Intelligent safety management and control method and system for gas station based on digital twinning

Abstract

The invention provides an intelligent safety management and control method and system for a gas station based on digital twinning, and relates to the field of digital twinning and artificial intelligence, and the method comprises the steps: collecting multi-source heterogeneous data in the gas station in real time, and constructing a total-factor digital twinning body of the gas station; extracting multi-dimensional feature data from the digital twin, inputting the multi-dimensional feature data into a pre-trained long-short-term memory network model, and predicting the future state of the equipment to obtain a prediction result; simulating a development process of at least one preset processing scheme in a virtual environment corresponding to the digital twin based on the prediction result and a preset initial accident scene to obtain a drilling result corresponding to each preset processing scheme; and determining an optimal disposal scheme corresponding to the initial accident scene based on the drill score. By implementing the method, the safety management and control level and the emergency preparation capability of the gas field station can be improved.

Description

Technical Field

[0001] This application relates to the fields of digital twins and artificial intelligence, and in particular to a smart safety management and control method and system for gas stations based on digital twins. Background Technology

[0002] Gas stations, as core infrastructure in urban energy supply networks, integrate complex processes such as natural gas reception, purification, storage, pressure regulation, and distribution. They have dense internal pipeline networks, diverse equipment, and handle natural gas, which inherently possesses high-risk characteristics of flammability and explosiveness. Therefore, the safe and stable operation of gas stations is directly related to urban energy security, public safety, and the safety of people's lives and property. How to implement effective and reliable safety management in this high-risk industrial setting to prevent accidents has always been a core issue that the gas industry has been continuously exploring and optimizing.

[0003] In related technologies, a commonly used monitoring and alarm system is based on distributed sensors and a central control room. Various types of sensors, such as those for pressure, temperature, flow rate, and gas concentration, are deployed in key equipment and areas of the site. These sensors independently collect single physical parameters at their locations. All data collected by the sensors is transmitted in real-time to a monitoring computer in the central control room. Maintenance personnel observe the real-time data at each measuring point through an interface presented in the form of a two-dimensional process diagram or data list on the monitoring computer. Simultaneously, the system pre-sets fixed, static safety thresholds for key parameters. When the data at a measuring point exceeds its preset threshold range, the system triggers an audible and visual alarm to alert maintenance personnel to an over-limit event, allowing for manual intervention according to standard emergency plans.

[0004] However, the relevant technical solutions assess risk by comparing real-time data from individual measuring points with static thresholds, making their control logic essentially passive and reactive. In other words, the system can only issue an alarm after a parameter has clearly exceeded safety boundaries and an abnormal state has become a definite fact. It struggles to identify systemic deterioration trends or early signs of failure exhibited by the coordinated changes of multiple parameters within their respective safe ranges. For example, the pressure, temperature, and vibration data of a device may not all exceed limits, but their combined dynamic changes might already indicate a leak within the next few hours. This could lead to missing the optimal window for preventative intervention before an accident occurs, posing a significant safety hazard. Summary of the Invention

[0005] This application provides a smart safety management and control method and system for gas stations based on digital twins, which addresses the problem that passive response control methods used in gas stations in related technologies are difficult to identify and intervene in before a fault actually occurs.

[0006] In a first aspect, this application provides a method for intelligent safety management and control of gas stations based on digital twins, applied to an intelligent safety management and control system for gas stations based on digital twins, the method comprising: Real-time collection of multi-source heterogeneous data within the gas station, including equipment data, station environmental data, and process flow data; A full-element digital twin of the gas station is constructed based on the multi-source heterogeneous data. The full-element digital twin integrates the geometric model, physical model and behavioral model of the equipment. Multidimensional feature data is extracted from the digital twin and input into a pre-trained long short-term memory network model to predict the future state of the equipment, and the prediction result is obtained. The multidimensional feature data includes historical operating data, real-time operating data, site environment data and process flow data of the equipment. Based on the prediction results and the preset initial accident scenario, the development process of at least one preset processing scheme is simulated in the virtual environment corresponding to the digital twin, and the exercise results corresponding to each preset processing scheme are obtained. The optimal response plan for the initial accident scenario is determined based on the exercise score corresponding to each exercise result.

