harmonics reflect rectifier problems, while the voltage fluctuation of the fuel cell is more indicative of internal or external problems, such as disturbances transmitted by inverter failure or compressor surge. The compressor is stabilized by a frequency converter, and its voltage signal responds weakly to mechanical failures (such as surge). In order to reduce the amount of data, in a preferred embodiment, only the current data of the electrolyzer and the compressor are subjected to 1024-point FFT, the fundamental wave and harmonic effective values ​​are calculated, and the compressor harmonic distortion rate of the electrolyzer is obtained; the current signal at the input end of the compressor drive motor is subjected to wavelet decomposition, and the voltage signal at the output end of the fuel cell is subjected to wavelet decomposition.

[0039] The embodiment of the present invention constructs an energy transmission path, extracts features such as harmonic distortion rate, and integrates convolution operations to obtain energy propagation features. This feature can make up for the problem of missing detail features in the global feature, so that the global feature pays more attention to the physical process of energy transmission.

[0040] In an embodiment of the present invention, hydrogen propagation characteristics are extracted from the operating status data based on the hydrogen transmission path, including: constructing a hydrogen transmission path based on the transmission process of hydrogen in the hydrogen energy storage system; extracting flow volatility and pressure gradient from the operating status data based on the hydrogen transmission path; reconstructing features based on the flow volatility and the pressure gradient to obtain a hydrogen feature matrix; and convolving the hydrogen feature matrix to obtain hydrogen propagation characteristics.

[0041] In a possible implementation, the hydrogen transmission path is electrolyzer → compressor → hydrogen storage tank → fuel cell, covering the entire process of hydrogen generation, compression, storage and use. Along this path, the flow signal and pressure signal of the electrolyzer outlet, the flow signal and pressure signal of the compressor inlet and outlet, the flow signal and pressure signal of the hydrogen storage tank inlet and outlet, and the pressure signal and flow signal of the fuel cell inlet are extracted from the operating status data, and the flow fluctuation rate is calculated for the flow signal, and the flow fluctuation rate reflects the impact of instantaneous flow changes on system stability; the pressure gradient is calculated for the pressure signal, and the pressure difference abnormality of the upstream and downstream of the pressure gradient pipeline is quantified, and the hydrogen path node position coding is generated based on the hydrogen transmission path, and the corresponding hydrogen feature moment is generated based on the hydrogen path node position coding, flow fluctuation rate, and pressure gradient. The convolution network is then used to convolve the hydrogen feature moment to obtain the hydrogen propagation feature. This method constructs an energy transmission path, extracts flow fluctuation rate and pressure gradient, and integrates spatiotemporal convolution operations to obtain hydrogen propagation features. This feature can make up for the problem of missing detail features in the global feature, so that the global feature pays more attention to the physical process of hydrogen transmission. However, in practical applications, due to the long transmission pipeline and the great hazard of hydrogen leakage, only obtaining the flow signal and pressure signal of the electrolyzer outlet, the flow signal and pressure signal of the compressor inlet and outlet, and the flow signal and pressure signal of the hydrogen storage tank inlet and outlet cannot well characterize the hydrogen transmission process. Therefore, it is necessary to further analyze the pressure signal and flow signal to obtain more characteristic information.

[0042] In an embodiment of the present invention, the flow signal and the pressure signal are feature reconstructed to obtain a hydrogen propagation matrix, including: determining the flow fluctuation rate according to the flow signal, and determining the pressure gradient according to the pressure signal; inputting the flow signal and the pressure signal into a preset physical information neural network model to obtain a predicted flow fluctuation rate and a predicted pressure gradient; and generating a hydrogen propagation matrix according to the flow fluctuation rate, the pressure gradient, the predicted flow fluctuation rate and the predicted pressure gradient.

[0043] In one possible implementation, a physical information neural network model is pre-constructed, which can determine the flow signals and pressure signals at other key locations on the transmission pipeline based on the pressure signals and flow signals at the inlets and outlets of each key equipment, the hydrogen production efficiency of the electrolyzer, and the speed of the compressor, and then calculate the flow fluctuation rate of the flow signal and the pressure gradient of the pressure signal, so as to obtain more hydrogen propagation characteristic information.

[0044] The embodiment of the present invention simulates and predicts hydrogen propagation through a neural network, thereby obtaining more information and enriching local details.

[0045] In a large-scale hydrogen energy storage system, electrolyzers, compressors and fuel cells are arranged in arrays. Unbalanced loads within these arrays will increase equipment risks. Therefore, the present invention provides a local feature determination method that takes load balancing into consideration.

