Hydrogen filling station safety interlock monitoring method and system

Abstract

The application relates to the technical field of intelligent control, and discloses a hydrogen filling station safety interlocking monitoring method and system, the method comprising: generating an abnormal state identification set from multi-source monitoring data through feature extraction, and analyzing the space-time correlation thereof in combination with process layout to intelligently deduce an abnormal conduction path. Further, by performing integral operation on parameter changes on the path, risk diffusion characteristics are quantitatively obtained, and safety operation instructions are dynamically matched and arranged according to the risk diffusion characteristics to generate a disposal sequence. Disturbance data and environmental parameters after execution of the instructions are synchronously collected by the system, state feedback data are formed after fusion processing, and the abnormal conduction path is iteratively updated by using the state feedback data, so that closed-loop perception and optimization of the risk conduction situation are finally realized; and the application can improve the self-adaptive capacity of hydrogen filling station safety interlocking monitoring.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, and in particular to a safety interlock monitoring method and system for hydrogen filling stations. Background Technology

[0002] In existing safety monitoring technologies for hydrogen filling stations, the processing of multi-source heterogeneous safety monitoring data often remains at the level of independent threshold alarms and simple logical interlocks. This method can only identify local parameter exceedances and cannot comprehensively analyze the transmission relationship of abnormal states from the perspectives of spatial topology and temporal correlation. As a result, there is a lack of effective judgment on the evolution path and diffusion trend of risks, making it difficult to grasp the dynamic safety situation within the station as a whole.

[0003] Meanwhile, existing technologies rely on pre-set fixed procedures for safety responses, which lack sufficient alignment between response logic and real-time risk scenarios, and cannot be dynamically adjusted or updated based on the actual effects of the responses. This limits the targetedness, timeliness, and adaptability of safety interlocking actions, and the system as a whole lacks the ability to self-optimize and iterate based on feedback data, resulting in lagging prevention and control and low efficiency when dealing with complex and cascading safety risks. Summary of the Invention

[0004] This invention provides a method and system for monitoring safety interlocks at hydrogen filling stations, the main purpose of which is to solve the problem of insufficient adaptive capability in monitoring safety interlocks at hydrogen filling stations.

[0005] To achieve the above objectives, the present invention provides a safety interlock monitoring method for hydrogen filling stations, comprising:

[0006] Feature extraction is performed on the multi-source heterogeneous safety monitoring data of the filling station to obtain the abnormal state identifier set of the filling station;

[0007] By analyzing the topological and temporal correlations of the abnormal status identifier set in spatial location based on the process layout of the filling station, the abnormal propagation path of the filling station can be obtained;

[0008] By performing path integration on the rate of change and direction of change of the parameters corresponding to the anomaly identifier in the anomaly propagation path, the risk diffusion characteristics of the filling station can be obtained;

[0009] Based on the risk diffusion characteristics, the corresponding safety operation instructions are matched to the risk scenarios constructed, and the safety operation instructions are cascaded and arranged to obtain the dynamic safety handling sequence of the filling station;

[0010] The disturbance process data and environmental parameters generated during the execution of the dynamic safety handling sequence are fused and processed to obtain the status feedback data of the filling station;

[0011] Based on the status feedback data, the abnormal transmission path is iteratively updated to obtain the risk transmission status of the filling station.

[0012] In a preferred embodiment, the step of extracting features from the multi-source heterogeneous safety monitoring data of the filling station to obtain the abnormal state identifier set of the filling station includes:

[0013] By acquiring equipment status parameters and environmental monitoring parameters, multi-source heterogeneous safety monitoring data of the filling station can be obtained;

[0014] The multi-source heterogeneous safety monitoring data is subjected to sliding window normalization to obtain the time-series standardized parameters of the filling station;

[0015] Abnormal parameter fragments of the filling station are obtained by performing abrupt change detection on the time-series standardized parameters;

[0016] Assign an abnormal identifier to the abnormal parameter fragment, and associate and store the abnormal identifier, the corresponding parameter type, the amount exceeding the threshold, and the timestamp to obtain the abnormal status identifier set of the filling station.

[0017] In a preferred embodiment, the step of analyzing the topological and temporal correlations of the abnormal state identifier set in spatial location, combined with the process layout of the filling station, to obtain the abnormal propagation path of the filling station, includes:

[0018] The abnormal status identifier set is mapped to the process piping and equipment layout diagram of the filling station to obtain the corresponding physical location coordinates;

[0019] Based on the physical connection relationships between the process pipelines and the equipment in the equipment layout diagram, construct an intra-station topology network diagram with the equipment in the equipment layout diagram and the process pipelines as vertices and the physical connection relationships as edges;

[0020] Based on the physical location coordinates, the transmission calculation of the process nodes in the station's internal topology network diagram is performed to obtain the abnormal transmission path of the filling station.

[0021] In a preferred embodiment, the step of performing transmission calculations on process nodes in the station's internal topology network diagram based on the physical location coordinates to obtain the abnormal transmission path of the filling station includes:

[0022] In the station topology network diagram, starting from the physical location coordinates, the probability intensity of the propagation of abnormal states from the abnormal state identifier set to adjacent nodes is calculated;

[0023] Based on the occurrence order and time interval represented by the timestamps of the anomaly identifiers in the anomaly status identifier set, a corresponding time-series correlation index is assigned to each anomaly identifier pair;

[0024] Based on the probability intensity and the temporal correlation index, high-risk transmission sequences in the station's internal topology network diagram are identified, and abnormal transmission paths of the filling station are obtained.

[0025] In a preferred embodiment, the step of performing path integration on the rate and direction of change of the parameter corresponding to the anomaly identifier in the anomaly propagation path to obtain the risk diffusion characteristics of the filling station includes:

[0026] Extract the risk sequence of the parameters corresponding to the anomaly identifier within the time window;

[0027] The first derivative of the risk sequence is used as the rate of change vector of the filling station;

[0028] Determine the spatial connection direction vector between adjacent anomaly identifiers on the anomaly propagation path based on the station topology network diagram;

[0029] The intensity integral value of the filling station is obtained by integrating the rate of change vector based on the spatial connection direction vector.

[0030] By binding the intensity integral value with the endpoint location information of the abnormal transmission path, the risk diffusion characteristics of the filling station are obtained.

[0031] In a preferred embodiment, the formula for calculating the intensity integral value includes:

[0032]

[0033] in, The intensity integral value, For path segment index, This represents the total number of path segments. For the first Equipment-based risk factors for each path segment For the first The abnormal state evolution factor of each path segment, For the first The rate of change vector of each path segment For the first Spatial connection direction vectors of path segments, The intensity coefficient between adjacent path segments. This is the saturation effect adjustment coefficient. It is the hyperbolic tangent function. For the first The vector of the rate of change of parameters at each path segment For the first Spatial connection direction vectors of parameters at each path segment.

[0034] In a preferred embodiment, the step of matching the corresponding safety operation instructions to the risk scenario constructed based on the risk diffusion characteristics, and cascading the safety operation instructions to obtain the dynamic safety handling sequence of the filling station, includes:

[0035] Based on the accident data, equipment process characteristics, and safety requirements of the filling station, a risk scenario template for the filling station is constructed;

[0036] The risk accumulation intensity and potential risk arrival location in the risk diffusion characteristics are matched with the risk scenario template to filter out the target risk scenarios of the filling station;

[0037] Retrieve basic safety operation instructions associated with the target risk scenario from the safety operation instruction library of the filling station;

[0038] Based on the process logic dependencies and equipment action response time of the filling station, the basic safety operation instructions are logically verified to obtain the dynamic safety handling sequence of the filling station.

[0039] In a preferred embodiment, the step of fusing the disturbance process data and environmental parameters generated by executing the dynamic safety handling sequence to obtain the status feedback data of the filling station includes:

[0040] Synchronously collect equipment status change data affected by the dynamic safety response sequence;

[0041] Extract environmental monitoring parameters from key monitoring points in the filling station;

[0042] The equipment status change data and the environmental monitoring parameters are spatiotemporally aligned to obtain the fusion monitoring sequence of the filling station;

[0043] The deviation of the filling station is obtained by performing differential calculation on the fusion monitoring sequence based on the dynamic safety handling sequence.