[0007] By adopting the above technical solution, the system can aggregate multi-source data from gas stations in real time to construct a fully-fledged digital twin that accurately maps to the physical entity. Based on this, the system utilizes a Long Short-Term Memory (LSTM) network model to perform in-depth analysis of the multi-dimensional features extracted from this twin, thereby achieving forward-looking predictions of the equipment's future state. This allows the aforementioned regulatory methods to respond not only after an anomaly occurs, but also to simulate and evaluate various pre-set emergency response plans in a virtual environment based on the prediction results. This enables the system to shift from passive response to proactive prevention, identifying risks and determining the optimal handling strategy before an accident occurs, fundamentally improving the safety management level and emergency preparedness capabilities of gas stations.

[0008] In some embodiments, the step of constructing a full-element digital twin of the gas station based on the multi-source heterogeneous data specifically includes: An initial digital twin of the device is constructed based on the multi-source heterogeneous data, the initial digital twin including a geometric model, a physical model and a behavioral model; Obtain the performance parameter deviation between the simulation output of the initial digital twin and the actual operating data of the physical entity; The performance degradation parameters of the device are calculated based on the Bayesian inference algorithm and the performance parameter deviation inversion. Based on the performance degradation parameters, the physical model and the behavioral model are modified to obtain a full-element digital twin.

[0009] By employing the aforementioned technical solution, the system constructs an initial digital twin. Then, by continuously comparing the performance deviations between the twin's simulation output and the actual operating data of the physical entity, a Bayesian inference algorithm is used to inversely calculate the actual performance degradation parameters of the device. Finally, these inferred degradation parameters are used to correct the physical and behavioral models of the digital twin. This iterative process of "virtual-real comparison and reverse correction" ensures that the digital twin can dynamically and accurately reflect the performance degradation of the physical device caused by wear, aging, and other factors, maintaining high fidelity at all times.

[0010] In some embodiments, the step of calculating the performance degradation parameters of the device based on the Bayesian inference algorithm and the performance parameter deviation inversion specifically includes: A prior probability distribution model of performance degradation parameters is established based on the performance degradation patterns in the historical operating data and maintenance records of the equipment. Based on the performance parameter deviation and the simulation model of the initial digital twin, a likelihood function is constructed between the observed values ​​and the predicted values. The likelihood function is used to characterize the probability of generating the performance parameter deviation under a given performance degradation parameter. The prior probability distribution model is multiplied by the likelihood function according to Bayes' theorem to obtain the posterior probability distribution of the performance degradation parameters. The performance degradation parameters of the device are determined based on the posterior probability distribution.

[0011] By adopting the above technical solution, the system uses a Bayesian inference method when inverting and calculating equipment performance degradation parameters. By combining a prior probability model based on historical data and maintenance records with a likelihood function, the system can derive a posterior probability distribution that better reflects reality, thereby determining the most probable performance degradation parameters. This avoids the limitations of relying solely on experience or a single data source, improving the accuracy and reliability of degradation parameter estimation.

[0012] In some embodiments, the step of simulating the development process of at least one preset processing scheme in the virtual environment corresponding to the digital twin based on the prediction results and preset initial accident scenarios specifically includes: The fault type and fault status of the faulty equipment are determined based on the prediction results; An accident simulation scenario is constructed in the virtual environment based on the fault type, the fault state, and the scenario parameters of the initial accident scenario. Obtain at least one preset processing scheme, each preset processing scheme including a sequence of processing operations and an operation timing; The accident development process is simulated by sequentially executing the handling operation sequence of each preset handling scheme in the accident simulation scenario, and the exercise results corresponding to each preset handling scheme are obtained.