[0046] In an embodiment of the present invention, energy propagation characteristics are extracted from the operating status data based on the energy transmission path, hydrogen propagation characteristics are extracted from the operating status data based on the hydrogen transmission path, and the energy propagation characteristics and the hydrogen propagation characteristics are fused to obtain local characteristics, and also include: when the system equipment is in cluster mode, calculating the load balancing index between devices according to the load rate; when the system equipment is in cluster mode, extracting the operating power from the operating status data; calculating the load rate of the system equipment according to the operating power, and calculating the load balancing index between devices according to the load rate; generating a load balancing feature based on the load balancing index between devices; fusing the load balancing feature between devices, the energy propagation feature and the hydrogen propagation feature to obtain local characteristics.

[0047] After obtaining local features, local features need to be fused with global features. Traditional methods usually fuse local features with global features through weighted averaging or simple splicing. Simple linear operations cannot model the complex association between global and local features, leading to misjudgment of cross-device risk transmission.

[0048] Step 5: Fusing the local features with the global features to obtain enhanced global features.

[0049] In the feature fusion stage, global features and local features are input into the cross-attention module, where global features are used as queries and local features are used as keys and values. The association weights between features are dynamically calculated through the multi-head attention mechanism. Specifically, the cross-attention mechanism constructs a cross-modal feature association model by using global features as queries and local features as keys and values. Global features represent the overall trend of the system, while local features capture the dynamic details of the physical process. Cross-attention dynamically allocates weights by calculating the correlation between the two. For example, when the harmonic distortion rate of the electrolyzer suddenly increases to 20%, cross-attention automatically increases the weight of the energy propagation feature. Even if the global feature does not reach the threshold due to mean smoothing, it can still accurately trigger an early warning; conversely, when the system is in steady-state operation, the global feature dominates the risk assessment to avoid excessive response to noise. The multi-head attention mechanism is further introduced to allow parallel attention to the interaction of features in different subspaces, thereby enhancing the global feature's perception of single anomalies and cross-device anomalies in the two core physical processes of power transmission and hydrogen transmission.

[0050] Step 6: Input the enhanced global features and the local features into a preset risk assessment model to obtain a risk assessment result, and generate alarm information based on the risk assessment result.

[0051] Specifically, the safety management of the hydrogen energy storage system needs to take into account both global state perception and local anomaly capture. For this purpose, the present invention constructs a dual-channel detection network, including a fully enhanced global feature channel and a local feature independent detection channel, wherein the fully enhanced global feature channel is a bidirectional LSTM network and the local feature channel is a convolutional neural network. The enhanced global features and local features are independently processed through the two channels to obtain enhanced global feature scores and local feature scores, and then the enhanced global feature scores and local feature scores are weightedly summed to obtain the final risk score, and risk warnings are issued based on the final risk score.

[0052] For the same inventive concept, please refer to Figure 2 The embodiment of the present invention also provides a safety management system for a hydrogen energy storage system, the system comprising: an acquisition module for acquiring operating status data of the hydrogen energy storage system; a global feature extraction module for extracting device-level global features and sensor-level global features from the operating status data, and fusing the device-level global features with the sensor-level global features to obtain global features; a local feature extraction module for extracting energy propagation features from the operating status data based on an energy transmission path, and extracting hydrogen propagation features from the operating status data based on a hydrogen transmission path, and fusing the energy propagation features with the hydrogen propagation features to obtain local features; a global feature enhancement module for fusing the local features with the global features to obtain enhanced global features; a risk assessment module for inputting the enhanced global features and the local features into a preset risk assessment model to obtain a risk assessment result; and an alarm module for generating alarm information based on the risk assessment result.

[0053] It should be understood that the safety management system of a hydrogen energy storage system provided in an embodiment of the present invention and the safety management system of a hydrogen energy storage system provided in the above embodiment are based on the same inventive concept. Therefore, the working process of each module in the safety management system of a hydrogen energy storage system provided in an embodiment of the present invention is the same as the safety management method of a hydrogen energy storage system provided in the above embodiment, and will not be repeated in the embodiment of the present invention.

[0054] The optional implementation modes of the embodiments of the present invention are described in detail above in conjunction with the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above implementation modes. Within the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical scheme of the embodiments of the present invention, and these simple modifications all belong to the protection scope of the embodiments of the present invention.

[0055] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not further describe various possible combinations.

[0056] Those skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a program, and the program is stored in a storage medium, including several instructions to enable a single-chip microcomputer, a chip or a processor to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0057] In addition, various implementation modes of the embodiments of the present invention may be arbitrarily combined, and as long as they do not violate the concept of the embodiments of the present invention, they should also be regarded as the contents disclosed by the embodiments of the present invention.