[0044] The deviation, the state change data, and the environmental monitoring parameters are collectively encapsulated into the state feedback data of the filling station.

[0045] In a preferred embodiment, the step of iteratively updating the abnormal transmission path based on the status feedback data to obtain the risk transmission status of the filling station includes:

[0046] Based on the statistical characteristics in the status feedback data, the transmission probability weights of the corresponding nodes in the abnormal transmission path are corrected to obtain the target weight of the filling station;

[0047] High-risk transmission edges are selected in the site topology network diagram based on the target weight;

[0048] The visualized network diagram containing the abnormal transmission path and the selected high-risk transmission edge is used as the risk transmission status of the filling station.

[0049] To address the above problems, the present invention also provides a safety interlock monitoring system for hydrogen filling stations, the system comprising:

[0050] The abnormal status module extracts features from the multi-source heterogeneous safety monitoring data of the filling station to obtain the abnormal status identifier set of the filling station;

[0051] The anomaly propagation path module analyzes the topological and temporal correlations of the anomaly status identifier set in spatial location, combined with the process layout of the filling station, to obtain the anomaly propagation path of the filling station;

[0052] The risk feature module performs path integration on the rate and direction of change of the parameters corresponding to the anomaly identifiers in the anomaly propagation path to obtain the risk diffusion characteristics of the filling station;

[0053] The safety handling sequence module matches corresponding safety operation instructions to the risk scenarios constructed based on the risk diffusion characteristics, and cascades and arranges the safety operation instructions to obtain the dynamic safety handling sequence of the filling station;

[0054] The status feedback module fuses the disturbance process data and environmental parameters generated by executing the dynamic safety handling sequence to obtain the status feedback data of the filling station.

[0055] The risk transmission status module iteratively updates the abnormal transmission path based on the status feedback data to obtain the risk transmission status of the filling station.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] 1. This invention achieves dynamic, quantitative, and spatial assessment of safety risks at hydrogen filling stations by constructing an intelligent analysis chain covering multi-source monitoring data, an internal station topology network, and a safety knowledge base. Its beneficial effects lie in the system's ability to identify abnormal states in real time, automatically deduce their propagation paths within the process network, and accurately calculate the cumulative intensity and expected arrival location of risks along the path. This elevates discrete alarm signals to a deep understanding and predictive judgment of risk evolution, significantly improving the accuracy of safety monitoring and the foresight of decision-making.

[0058] 2. This invention further matches real-time risk characteristics with pre-set risk scenario templates to automatically generate dynamic safety response sequences that highly match the current situation, and iteratively updates the risk transmission model based on multi-dimensional feedback data after execution. This process transforms safety response from passive alarms and fixed plan execution to proactive closed-loop protection based on real-time situational awareness, intelligent decision-making, and continuous self-learning, thereby effectively improving the accuracy, adaptability, and overall reliability of safety interlocking control under complex operating conditions. Attached Figure Description

[0059] Figure 1 This is a flowchart illustrating a safety interlock monitoring method for hydrogen filling stations according to an embodiment of the present invention.

[0060] Figure 2 This is a functional block diagram of a hydrogen filling station safety interlock monitoring system provided in an embodiment of the present invention;

[0061] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0062] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0063] This application provides a method for monitoring the safety interlock of a hydrogen filling station. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0064] Reference Figure 1 The diagram shown is a flowchart illustrating a safety interlock monitoring method for hydrogen filling stations according to an embodiment of the present invention. In this embodiment, the safety interlock monitoring method for hydrogen filling stations includes:

[0065] In this embodiment of the invention, when extracting features from the multi-source heterogeneous safety monitoring data of the filling station to obtain the abnormal state identifier set of the filling station, the specific steps are as follows:

[0066] By acquiring equipment status parameters and environmental monitoring parameters, multi-source heterogeneous safety monitoring data of the filling station can be obtained;

[0067] The multi-source heterogeneous safety monitoring data is subjected to sliding window normalization to obtain the time-series standardized parameters of the filling station;

[0068] Abnormal parameter fragments of the filling station are obtained by performing abrupt change detection on the time-series standardized parameters;

[0069] Assign an abnormal identifier to the abnormal parameter fragment, and associate and store the abnormal identifier, the corresponding parameter type, the amount exceeding the threshold, and the timestamp to obtain the abnormal status identifier set of the filling station.

[0070] Specifically, real-time values ​​are directly read from sensors installed on the process equipment of the hydrogen filling station. These sensors include pressure transmitters, temperature sensors, flow meters, valve opening feedback devices, and combustible gas concentration detectors. At the same time, ambient temperature, wind speed, and humidity data are obtained from environmental monitoring units deployed in the station area. All these real-time read values ​​are aggregated according to a preset acquisition frequency to form multi-source heterogeneous safety monitoring data that includes equipment status parameters and environmental monitoring parameters.

[0071] Specifically, a sliding window with a fixed time length is used to process multi-source heterogeneous safety monitoring data. An independent time window is set for each type of parameter. This window contains all sampling points within the current time and a continuous period before it. The maximum and minimum values ​​of the data within this window are calculated.

[0072] Specifically, the difference between two adjacent data points in the time-series standardized parameter sequence is calculated sequentially. When the absolute value of the difference exceeds a fixed threshold obtained from the statistical analysis of the parameter's historical stable operation data, a sudden change is determined to have occurred at that moment.

[0073] Specifically, a globally unique anomalous identifier is assigned to each identified anomalous parameter fragment. This identifier is generated by a combination of letters and numbers. At the same time, the original parameter type that generated the fragment is recorded. The difference between the average value of all data points in the anomalous parameter fragment and the preset safety threshold is calculated to obtain the amount exceeding the threshold. The timestamp of the first time the fragment triggers mutation detection is accurately recorded.

[0074] Furthermore, the data acquisition modules deployed on-site convert various sensor signals into digital signals, add precise timestamps, organize them in a structured manner according to data source type and device number, and transmit them to the real-time database of the central monitoring server for storage.

[0075] Furthermore, for each original data point within the window, it is linearly scaled to a range of zero to one by subtracting the minimum value of the window and then dividing by the range of the window. This process is called sliding window normalization. After this process, all parameters, regardless of their original dimensions and numerical range, are transformed into a standardized numerical sequence that fluctuates within the same amplitude range, thus obtaining the time-series standardized parameters of the filling station.

[0076] Furthermore, starting from this mutation point, the data is continuously checked backward until the absolute value of the difference between multiple consecutive data points falls back to within the threshold. At this point, the end point of the sequence is marked, and the parameter data between the start point and the end point is extracted, which is the abnormal parameter segment of the filling station.

[0077] Furthermore, a one-to-one correspondence is established between the four elements: anomaly identifier, parameter type, threshold exceedance, and timestamp. This set of correspondences is stored as a complete record entry in a dedicated anomaly status record table. The collection of all such entries constitutes the anomaly status identifier set of the filling station.

[0078] In summary, by employing a sliding window mechanism to locally normalize multi-source heterogeneous safety monitoring data, the analytical obstacles caused by differences in dimensions and numerical ranges of different monitoring parameters (such as pressure, temperature, and concentration) are effectively eliminated. It dynamically scales the historical data of each parameter within the window to a uniform range, enabling subsequent algorithms to process signals from different devices equally. This lays a reliable data foundation for accurately identifying global or local anomalies and improves the consistency of condition assessment.

[0079] In summary, abrupt change detection of standardized time-series parameters can accurately capture drastic changes or deviations from normal patterns in parameter values ​​within a short period. This method directly targets the most critical instantaneous anomalies in safety monitoring, avoiding interference from slow drift or normal fluctuations. It efficiently and accurately locates and extracts truly representative anomaly parameter segments from the continuous data stream, significantly improving the sensitivity and accuracy of initial warnings.

[0080] In summary, by assigning a unique identifier to each anomaly fragment and associating it with key attributes, unstructured raw anomaly data is transformed into structured knowledge entries. The resulting set of anomaly status identifiers for filling stations not only fully records specific core information but also integrates it into standardized objects that can be directly accessed for subsequent topology and time-series analysis. This greatly facilitates the retrieval, correlation analysis, and tracing of anomaly data, providing clear and standardized input for subsequent anomaly propagation path analysis and risk quantification. It is a crucial link in the entire dynamic intelligent monitoring chain.