[0013] By adopting the above technical solution, the system deeply integrates equipment status prediction results with emergency response plan simulation exercises. When the system predicts a specific fault, it does not simply issue an alarm, but automatically constructs an accident simulation scenario in a digital twin virtual environment that is highly consistent with the predicted fault type, status, and on-site environment. Then, the system exercises a preset processing plan containing specific operation sequences and timing within this virtual scenario, allowing the abstract emergency response plan to be tested in specific, dynamically evolving accident scenarios, intuitively demonstrating the accident development process under different handling measures.

[0014] In some embodiments, the step of determining the optimal handling plan corresponding to the initial accident scenario based on the exercise score corresponding to each exercise result specifically includes: Quantitative evaluation indicators are extracted from the results of each exercise, including accident control time, total leakage amount, and scope of impact. Calculate the exercise score corresponding to each preset processing scheme based on the quantitative evaluation indicators; The preset handling scheme with the highest exercise score is selected as the optimal handling scheme corresponding to the initial accident scenario.

[0015] By adopting the above technical solution, the system automatically extracts several core quantitative evaluation indicators from the virtual exercise results after the exercise, such as accident control time, total leakage, and scope of impact. These indicators directly reflect the core objectives of emergency response—"rapid control, minimal leakage, and minimal impact." By weighting these indicators, a comprehensive exercise score is obtained for each plan. Finally, the plan with the highest score is selected as the optimal response plan.

[0016] In some embodiments, after the step of selecting the preset handling scheme with the highest exercise score as the optimal handling scheme corresponding to the initial accident scenario, the method further includes: Obtain historical execution data of the operators who are scheduled to execute the optimal handling plan, including historical average execution time and operation success rate; Based on the historical average execution time and the operation success rate, the reserved time margin of each disposal operation in the optimal disposal plan is adjusted to obtain a personalized disposal plan.

[0017] By adopting the above technical solution, after determining the optimal handling plan, the system retrieves the historical execution data of the designated operators and analyzes their individual ability indicators, such as average time consumption and success rate in similar operations. Based on this objective data, the system dynamically adjusts the time margin in the theoretically optimal plan, reserving more time for personnel with weaker execution capabilities or optimizing processes for highly efficient personnel, thereby improving the effectiveness of emergency plans in real-world scenarios.

[0018] In some embodiments, after the step of obtaining the performance parameter deviation between the simulation output of the initial digital twin and the actual operating data of the physical entity, the method further includes: At least one acoustic exciter is provided on the device for actively transmitting a preset acoustic excitation signal; The real-time acoustic resonance fingerprint data formed during the propagation of the acoustic excitation signal is collected by multiple acoustic sensors deployed in the device. The real-time acoustic resonance fingerprint data is used to characterize the current physical structural integrity of the device. The difference between the real-time acoustic resonance fingerprint data and the reference acoustic resonance fingerprint is used as part of the performance parameter deviation, which is determined in advance in the digital twin under ideal health conditions.

[0019] By employing the aforementioned technical solution, the system actively emits acoustic excitation signals and collects their response data after propagation within the equipment structure to obtain an "acoustic resonance fingerprint" that characterizes the current physical structural integrity of the equipment. Unlike monitoring only indirect operating parameters such as pressure and temperature, this method can more directly and sensitively capture structural changes caused by micro-cracks, material fatigue, etc. Comparing this fingerprint data with a baseline in a healthy state, the difference is used as part of the performance parameter deviation, providing a higher-dimensional and precise input regarding the physical health of the equipment for the correction of the digital twin model, enhancing the system's ability to detect early structural defects.

[0020] Secondly, this application provides a smart safety management and control system for gas stations based on digital twins, the system comprising: one or more processors and a memory; The memory is coupled to the one or more processors. The memory is used to store computer program code, which includes computer instructions. The one or more processors call the computer instructions so that the system can implement the intelligent safety management and control method for gas stations based on digital twins provided in the above embodiments, which will not be described in detail here.