Claims

1. A safety management method for a hydrogen energy storage system, characterized in that: include: Acquiring operating status data of the hydrogen energy storage system; extracting global features from the operating status data; Extracting energy propagation features from the operating state data based on the energy transmission path, extracting hydrogen propagation features from the operating state data based on the hydrogen transmission path, and fusing the energy propagation features and the hydrogen propagation features to obtain local features; The local feature and the global feature are fused to obtain an enhanced global feature; Inputting the enhanced global features and the local features into a preset risk assessment model to obtain a risk assessment result; Generate alarm information based on risk assessment results.

2. A safety management method for a hydrogen energy storage system according to claim 1, characterized in that: Extracting global features from the operating status data includes: extracting the operation mode and the importance of the equipment from the operation status data; Determining a device weight based on the operation mode and the device importance; The operating status data is fused based on the device weight to obtain a global feature.

3. A safety management method for a hydrogen energy storage system according to claim 1, characterized in that: Extracting energy propagation features from the operating status data based on the energy transmission path includes: Based on the energy transmission process of the hydrogen energy storage system, construct an energy transmission path; extracting current data and voltage data of nodes of the energy transmission path from the operating status data based on the energy transmission path; Performing feature reconstruction on the current data and the voltage data to obtain an energy propagation matrix; The energy propagation matrix is ​​convolved to obtain energy propagation features.

4. A safety management method for a hydrogen energy storage system according to claim 3, characterized in that: The current data and the voltage data are subjected to feature reconstruction to obtain an energy propagation matrix, including: Performing FFT analysis on the current data, and determining the harmonic distortion rate according to the FFT analysis result; Performing wavelet decomposition on the current data and the voltage data respectively, and determining the current spectrum energy proportion and the current spectrum energy proportion according to the wavelet decomposition results; The node weights of the nodes of each energy transmission path are determined according to the energy transmission path, and an energy propagation matrix is ​​constructed based on the node weights, the harmonic distortion rate, the current spectrum energy proportion and the current spectrum energy.

5. A safety management method for a hydrogen energy storage system according to claim 1, characterized in that: Extracting hydrogen propagation characteristics from the operating status data based on the hydrogen transmission path includes: Establishing a hydrogen transmission path based on the transmission process of hydrogen in the hydrogen energy storage system; extracting a pressure signal and a flow signal from the operating status data based on the hydrogen transmission path; Performing feature reconstruction on the flow signal and the pressure signal to obtain a hydrogen propagation matrix; The hydrogen propagation matrix is ​​convolved to obtain hydrogen propagation features.

6. A safety management method for a hydrogen energy storage system according to claim 5, characterized in that: The flow signal and the pressure signal are subjected to feature reconstruction to obtain a hydrogen propagation matrix, including: determining a flow rate fluctuation rate according to the flow signal, and determining a pressure gradient according to the pressure signal; Inputting the flow signal and the pressure signal into a preset physical information neural network model to obtain a predicted flow fluctuation rate and a predicted pressure gradient; A hydrogen propagation matrix is ​​generated according to the flow fluctuation rate, the pressure gradient, the predicted flow fluctuation rate and the predicted pressure gradient.

7. A safety management method for a hydrogen energy storage system according to claim 1, characterized in that: The energy propagation feature and the hydrogen propagation feature are merged to obtain a local feature, further comprising: When the system device is in cluster mode, extracting operating power from the operating status data; Calculate the load rate of the system equipment according to the operating power, and calculate the load balancing index between the equipment according to the load rate; generating a load balancing feature based on the inter-device load balancing index; The inter-device load balancing feature, the energy propagation feature and the hydrogen propagation feature are integrated to obtain a local feature.

8. A safety management system for a hydrogen energy storage system, characterized in that: The system comprises: An acquisition module, used to acquire operating status data of the hydrogen energy storage system; A global feature extraction module, used to extract global features from the operating status data; A local feature extraction module, configured to extract energy propagation features from the operating state data based on an energy transmission path, extract hydrogen propagation features from the operating state data based on a hydrogen transmission path, and fuse the energy propagation features with the hydrogen propagation features to obtain local features; A global feature enhancement module, used for fusing the local feature with the global feature to obtain an enhanced global feature; A risk assessment module, used for inputting the enhanced global features and the local features into a preset risk assessment model to obtain a risk assessment result; The alarm module is used to generate alarm information based on risk assessment results.

Images