[0081] In this embodiment of the invention, when analyzing the topological and temporal correlations of the abnormal state identifier set in spatial location based on the process layout of the filling station to obtain the abnormal propagation path of the filling station, it is specifically used for:

[0082] The abnormal status identifier set is mapped to the process piping and equipment layout diagram of the filling station to obtain the corresponding physical location coordinates;

[0083] Based on the physical connection relationships between the process pipelines and the equipment in the equipment layout diagram, construct an intra-station topology network diagram with the equipment in the equipment layout diagram and the process pipelines as vertices and the physical connection relationships as edges;

[0084] Based on the physical location coordinates, the transmission calculation of the process nodes in the station's internal topology network diagram is performed to obtain the abnormal transmission path of the filling station.

[0085] Specifically, the parameter type information of each record in the abnormal status identifier set is extracted, and the unique installation point number of the sensor or monitoring device that generated the abnormal parameter is determined in the filling station based on the pre-established parameter type and monitoring point number association table.

[0086] Specifically, after obtaining the physical location coordinates, the system reads a topology description file of the same process piping and equipment layout diagram. This file explicitly defines the physical connections between all equipment within the station, between equipment and piping, and between piping in a list format. Based on this description file, the system abstracts each independent equipment entity in the layout diagram, such as compressors, hydrogen storage tank groups, hydrogen dispensers, etc., and each section of process piping, into a point, called a vertex.

[0087] Specifically, starting from this initial point, and connecting all adjacent vertices directly via edges, a basic probability value is pre-set for each connecting edge, representing the likelihood of an abnormal state propagating along this path, based on process principles and historical accident data. During calculation, the system compares the parameter type represented by the current abnormal state identifier with the edge's attributes; if a match is found, the basic probability value is invoked.

[0088] Furthermore, the monitoring system loads the process piping and equipment layout diagrams of the filling station stored in the digital asset library. These diagrams are two-dimensional or three-dimensional digital drawings with precise dimensions. Each equipment symbol and pipe segment in the drawing is associated with a unique identifier, and each identifier corresponds to a set of coordinate values ​​in the coordinate system of the drawing. By matching the monitoring point number with the equipment or pipe identifier in the drawing, the system automatically retrieves and reads the corresponding coordinate values ​​from the drawing database, assigns these coordinate values ​​to the corresponding record in the abnormal status identifier set, thereby completing the mapping from abstract data records to specific spatial locations, and ultimately obtaining the physical location coordinates of each abnormal status on the layout diagram.

[0089] Furthermore, each physical connection defined in the file, such as "compressor A outlet flange connects to pipe P1" and "valve V1 is installed on pipe P1," is abstracted as a line segment connecting two vertices, called an edge. The system uses graph theory data structures to construct a network model composed of these vertices and edges in memory. This complete and accurate network model reflecting the physical connection structure of the filling station is constructed as the station's internal topology network graph, which depicts the spatial connectivity possibilities of all process units within the station.

[0090] Furthermore, the system iterates through the set of abnormal state identifiers, calculates the time interval between each pair of abnormal events based on the chronological order of their timestamps, and maps short time intervals to high temporal correlation strength values ​​according to preset rules. For each possible propagation path formed by the starting point and its adjacent vertices, the system multiplies the basic probability value of the path with the temporal correlation strength values ​​of multiple abnormal points along the path to obtain a comprehensive propagation evaluation value. Among all possible paths originating from the starting point, the system selects a sequence of consecutive vertices and edges whose comprehensive propagation evaluation value exceeds a set threshold. This sequence represents the most likely physical route for the current abnormality to spread, which is the abnormal propagation path of the filling station.

[0091] In summary, by precisely associating the abstract set of abnormal status identifiers with the specific process piping and equipment layout of the filling station, each abnormal data point is given clear spatial location information. This achieves a transformation from signal space to physical space, enabling subsequent risk analysis to be based on the actual equipment layout and geographical location. This lays a solid foundation for accurately understanding the specific equipment node where the anomaly occurred and assessing its spatial impact range within the station, overcoming the problem of disconnect between alarm information and physical location in traditional monitoring.

[0092] In summary, based on the physical connections between equipment, the entire filling station's process structure is abstracted as a network graph consisting of vertices and edges. This network graph accurately depicts all connected paths through which physical quantities such as hydrogen, pressure, and temperature may be conducted or diffused within the station, providing a complete structural model for systematically analyzing how abnormal states propagate along process pipelines and the equipment network. It connects isolated equipment into a unified analytical object, making the analysis of cascading risks possible.

[0093] In summary, by performing propagation calculations based on physical location coordinates and the station's internal topology network diagram, the most likely propagation or transmission route of the current abnormal state can be dynamically deduced, thus obtaining the abnormal propagation path of the filling station. This represents a leap from identifying "point anomalies" to predicting "line risks" and even "area situations," giving safety monitoring a certain degree of foresight. It helps operators and management systems to pay attention to the direction and key nodes where risks may spread in advance, providing direct and crucial decision-making basis for formulating accurate and proactive safety response strategies.

[0094] In this embodiment of the invention, when performing transmission calculations on process nodes in the station's internal topology network diagram based on the physical location coordinates to obtain the abnormal transmission path of the filling station, the specific method is as follows:

[0095] In the station topology network diagram, starting from the physical location coordinates, the probability intensity of the propagation of abnormal states from the abnormal state identifier set to adjacent nodes is calculated;

[0096] Based on the occurrence order and time interval represented by the timestamps of the anomaly identifiers in the anomaly status identifier set, a corresponding time-series correlation index is assigned to each anomaly identifier pair;

[0097] Based on the probability intensity and the temporal correlation index, high-risk transmission sequences in the station's internal topology network diagram are identified, and abnormal transmission paths of the filling station are obtained.

[0098] Specifically, starting from the starting point, all adjacent vertices directly connected by edges are traversed; for each edge connecting the starting point and an adjacent vertex, the system retrieves a pre-set process knowledge base, which defines a basic transmission probability value for the possibility of different types of abnormal states propagating along a specific edge based on equipment type, pipe material, medium flow direction and historical fault data.

[0099] Specifically, the timestamps of all exception identifiers in the exception status identifier set are read and strictly sorted from earliest to latest to obtain the clear order of occurrence of the exception events. The specific method for assigning the time-series correlation index is that the system processes two adjacent exception identifiers in the sorted list in sequence to form an exception identifier pair; the difference between the two timestamps in the pair is calculated to obtain the precise time interval.

[0100] Specifically, starting from the starting point, a path search is performed in the station's topology network graph; for each potential path found, the system extracts the probability intensity of each edge segment on the path and multiplies these probability intensity values ​​to obtain the overall propagation probability base value of the path.

[0101] Furthermore, the parameter type recorded by the current abnormal state identifier is matched with the entries in the knowledge base to extract the corresponding basic transmission probability value. Then, the system checks the real-time state attributes of the edge, such as whether the valve is open or whether the pipeline has a blockage alarm. If the state is abnormal, the basic probability value is increased according to the rules; if the state is normal, the original value is maintained. This value after real-time state correction is the probability intensity of the abnormal state propagating from the starting point to the specific neighboring node. The system calculates and stores a probability intensity value for each edge starting from the starting point, forming a quantitative set of propagation possibilities from the starting point to each neighboring node.

[0102] Furthermore, a preset time window threshold is defined. If the calculated time interval is less than or equal to this threshold, it indicates that the two anomalous events occurred in close succession, and the system assigns a higher temporal correlation index to the pair. If the time interval is greater than the threshold, a lower temporal correlation index is assigned. The index value is represented by a specific numerical value, such as a range from zero to one. The shorter the time interval, the closer the index value is to one, indicating a stronger temporal correlation. The system performs this calculation and assignment operation on all sequentially adjacent pairs of anomalous identifiers, ultimately generating a corresponding temporal correlation index for each pair and recording the association between this index and the two anomalous identifiers involved.