[0021] Thirdly, this application provides a computer-readable storage medium including instructions that, when executed on a digital twin-based intelligent safety management and control system for gas stations, enable the system to implement the digital twin-based intelligent safety management and control method for gas stations provided in the above embodiments, which will not be elaborated here.

[0022] Fourthly, this application provides a computer program product that, when running on a digital twin-based intelligent safety management and control system for gas stations, enables the system to implement the digital twin-based intelligent safety management and control method for gas stations provided in the above embodiments, which will not be elaborated here.

[0023] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. By constructing a digital twin and using AI models to predict future equipment failure states, the system can "see the future" before an accident occurs. It transforms this predictive capability into action guidance, simulating and quantifying various emergency plans in a virtual environment. This allows the optimal response strategy to be determined in advance when risks are still in their nascent stages, achieving truly proactive and preventative safety management.

[0024] 2. The system continuously compares the performance deviations between the twin and the physical entity, and introduces active acoustic resonance fingerprint technology to directly perceive subtle changes in the device's physical structure. It utilizes a Bayesian inference algorithm, integrating historical patterns and real-time evidence, to reverse-calculate the device's true performance degradation parameters, and dynamically corrects the twin model accordingly, ensuring high fidelity throughout the twin model's entire lifecycle.

[0025] 3. By analyzing historical execution data (such as average time consumption and success rate) of specific operators, the system makes customized adjustments to the operation sequence and time margin in the optimal solution, resulting in personalized instructions optimized for specific personnel that are more operable and have a higher success rate. This greatly bridges the gap between theoretical optimization and actual execution, improving the success rate of emergency response in real-world scenarios. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating a smart safety management and control method for gas stations based on digital twins, as described in this application. Figure 2 This is another flowchart illustrating a smart safety management and control method for gas stations based on digital twins, as described in this application. Figure 3 This is a schematic diagram of the physical device structure of a smart safety management and control system for gas stations based on digital twins, as described in this application. Detailed Implementation

[0027] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0028] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0029] For ease of understanding, the method provided in this implementation is described in process below. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating a smart safety management and control method for gas stations based on digital twins, as described in this application.

[0030] S101. Real-time acquisition of multi-source heterogeneous data within gas stations.

[0031] Among them, multi-source heterogeneous data refers to datasets that come from different collection sources, have different data formats and types, but are related to each other; equipment data refers to relevant data representing the working status of various equipment in the gas station during operation; station environment data refers to data related to the natural and artificial environment of the space where the gas station is located; process flow data refers to relevant data generated in the process links of gas receiving, purification, storage, pressure regulation and transmission and distribution.

[0032] Specifically, the system collects various data related to safety management within the gas station, ensuring that the data covers key dimensions such as equipment operation, environmental conditions, and process flow. Equipment data includes core parameters such as compressor vibration, temperature, and speed; storage tank pressure, liquid level, and material fatigue; and valve opening / closing status and sealing performance. Station environmental data includes environmental factors affecting station safety such as temperature, humidity, wind speed, combustible gas concentration, and rainfall. Process flow data includes information related to process execution, such as natural gas flow rate, pressure regulation curves, and equipment start-up and shutdown operation records.

[0033] In some embodiments, real-time data acquisition can be achieved through the linkage of sensors and the system. Optionally, the system can deploy corresponding dedicated sensors on key equipment (compressors, storage tanks, valves, etc.), deploy environmental sensors such as temperature and humidity, wind speed, and combustible gas concentration in the site environment, and install flow and pressure sensors at key nodes in the process flow; the sensors collect corresponding data in real time at a preset frequency (e.g., 10Hz), and automatically filter out obviously abnormal invalid data during the acquisition process; the sensors transmit the collected data to the system's data receiving interface through wired or wireless communication modules, and the system performs preliminary classification and storage of the received data.

[0034] S102. Construct a full-element digital twin of a gas station based on multi-source heterogeneous data.