[0103] Furthermore, the system examines the vertex sequence traversed by the path. If the abnormal events corresponding to adjacent vertices in the sequence exist in the abnormal identifier set and form an abnormal identifier pair, the system extracts the temporal correlation index of the pair and multiplies these index values ​​to obtain the path's temporal correlation factor. The system multiplies the overall propagation probability base value of the path with the temporal correlation factor to obtain a comprehensive risk score. The system traverses all possible paths originating from the starting point, calculates the comprehensive risk score for each path, and then filters out paths whose comprehensive risk scores exceed a preset risk threshold. These filtered paths are sorted from highest to lowest score, and the vertex and edge sequence represented by the path with the highest score is identified as a high-risk propagation sequence in the site's topology network graph.

[0104] In summary, based on the station's topology network diagram and physical location coordinates, the probability of abnormal states propagating to adjacent equipment or pipeline nodes is quantitatively calculated. By combining basic equipment risk factors with real-time status to correct the preset propagation probability, this calculation transforms process knowledge and real-time monitoring data into specific and comparable probability strength values. This achieves an objective measurement of the spatial diffusion capacity of risk, avoiding the one-sidedness of relying solely on experience, and provides accurate numerical basis for subsequent screening of key propagation paths, enabling the system to prioritize physically connected directions with a high probability of risk propagation.

[0105] In summary, by analyzing the occurrence order and interval of the timestamps of anomalous state identifiers, a temporal correlation index is assigned to consecutive pairs of anomalous events. This operation links temporal proximity with causal probability, effectively distinguishing between random independent events and consecutive events that are closely related in time and may have a causal chain. It enhances the system's ability to identify the temporal coherence of anomalous evolution processes and provides crucial temporal correlation evidence for determining whether a series of anomalies constitutes a continuous diffusion process.

[0106] In summary, by combining probability intensity and temporal correlation index, path search and scoring are performed in the site's topology network diagram to ultimately identify high-risk transmission sequences as abnormal transmission paths. This process organically combines spatial transmission probability with temporal occurrence sequence, enabling intelligent reasoning and localization of dynamic risk diffusion paths. Its effectiveness lies in automatically and accurately outlining the most threatening risk transmission chain from complex network topology and a large amount of alarm information, thereby transforming multiple abstract anomalies into a specific, directional path. This greatly focuses the attention of monitoring and response, laying a core decision-making foundation for implementing precise and efficient interceptive security measures.

[0107] In this embodiment of the invention, when performing path integration on the rate and direction of change of the parameter corresponding to the anomaly identifier in the anomaly propagation path to obtain the risk diffusion characteristics of the filling station, it is specifically used for:

[0108] Extract the risk sequence of the parameters corresponding to the anomaly identifier within the time window;

[0109] The first derivative of the risk sequence is used as the rate of change vector of the filling station;

[0110] Determine the spatial connection direction vector between adjacent anomaly identifiers on the anomaly propagation path based on the station topology network diagram;

[0111] The intensity integral value of the filling station is obtained by integrating the rate of change vector based on the spatial connection direction vector.

[0112] By binding the intensity integral value with the endpoint location information of the abnormal transmission path, the risk diffusion characteristics of the filling station are obtained.

[0113] Specifically, the system extracts all sampled data of a specific monitoring parameter within a set fixed-length time window from the real-time historical database of the filling station, based on the unique identifier pointing to that parameter. This time window traces back a continuous period of time, ending at the current calculation time.

[0114] Specifically, to perform calculations on two adjacent data points in the obtained risk sequence, the risk value of the subsequent data point is subtracted from the risk value of the preceding data point to obtain a difference; this difference is then divided by the time interval between the subsequent and preceding data points to calculate an average rate of change.

[0115] Specifically, the precise coordinates of each pair of adjacent vertices on the path in the digital coordinate system of the filling station layout are obtained, and the vector pointing from the coordinates of the previous vertex to the coordinates of the next vertex is calculated. The system extracts the three coordinate components of this vector that represent the direction in space, and by calculating the length of the vector, divides each coordinate component by the length to obtain a unit direction vector with a fixed length.

[0116] Specifically, along the abnormal propagation path, each segment of the path is processed sequentially. For each segment, the system takes the instantaneous rate of change value in the rate of change vector corresponding to the starting time of the segment and performs a dot product operation with the spatial connection direction vector of the segment. This operation projects the rate value onto the spatial direction to obtain a scalar value.

[0117] Specifically, the device number corresponding to the last vertex in the anomaly propagation path and its physical location coordinates in the layout diagram are obtained as the endpoint location information. The specific method for binding the intensity integral value with the endpoint location information is that the system creates a structured data record.

[0118] Furthermore, the system precisely extracts the parameter values ​​corresponding to each sampling time point within this window and arranges these values ​​into an ordered list according to the timestamp from earliest to latest. Each value in this list represents the original measurement value of the parameter at the corresponding time. This list of original parameter values ​​arranged in chronological order is the basic data set for subsequent processing. The system will transform this list to characterize the risk level.

[0119] Furthermore, from the start point to the end point of the sequence, this subtraction and division operation is repeated for each pair of adjacent data points to obtain a series of instantaneous rate of change values ​​arranged by time points. These instantaneous rate of change values ​​are stored sequentially in an array, where each element corresponds to a specific time point, representing the rate and direction of change of the risk value over time at that moment; that is, positive values ​​indicate an increase, and negative values ​​indicate a decrease. This array is the rate of change vector of the filling station.

[0120] Furthermore, this unit direction vector uniquely represents the spatial path direction from the previous anomaly location to the next anomaly location. A unit direction vector is calculated for each adjacent vertex on the anomaly propagation path in this way. The ordered set of these vectors constitutes a spatial connection direction vector group used to describe the spatial orientation of the diffusion path.

[0121] Furthermore, this scalar value represents the intensity component of the risk change along the spatial direction at that moment on that path segment. The system sequentially adds up these intensity components calculated for all segments along the path. This summation process simulates the continuous calculation of the risk change intensity along the entire spatial path, and the final summation result is the intensity integral value of the filling station, which quantifies the comprehensive intensity of risk diffusion along a specific path.

[0122] Furthermore, the record contains two main fields: the first field stores the specific numerical value of the calculated intensity integral; the second field stores the endpoint location information, including the unique identifier of the endpoint device and its coordinates. The system saves and outputs this data record as an indivisible whole data object. This data object, which simultaneously contains the cumulative risk intensity value and the spatial location of the risk's final impact, is defined as the risk diffusion characteristic of the filling station. It fully describes the intensity and final destination of the current anomaly spreading along a specific path.

[0123] In summary, time-series data of parameters directly related to anomaly identifiers are extracted from historical and real-time data to construct a risk sequence. This enables continuous tracking and quantitative characterization of specific risk points, transforming the abstract "anomaly" into an analyzable and calculable data trajectory. By defining a clear time window, this step focuses on the critical periods before and after the anomaly occurs, effectively filtering out irrelevant historical data and providing high-quality, highly relevant input data for subsequent accurate analysis of the dynamic evolution of risks, laying the foundation for quantitative analysis.

[0124] In summary, by calculating the first derivative of the risk sequence, a rate of change vector is obtained. This achieves a crucial transformation from describing the "state" of risk to describing the "trend and severity of risk change." This vector not only quantifies the speed of change of risk parameters, but its direction also more intuitively reveals whether the risk is intensifying or mitigating. This step enables the system to sensitively capture the dynamic characteristics of risk, incorporating the dynamic evolution speed of risk into the assessment system, and providing a core dynamic indicator for predicting its future development trend.

[0125] In summary, based on the site's internal topology network diagram, the spatial relationships between points along the anomaly propagation path are quantified as direction vectors. This achieves a geometric description of the physical path of risk diffusion, transforming "connection" relationships into mathematical objects with direction and magnitude. By clearly defining the specific direction of each propagation path segment in three-dimensional space, this step seamlessly embeds spatial topology information into subsequent mathematical calculations, ensuring that the calculation of risk intensity accumulation is performed along the actual physical spatial path, giving the assessment results clear physical spatial meaning.

[0126] In summary, the intensity integral value is calculated by integrating the rate of change along the spatial connection direction. This calculation organically combines the "speed of risk change" with the "path of spatial transmission," enabling a comprehensive quantitative assessment of the overall intensity of risk accumulation and diffusion along a specific path. Its effectiveness lies in overcoming the limitations of single-point instantaneous rate assessment, reflecting the continuous cumulative effect and path-dependent characteristics of risk during spatial transmission, and ultimately outputting a scalar value characterizing the severity of risk along the entire path, providing a unified quantitative basis for comparing and classifying different risk scenarios.

[0127] In summary, by binding the intensity integral value representing the comprehensive risk of a path with the path's endpoint location information, a complete risk diffusion characteristic is formed. This creates a composite data object that simultaneously contains "how great the risk is" and "where the risk is going." Its effect is that it ultimately integrates the preceding quantitative analysis with spatial positioning, providing a direct and clear target for risk management decisions. Based on this characteristic, the system can not only know the current comprehensive risk intensity level but also accurately pinpoint the expected location of risk spread, thereby guiding resources to be prioritized and protected at the most critical potentially affected points, achieving precision and spatialization in risk situation assessment.

[0128] In this embodiment of the invention, the formula for calculating the intensity integral value is specifically used for:

[0129]

[0130] in, The intensity integral value, For path segment index, This represents the total number of path segments. For the first Basic equipment risk factors for each path segment For the first The abnormal state evolution factor of each path segment, For the first The rate of change vector of each path segment For the first Spatial connection direction vectors of path segments, The intensity coefficient between adjacent path segments. This is the saturation effect adjustment coefficient. It is the hyperbolic tangent function. For the first The vector of the rate of change of parameters at each path segment For the first Spatial connection direction vectors of parameters at each path segment.

[0131] Specifically, the sources of the parameters in the formula are clearly defined. The equipment basic risk factor γ_i comes from the filling station equipment management database, which stores the model, service life, design safety level, and historical fault statistics of each piece of equipment. The system directly retrieves the corresponding value by matching the equipment number on the anomaly propagation path. The anomaly state evolution factor δ_i comes from the threshold exceedance amount and timestamp information of each record in the anomaly state identifier set. The system obtains this by calculating the product of the average magnitude of the threshold exceedance over the duration of the anomaly parameter segment and the duration, and then normalizing it. The rate of change vector v_i comes from the first derivative calculation of the risk sequence extracted from the parameters corresponding to the anomaly identifier within a fixed time window. That is, the difference between adjacent data points in the sequence is calculated sequentially and divided by the time interval to form an ordered list of values. The spatial connection direction vector u_i comes from the geometric data of the station's topology network diagram. The system calculates the unit vector from one vertex to the next based on the physical coordinates of the vertices corresponding to adjacent anomaly identifiers on the anomaly propagation path. The total number of path segments n and the path segment index i come from the structure of the anomaly propagation path itself. The path is sequentially divided into continuous segments, each segment connecting two adjacent anomaly identifiers. The intensity coefficient α and the saturation effect adjustment coefficient β between adjacent path segments are derived from the configuration file loaded during system initialization. The values ​​in this file are directly set as constants by safety engineers based on the results of simulation analysis of a large number of historical accidents.

[0132] Furthermore, the formula's significance lies in its mathematical calculation of the overall intensity of risk diffusion along a specific spatial path, outputting an intensity integral value S. The formula consists of two summation parts. The first part calculates the direct risk contribution of each independent segment along the path. This is achieved by multiplying the segment's equipment-based risk factor, abnormal state evolution factor, and the projection of the rate of change vector onto the spatial connection direction vector. The results for all segments are then summed. This reflects the static and dynamic accumulation of risk at each point along the path based on its own state and rate of change. The second summation part calculates the contribution of the interaction between adjacent path segments to the total intensity. This is achieved by inputting the projection value of the preceding segment into a hyperbolic tangent function, multiplying the output by the projection value of the following segment, and summing all such adjacent segment pairs. Finally, an intensity coefficient is multiplied. This simulates the triggering or amplifying effect of the preceding segment on the following segment during risk transmission, as well as the saturation characteristics of this effect. The significance of the entire formula is to integrate information from multiple dimensions, such as inherent risks of equipment, real-time evolution of anomalies, rate of parameter change, spatial transmission direction, and continuous influence of paths, into a single scalar indicator.

[0133] In general, the formula trend exhibits a specific behavioral pattern in which the intensity integral value changes with the input parameters. When the projection value of the rate of change vector of any path segment in the spatial direction increases, the intensity integral value tends to increase because this projection value directly appears as a multiplier in both summation parts. The introduction of the hyperbolic tangent function causes its output to approach 1 when the projection value is too large, resulting in the contribution of the interaction term between adjacent segments in the second part no longer increasing linearly with the projection value, thus producing a saturation effect and preventing excessive dominance of the overall result by local drastic changes. The magnitude of the intensity coefficient directly regulates the contribution ratio of the second part to the total intensity integral value; the larger the intensity coefficient, the more significant the risk enhancement effect brought about by path continuity. The equipment basic risk factor and the abnormal state evolution factor act as multipliers; their increase will linearly amplify the contribution value of the corresponding path segment. Overall, under the condition of enhanced parameters, the intensity integral value shows a monotonically increasing trend, but due to the nonlinearity of the hyperbolic tangent function, the growth rate slows down in the high-rate projection value range. The overall output is a weighted sum of the independent contributions of each path segment and the nonlinear correlation contributions between adjacent segments.

[0134] In this embodiment of the invention, when matching the corresponding safety operation instructions to the risk scenario constructed based on the risk diffusion characteristics, and cascading and arranging the safety operation instructions to obtain the dynamic safety handling sequence of the filling station, the specific usage is as follows:

[0135] Based on the accident data, equipment process characteristics, and safety requirements of the filling station, a risk scenario template for the filling station is constructed;

[0136] The risk accumulation intensity and potential risk arrival location in the risk diffusion characteristics are matched with the risk scenario template to filter out the target risk scenarios of the filling station;

[0137] Retrieve basic safety operation instructions associated with the target risk scenario from the safety operation instruction library of the filling station;

[0138] Based on the process logic dependencies and equipment action response time of the filling station, the basic safety operation instructions are logically verified to obtain the dynamic safety handling sequence of the filling station.

[0139] Specifically, all accident reports and alarm logs recorded during the historical operation of the filling station are collected. Key information is extracted from these unstructured documents using natural language processing, including the equipment that triggered the accident, the direct cause, the evolution process, and the final scope of impact, forming a structured accident case library. At the same time, the process characteristic specifications provided by the equipment manufacturers are integrated to extract the inherent attributes of each piece of equipment, such as the permissible operating pressure range, upper temperature limit, flow limit, and material compatibility.

[0140] Specifically, based on the potential arrival location information of the risk, a precise search is performed in the risk scenario template library to locate all template records that completely match that location. Next, the system compares the cumulative risk intensity value parsed from the risk diffusion characteristics with each predefined risk intensity level interval in these located template records to determine which specific interval the value falls into.

[0141] Specifically, the system maintains a safety operation instruction library for a filling station. This library is stored in a structured table format. Each basic safety operation instruction includes an instruction number, the target device, the specific control action, and one or more risk scenario numbers to which the instruction applies.

[0142] Specifically, the system loads a process logic dependency table for the filling station, which defines the sequential constraints between equipment actions. Simultaneously, the system includes built-in rated action response time data for each equipment actuator. The specific method for obtaining the dynamic safety handling sequence through logical verification is as follows: the system first sorts the retrieved set of basic safety operation instructions according to the process logic dependencies of their action objects, ensuring that all prerequisite action instructions are placed before the action instructions that depend on them.

[0143] Furthermore, mandatory clauses in safety technical specifications, such as safety distance requirements, interlocking cut-off conditions, and emergency response procedures, are transformed into explicit rule entries. The specific method for constructing risk scenario templates is as follows: the system uses the potential risk's final arrival location, such as a specific piece of equipment or pipeline segment, as a template index, creating a template record for each location. Within this record, based on the accident case library and process characteristics, one or more risk intensity level intervals are defined, each interval corresponding to a different description of potential consequences. And according to safety specification clauses, the system pre-associates the macro-level handling targets that should be triggered under this scenario, ultimately forming a structured data set indexed by location and containing intensity classifications and consequence contingency plans—that is, the risk scenario template for the filling station.