[0035] Among them, the full-element digital twin refers to a virtual model that integrates the geometric model, physical model and behavioral model of the equipment and can accurately map the physical state of the gas station; the geometric model refers to a three-dimensional spatial model built based on the actual spatial layout of the station and the external structure of the equipment; the physical model refers to a model that defines the physical attributes of the equipment and the physical laws of failure; and the behavioral model refers to a model that simulates the operating behavior of the equipment and the linkage logic of the process.

[0036] Specifically, the system uses multi-source heterogeneous data collected in step S101 to model from three core dimensions: geometry, physics, and behavior, ultimately forming a full-element digital twin corresponding to the physical gas station at a 1:1 scale. In the geometric modeling stage, the system combines CAD drawings of the station with on-site laser scanning data to accurately recreate the three-dimensional spatial structure of equipment, pipelines, and buildings, ensuring complete consistency in spatial position and dimensional parameters between the virtual model and the physical entity. In the physical modeling stage, physical properties such as the equipment's material, rated pressure, and maximum temperature are entered, defining fault physical laws such as changes in sealing performance when pressure exceeds the rated value. In the behavioral modeling stage, the system simulates the equipment's start-up and shutdown processes, valve opening and closing response times, and process linkage logic such as automatically starting the feed pump when the tank level is low. During the construction process, it is necessary to ensure deep integration of the three types of models, enabling the twin to comprehensively and realistically reflect the operating status of the physical station and support real-time synchronous data updates.

[0037] S103. Extract multi-dimensional feature data from the digital twin and input it into a pre-trained long short-term memory network model to predict the future state of the device and obtain the prediction results.

[0038] Among them, multidimensional feature data refers to the comprehensive set of features extracted from the digital twin, including historical operating data, real-time operating data, site environment data, and process flow data of the equipment; the Long Short-Term Memory (LSTM) network model is a deep learning model that is good at processing time-series data and can capture long-term dependencies, and is used to predict the future state of the equipment; the future state of the equipment refers to the possible operating state of the equipment in the future period of time, including normal operation, fault warning, fault type, etc.; the prediction result refers to the judgment information about the future state of the equipment output by the LSTM model, including remaining life, risk level, failure probability, etc.

[0039] Specifically, the system extracts multi-dimensional feature data reflecting the equipment's operating status from a full-element digital twin. This data covers the equipment's historical operating records over a period of time, current real-time operating parameters, real-time environmental data of the site, and related process flow data, ensuring that the feature dimensions are comprehensive and correlated. The extracted multi-dimensional feature data is input into a pre-trained LSTM model, which has been trained based on nearly 5 years of site fault data (including various fault types such as valve internal leakage, compressor failure, and pressure anomalies), and has achieved a prediction accuracy that meets preset requirements. By analyzing the changing trends and correlations of the multi-dimensional feature data, the model predicts the equipment's operating status over a future period, outputting prediction results such as the equipment's remaining lifespan, risk level, failure probability, and failure type.

[0040] S104. Based on the prediction results and the preset initial accident scenario, simulate the development process of at least one preset handling scheme in the virtual environment corresponding to the digital twin, and obtain the exercise results corresponding to each preset handling scheme. Among them, the preset initial accident scenario refers to the standard scenario that the system has pre-set, covering various types of accidents that may occur in gas stations, including leaks, fires, and equipment shutdowns; the virtual environment refers to the virtual simulation space corresponding to the full-element digital twin, which can simulate the operating environment and accident development patterns of the physical station; the preset handling plan refers to the emergency response plan that has been pre-formulated for various initial accident scenarios, including the sequence of handling operations and the timing of operations; and the exercise result refers to the quantitative data and analysis results on the accident development process and handling effect obtained after simulating the execution of the preset handling plan in the virtual environment.