[0144] Furthermore, a strict one-to-one matching process is implemented, selecting only one template record whose cumulative risk intensity value falls completely within its predefined intensity range. This uniquely determined template record that simultaneously satisfies both location matching and intensity range matching is selected as the target risk scenario for the filling station. This scenario clearly indicates the risk point, intensity level, and preset consequences.

[0145] Furthermore, the specific method for retrieving basic safety operation instructions from the safety operation instruction library is as follows: The system reads the unique scenario number carried by the target risk scenario obtained in the previous step, uses this number as the query condition, and performs a full table scan and matching in the applicable risk scenario field of the safety operation instruction library. The system finds all record rows that contain the target risk scenario number in the applicable risk scenario field, and extracts the complete instruction content corresponding to these record rows, including the instruction number, action object, and specific control action, to form a temporary instruction set. Each instruction in this set is a basic safety operation instruction associated with the target risk scenario.

[0146] Furthermore, a theoretical execution start time is calculated for each ordered instruction: the start time of the first instruction is set to the current time; the start time of the next instruction is equal to the start time of the previous instruction plus the action response time of the device corresponding to that previous instruction. The system calculates recursively according to this rule, forming an ordered list containing the content of each instruction, a strict logical order, and a precise theoretical execution time. This ordered list of instructions, verified by logic and time, is the dynamic safety handling sequence of the filling station.

[0147] In summary, the system systematically integrates historical accident data from filling stations, inherent process characteristics of equipment, and external safety regulations to construct a structured risk scenario template library. This transforms scattered, experiential safety knowledge into standardized, computable data models, providing a complete reference benchmark for subsequent automated risk matching and laying a solid foundation for rapidly matching real-time risk characteristics to corresponding contingency plans.

[0148] In summary, the system accurately matches and filters real-time calculated risk diffusion characteristics against pre-defined risk scenario templates. By comparing the cumulative risk intensity with the intensity range in the template, and the potential risk arrival location with the pre-defined location in the template, it achieves an accurate mapping from dynamic, quantitative real-time risk data to static, structured known risk categories. This step effectively identifies which known accident scenario best matches the current risk situation, significantly shortening the decision-making time from risk perception to contingency plan activation.

[0149] In summary, based on the identified target risk scenario, all relevant basic operating instructions are automatically retrieved from a pre-set safety operation instruction library. This enables rapid, batch recall of emergency response plans, avoiding delays and omissions caused by relying on manual memorization or consulting paper procedures in emergency situations. Its effectiveness lies in ensuring that every retrieved operating instruction is a pre-verified, directly relevant, safe, and effective mature action.

[0150] In summary, based on the rigorous process logic dependencies and precise equipment response times of the filling station, the retrieved basic safety operation instructions were sorted and time-sequentially verified. This simulates the core logic of experienced operators arranging operational steps in emergency situations, ensuring that the instruction sequence is technically executable, logically conflict-free, and achieves the most efficient arrangement in terms of time. The resulting dynamic safety response sequence is an executable plan that conforms to safety regulations, fits the current real-time process status, and has been time-optimized. Its effect is to directly transform the risk analysis conclusions into a set of precise operational guidelines that can be automatically or manually executed, with clear steps and precise timing.

[0151] In this embodiment of the invention, when fusing the disturbance process data and environmental parameters generated by executing the dynamic safety handling sequence to obtain the status feedback data of the filling station, the specific method is as follows:

[0152] Synchronously collect equipment status change data affected by the dynamic safety response sequence;

[0153] Extract environmental monitoring parameters from key monitoring points in the filling station;

[0154] The equipment status change data and the environmental monitoring parameters are spatiotemporally aligned to obtain the fusion monitoring sequence of the filling station;

[0155] The deviation of the filling station is obtained by performing differential calculation on the fusion monitoring sequence based on the dynamic safety handling sequence.

[0156] The deviation, the state change data, and the environmental monitoring parameters are collectively encapsulated into the state feedback data of the filling station.

[0157] Specifically, while executing the dynamic safety handling sequence, the system initiates a parallel data acquisition process. This process continuously reads the real-time status parameters of the devices that are explicitly listed as action targets in each instruction of the sequence from the control units of these devices or the sensors directly connected to them.

[0158] Specifically, the system accesses a predefined list of key monitoring points, which is based on the safety risk assessment results of the filling station and clarifies the physical locations within the station that must be continuously monitored, such as around the hydrogen refueling machine, inside the hydrogen storage cylinder group dike, on the top of the compressor plant, and near the outlet of the vent pipe.

[0159] Specifically, a unified sequence of time points is first established, starting from the moment the dynamic safety handling sequence begins execution and increasing at fixed time intervals. For equipment status change data, at each unified time point, the system searches for the equipment data record with the timestamp closest to that point and directly uses the value of that record.

[0160] Specifically, based on the expected effects of the instructions in the dynamic safety response sequence, the system establishes a baseline curve for the expected parameter changes of each controlled device. This curve defines the values ​​that the key parameters of the device should reach over time from the moment the instruction is issued, under ideal conditions.

[0161] Specifically, a data packet with a fixed format is created, which contains three main data segments: the first data segment stores a list of raw, timestamped device status change data; the second data segment stores a list of raw, timestamped environmental monitoring parameters; and the third data segment stores a list of calculated deviation data aligned with a uniform time point sequence.

[0162] Furthermore, the specific parameter types collected are equipment-specific. For example, for valves, the opening feedback signal and valve position status are collected; for compressors, the operating current, speed, and outlet pressure are collected; and for hydrogen storage tanks, the pressure and temperature are collected. This data collection process is performed at a fixed frequency, higher than the command execution interval, to ensure that the complete change process of the equipment from the start of command execution to the stabilization of its state can be captured. All collected data are stamped with millisecond-level timestamps that are strictly synchronized with the central monitoring system. This set of real-time equipment readings with precise timestamps represents the equipment state change data affected by the dynamic safety response sequence.

[0163] Furthermore, based on the unique number of each monitoring point in the list, the system reads real-time data from the independent environmental monitoring sensor network deployed at the corresponding location. This data includes, but is not limited to, hydrogen volume concentration percentage, ambient temperature, wind speed, and atmospheric pressure. The environmental monitoring sensor network samples at its own stable frequency, and the system appends a timestamp generated by the same clock source synchronized with the equipment's data acquisition system when reading the data, ensuring that all monitoring data have a comparable time reference, thereby extracting the environmental monitoring parameters of key monitoring points in the filling station.

[0164] Furthermore, for environmental monitoring parameters, since their sampling frequencies may differ, the system finds the two closest environmental monitoring data points before and after each unified time point. Using a linear calculation method, based on the values ​​of these two data points and the time difference, it extrapolates the estimated environmental parameter values ​​for that unified time point. In this way, both equipment data and environmental data are converted to the same strictly aligned set of time points. The system then merges all equipment status parameter values ​​and all environmental monitoring parameter values ​​corresponding to each time point to form a complete record. Arranging all the records from all time points in chronological order yields the fused monitoring sequence of the filling station.

[0165] Furthermore, the specific method for calculating the deviation is as follows: the system iterates through each time point record in the fusion monitoring sequence. For each actual monitoring parameter value from the controlled equipment in the record, the system finds the expected value at the same time point on the corresponding expected baseline curve. Then, the system subtracts the expected value from the actual monitoring value to calculate a difference, which is the deviation of the parameter from the expected state at that moment. The system performs this subtraction operation on all relevant parameters at all time points in the sequence, thereby obtaining a new data sequence that corresponds one-to-one with the time points of the fusion monitoring sequence and contains the deviation values ​​of each parameter. This sequence, which completely records the difference between the actual response and the expected command effect, is the calculated deviation of the filling station.

[0166] Furthermore, the data packet also contains metadata, such as the unique number of the corresponding dynamic safety handling sequence and the start and end times of the data coverage. The system stores and transmits this composite data packet as a whole. This data packet, which fully encapsulates the multi-dimensional state response information of the system after the execution of the disturbance, is defined as the status feedback data of the filling station.

[0167] In summary, the status parameters of the relevant controlled equipment are collected simultaneously while the safety response sequence is being executed. This enables the immediate and accurate capture of the effectiveness of the response measures, forming a complete and precisely time-stamped equipment response record. Its benefit lies in providing firsthand, objective equipment data for assessing whether safety control actions have achieved their intended objectives, avoiding delays and distortions in post-event analysis, and constituting an indispensable real-world feedback signal source in closed-loop control.

[0168] In summary, parallel extraction of environmental monitoring data from key locations within the filling station incorporates external environmental conditions into the safety assessment system. This enables simultaneous monitoring of potential secondary environmental risks arising from the treatment measures or interference from the external environment with the treatment effect. Its effect lies in expanding the perspective of safety assessment from simple equipment control to a broader scope of environmental safety, ensuring that the safety assessment is comprehensive and multi-dimensional, meeting the stringent requirements of hydrogen facilities for the safety of the surrounding environment.

[0169] In summary, by synchronizing and spatially aligning device status data and environmental parameter data from different sources and sampling rates, a unified fused monitoring sequence is generated. This resolves the inconsistency in timestamps and spatial references between multi-source heterogeneous monitoring data, creating a comparable data set with all monitoring points at the same time reference and clearly defined spatial locations. Its benefit lies in providing a standardized data platform for subsequent accurate correlation analysis and effect evaluation, serving as a prerequisite for precise differential calculations and situational understanding.

[0170] In summary, the deviation sequence is obtained by point-by-point difference calculation between the actual observed values ​​in the integrated monitoring sequence and the ideal state values ​​expected based on the dynamic safety response sequence. This enables a quantitative measurement of the deviation of the response effect down to each time point and each parameter. Its effect is to transform vague qualitative judgments into clear quantitative indicators, providing a direct, data-driven decision-making basis for judging whether the response is sufficient and whether supplementary measures are needed.

[0171] In summary, deviations, raw equipment status change data, and raw environmental monitoring parameters are collectively encapsulated into a structured status feedback data package. This achieves complete archiving and integration of all key feedback information within a single response cycle. Its effect lies in creating a comprehensive and self-interpretive data object that can not only be used for immediate assessment of the current response but also serve as a historical case study, providing high-quality training and validation data for subsequent risk transmission model iterations and response strategy optimization. It is the core data carrier for enabling the entire safety monitoring system to learn and continuously improve.

[0172] In this embodiment of the invention, when iteratively updating the abnormal transmission path based on the state feedback data to obtain the risk transmission status of the filling station, it is specifically used for:

[0173] Based on the statistical characteristics in the status feedback data, the transmission probability weights of the corresponding nodes in the abnormal transmission path are corrected to obtain the target weight of the filling station;

[0174] High-risk transmission edges are selected in the site topology network diagram based on the target weight;

[0175] The visualized network diagram containing the abnormal transmission path and the selected high-risk transmission edge is used as the risk transmission status of the filling station.

[0176] Specifically, for each specific process node along the anomaly propagation path, all deviation data points related to the equipment at that node are selected. The average value of these data points is calculated to assess the average deviation level, and the standard deviation of these data points is calculated to assess the degree of fluctuation in deviation. These two calculated values ​​constitute the statistical characteristics used for correction.

[0177] Specifically, a unified high-risk threshold is set for the entire network. This threshold is a fixed value predetermined based on historical security data. The system traverses all edges in the network graph. For each edge, it checks whether it exists in the target weight set: if it exists, the target weight value is used directly; if it does not exist, that is, the edge is not on the abnormal propagation path of this update, its initial propagation probability weight is used as the current weight value.

[0178] Specifically, the system invokes a graphics visualization engine to load the basic layout data of vertices and edges of the site's topology network diagram. The specific method for generating a visualized network diagram as a risk transmission situation is that the engine first draws all vertices and edges in a standard style. Then, the engine highlights and renders the vertices and edges contained in the anomaly transmission path with bright colors and bold lines.

[0179] Furthermore, the specific method for correcting the transmission probability weight to obtain the target weight is as follows: the system reads the initial transmission probability weight of the edge associated with the node in the original abnormal transmission path and uses this weight as a base value; then, the calculated average deviation level is multiplied by a preset amplification factor to obtain an adjustment amount. If the average deviation level is positive, the adjustment amount is added to the base weight; if it is negative, the adjustment amount is subtracted from the base weight; then, the calculated standard deviation is multiplied by another preset sensitivity coefficient and added to the weight adjusted by the average deviation to reflect the uncertainty caused by volatility; after these two steps of linear superposition adjustment based on statistical characteristics, a new weight value is obtained. This weight updated for the outgoing edge of the node is recorded as the target weight of the filling station.

[0180] Furthermore, the system compares the current weight value of each edge with the set high-risk threshold, and marks the edges whose current weight is greater than or equal to the high-risk threshold. The set of all marked edges is then selected as high-risk transmission edges in the site topology network graph. These edges represent physical connections with a high probability of risk transmission under the current state feedback.

[0181] Furthermore, the engine then iterates through the selected set of high-risk transmission edges, drawing these edges with a different, more prominent color and dashed line style. If an edge belongs to both the abnormal transmission path and the high-risk transmission edge set, the abnormal transmission path highlighting style is prioritized. Finally, the engine adds a legend to the side of the graph, explaining the meaning of different colors and line styles, and labels the currently calculated system time as part of the title on the graph. This comprehensive graph, integrating static network topology, dynamic abnormal paths, and real-time high-risk edge screening, is rendered and output as a complete image, which is defined as the risk transmission status of the filling station.

[0182] Compared with the prior art, the present invention has the following beneficial effects:

[0183] 1. This invention achieves dynamic, quantitative, and spatial assessment of safety risks at hydrogen filling stations by constructing an intelligent analysis chain covering multi-source monitoring data, an internal station topology network, and a safety knowledge base. Its beneficial effects lie in the system's ability to identify abnormal states in real time, automatically deduce their propagation paths within the process network, and accurately calculate the cumulative intensity and expected arrival location of risks along the path. This elevates discrete alarm signals to a deep understanding and predictive judgment of risk evolution, significantly improving the accuracy of safety monitoring and the foresight of decision-making.

[0184] 2. This invention further matches real-time risk characteristics with pre-set risk scenario templates to automatically generate dynamic safety response sequences that highly match the current situation, and iteratively updates the risk transmission model based on multi-dimensional feedback data after execution. This process transforms safety response from passive alarms and fixed plan execution to proactive closed-loop protection based on real-time situational awareness, intelligent decision-making, and continuous self-learning, thereby effectively improving the accuracy, adaptability, and overall reliability of safety interlocking control under complex operating conditions.

[0185] like Figure 2 The diagram shown is a functional block diagram of a hydrogen filling station safety interlock monitoring system provided in an embodiment of the present invention.

[0186] The hydrogen filling station safety interlock monitoring system 100 of this invention can be installed in an electronic device. Depending on the functions implemented, the hydrogen filling station safety interlock monitoring system 100 may include an abnormal state module 101, an abnormal transmission path module 102, a risk characteristic module 103, a safety handling sequence module 104, a status feedback module 105, and a risk transmission situation module 106. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0187] In this embodiment, the functions of each module / unit are as follows:

[0188] The abnormal status module extracts features from the multi-source heterogeneous safety monitoring data of the filling station to obtain the abnormal status identifier set of the filling station;

[0189] The anomaly propagation path module analyzes the topological and temporal correlations of the anomaly status identifier set in spatial location, combined with the process layout of the filling station, to obtain the anomaly propagation path of the filling station;

[0190] The risk feature module performs path integration on the rate and direction of change of the parameters corresponding to the anomaly identifiers in the anomaly propagation path to obtain the risk diffusion characteristics of the filling station;

[0191] The safety handling sequence module matches corresponding safety operation instructions to the risk scenarios constructed based on the risk diffusion characteristics, and cascades and arranges the safety operation instructions to obtain the dynamic safety handling sequence of the filling station;

[0192] The status feedback module fuses the disturbance process data and environmental parameters generated by executing the dynamic safety handling sequence to obtain the status feedback data of the filling station.

[0193] The risk transmission status module iteratively updates the abnormal transmission path based on the status feedback data to obtain the risk transmission status of the filling station.

[0194] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0195] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0196] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0197] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0198] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0199] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A safety interlock monitoring method for hydrogen filling stations, characterized in that... The method includes: Feature extraction is performed on the multi-source heterogeneous safety monitoring data of the filling station to obtain the abnormal state identifier set of the filling station; By analyzing the topological and temporal correlations of the abnormal status identifier set in spatial location based on the process layout of the filling station, the abnormal propagation path of the filling station can be obtained; By performing path integration on the rate of change and direction of change of the parameters corresponding to the anomaly identifier in the anomaly propagation path, the risk diffusion characteristics of the filling station can be obtained; Based on the risk diffusion characteristics, the corresponding safety operation instructions are matched to the risk scenarios constructed, and the safety operation instructions are cascaded and arranged to obtain the dynamic safety handling sequence of the filling station; The disturbance process data and environmental parameters generated during the execution of the dynamic safety handling sequence are fused and processed to obtain the status feedback data of the filling station; Based on the status feedback data, the abnormal transmission path is iteratively updated to obtain the risk transmission status of the filling station.

2. The safety interlock monitoring method for hydrogen filling stations as described in claim 1, characterized in that... The step of extracting features from the multi-source heterogeneous safety monitoring data of the filling station to obtain the abnormal state identifier set of the filling station includes: By acquiring equipment status parameters and environmental monitoring parameters, multi-source heterogeneous safety monitoring data of the filling station can be obtained; The multi-source heterogeneous safety monitoring data is subjected to sliding window normalization to obtain the time-series standardized parameters of the filling station; Abnormal parameter fragments of the filling station are obtained by performing abrupt change detection on the time-series standardized parameters; Assign an abnormal identifier to the abnormal parameter fragment, and associate and store the abnormal identifier, the corresponding parameter type, the amount exceeding the threshold, and the timestamp to obtain the abnormal status identifier set of the filling station.

3. The safety interlock monitoring method for hydrogen filling stations as described in claim 1, characterized in that... The analysis of the abnormal status identifier set in terms of spatial topological and temporal correlation, combined with the process layout of the filling station, yields the abnormal propagation path of the filling station, including: The abnormal status identifier set is mapped to the process piping and equipment layout diagram of the filling station to obtain the corresponding physical location coordinates; Based on the physical connection relationships between the process pipelines and the equipment in the equipment layout diagram, construct an intra-station topology network diagram with the equipment in the equipment layout diagram and the process pipelines as vertices and the physical connection relationships as edges; Based on the physical location coordinates, the transmission calculation of the process nodes in the station's internal topology network diagram is performed to obtain the abnormal transmission path of the filling station.

4. The safety interlock monitoring method for hydrogen filling stations as described in claim 3, characterized in that... The step of performing transmission calculations on process nodes in the station's internal topology network diagram based on the physical location coordinates to obtain the abnormal transmission path of the filling station includes: In the station topology network diagram, starting from the physical location coordinates, the probability intensity of the propagation of abnormal states from the abnormal state identifier set to adjacent nodes is calculated; Based on the occurrence order and time interval represented by the timestamps of the anomaly identifiers in the anomaly status identifier set, a corresponding time-series correlation index is assigned to each anomaly identifier pair; Based on the probability intensity and the temporal correlation index, high-risk transmission sequences in the station's internal topology network diagram are identified, and abnormal transmission paths of the filling station are obtained.

5. The safety interlock monitoring method for hydrogen filling stations as described in claim 3, characterized in that... The step of performing path integration on the rate and direction of change of the parameters corresponding to the anomaly identifier in the anomaly propagation path to obtain the risk diffusion characteristics of the filling station includes: Extract the risk sequence of the parameters corresponding to the anomaly identifier within the time window; The first derivative of the risk sequence is used as the rate of change vector of the filling station; Determine the spatial connection direction vector between adjacent anomaly identifiers on the anomaly propagation path based on the station topology network diagram; The intensity integral value of the filling station is obtained by integrating the rate of change vector based on the spatial connection direction vector. By binding the intensity integral value with the endpoint location information of the abnormal transmission path, the risk diffusion characteristics of the filling station are obtained.

6. The safety interlock monitoring method for hydrogen filling stations as described in claim 5, characterized in that... The formula for calculating the intensity integral value includes: in, The intensity integral value, For path segment index, This represents the total number of path segments. For the first Equipment-based risk factors for each path segment For the first The abnormal state evolution factor of each path segment, For the first The rate of change vector of each path segment For the first Spatial connection direction vectors of path segments, The intensity coefficient between adjacent path segments. This is the saturation effect adjustment coefficient. It is the hyperbolic tangent function. For the first The vector of the rate of change of parameters at each path segment For the first Spatial connection direction vectors of parameters at each path segment.

7. The safety interlock monitoring method for hydrogen filling stations as described in claim 1, characterized in that... The step of matching corresponding safety operation instructions to the risk scenario constructed based on the risk diffusion characteristics, and cascading and arranging the safety operation instructions to obtain the dynamic safety handling sequence of the filling station, includes: Based on the accident data, equipment process characteristics, and safety requirements of the filling station, a risk scenario template for the filling station is constructed; The risk accumulation intensity and potential risk arrival location in the risk diffusion characteristics are matched with the risk scenario template to filter out the target risk scenarios of the filling station; Retrieve basic safety operation instructions associated with the target risk scenario from the safety operation instruction library of the filling station; Based on the process logic dependencies and equipment action response time of the filling station, the basic safety operation instructions are logically verified to obtain the dynamic safety handling sequence of the filling station.

8. The safety interlock monitoring method for hydrogen filling stations as described in claim 1, characterized in that... The process of fusing the disturbance process data and environmental parameters generated by executing the dynamic safety handling sequence to obtain the status feedback data of the filling station includes: Synchronously collect equipment status change data affected by the dynamic safety response sequence; Extract environmental monitoring parameters from key monitoring points in the filling station; The equipment status change data and the environmental monitoring parameters are spatiotemporally aligned to obtain the fusion monitoring sequence of the filling station; The deviation of the filling station is obtained by performing differential calculation on the fusion monitoring sequence based on the dynamic safety handling sequence. The deviation, the state change data, and the environmental monitoring parameters are collectively encapsulated into the state feedback data of the filling station.

9. The safety interlock monitoring method for hydrogen filling stations as described in claim 3, characterized in that... The step of iteratively updating the anomaly transmission path based on the status feedback data to obtain the risk transmission status of the filling station includes: Based on the statistical characteristics in the status feedback data, the transmission probability weights of the corresponding nodes in the abnormal transmission path are corrected to obtain the target weight of the filling station; High-risk transmission edges are selected in the site topology network diagram based on the target weight; The visualized network diagram containing the abnormal transmission path and the selected high-risk transmission edge is used as the risk transmission status of the filling station.

10. A safety interlock monitoring system for a hydrogen filling station, used to implement the safety interlock monitoring method for a hydrogen filling station as described in any one of claims 1-9, characterized in that... The system includes: The abnormal status module extracts features from the multi-source heterogeneous safety monitoring data of the filling station to obtain the abnormal status identifier set of the filling station; The anomaly propagation path module analyzes the topological and temporal correlations of the anomaly status identifier set in spatial location, combined with the process layout of the filling station, to obtain the anomaly propagation path of the filling station; The risk feature module performs path integration on the rate and direction of change of the parameters corresponding to the anomaly identifiers in the anomaly propagation path to obtain the risk diffusion characteristics of the filling station; The safety handling sequence module matches corresponding safety operation instructions to the risk scenarios constructed based on the risk diffusion characteristics, and cascades and arranges the safety operation instructions to obtain the dynamic safety handling sequence of the filling station; The status feedback module fuses the disturbance process data and environmental parameters generated by executing the dynamic safety handling sequence to obtain the status feedback data of the filling station. The risk transmission status module iteratively updates the abnormal transmission path based on the status feedback data to obtain the risk transmission status of the filling station.

Images