[0041] Specifically, based on the prediction results obtained in step S103, the system identifies the fault type (e.g., valve internal leakage, compressor high-temperature shutdown) and fault status (e.g., leakage amount, equipment temperature) of the faulty equipment. Then, combined with a preset initial accident scenario, it constructs an accident simulation scenario highly closely matching the actual operating conditions in the virtual environment corresponding to the digital twin. The scenario includes key factors such as the faulty equipment status and real-time environmental parameters (e.g., wind speed, wind direction). Subsequently, the system retrieves at least one preset handling plan for this type of accident scenario. Each plan specifies a concrete sequence of handling operations (e.g., valve closing sequence, equipment start-up and shutdown operations) and operation timing (e.g., execution time and interval time for each operation). The handling operations of each preset handling plan are executed sequentially in the virtual environment, simulating the development process of the accident under different handling methods. Key data such as accident control time, gas leakage amount, and impact range are recorded in real time, ultimately forming the exercise results corresponding to each preset handling plan.

[0042] S105. Determine the optimal handling plan for the initial accident scenario based on the exercise score corresponding to each exercise result.

[0043] Among them, the exercise score refers to the quantitative score calculated based on the quantitative evaluation indicators in the exercise results and through preset scoring rules, which is used to evaluate the effectiveness of the preset handling plan; the optimal handling plan refers to the handling plan with the highest exercise score among all preset handling plans, which can control the accident in the shortest time and with the least loss.

[0044] Specifically, the system extracts core quantitative evaluation indicators reflecting the effectiveness of each drill from the results. These indicators mainly include key dimensions such as accident control time (the time from initiating the response to the accident being effectively controlled), total leakage (the total amount of gas leaked during the accident simulation), and impact range (the spatial range of the accident's impact on the station and surrounding areas). Then, according to preset scoring rules, the quantitative evaluation indicators of each preset handling plan are weighted and calculated to obtain the corresponding drill score. The scoring rules must reflect the core principles of "shortest control time, smallest impact range, and optimal cost." Finally, the drill scores of all preset handling plans are ranked, and the plan with the highest score is selected as the optimal handling plan for the initial accident scenario. A detailed plan report is generated, clearly specifying the handling operation steps, precautions, and other key information.

[0045] Optionally, the system can define the scoring calculation logic by pre-setting the weight allocation of quantitative evaluation indicators (such as 40% for accident control time, 30% for total leakage, and 30% for the scope of impact); extract the specific values ​​of indicators such as accident control time, total leakage, and scope of impact from each exercise result, and calculate the exercise score of each plan according to the preset formula (such as score = (1 - accident control time / baseline time) × weight 1 + (1 - total leakage / baseline leakage) × weight 2 + (1 - scope of impact / baseline scope) × weight 3); sort all the exercise scores of the plans from high to low, select the plan with the highest ranking as the optimal response plan; and generate a detailed report of the optimal response plan, including operation steps, timing requirements, key precautions, etc.

[0046] In the above embodiments, the system can aggregate multi-source data from gas stations in real time to construct a fully-featured digital twin that accurately maps to the physical entity. Based on this, the system utilizes a Long Short-Term Memory (LSTM) network model to perform in-depth analysis of the multi-dimensional features extracted from the twin, thereby achieving forward-looking predictions of the equipment's future state. This allows the aforementioned regulatory method to respond not only after an anomaly occurs, but also to simulate and evaluate various preset emergency plans in a virtual environment based on the prediction results. This enables the system to shift from passive response to proactive prevention, identifying risks and determining the optimal handling strategy before an accident occurs, fundamentally improving the safety management level and emergency preparedness capabilities of gas stations.

[0047] The following provides a more detailed description of the process of the method provided in this implementation. Please refer to [link / reference]. Figure 2 This is another flowchart illustrating a smart safety management and control method for gas stations based on digital twins, as described in this application.

[0048] S201, Constructing an initial digital twin of the device based on multi-source heterogeneous data.

[0049] The initial digital twin refers to a virtual model that has been initially constructed based on multi-source heterogeneous data and has not yet undergone performance deviation correction. It includes the geometric model, physical model and behavioral model of the equipment. Multi-source heterogeneous data refers to datasets from different acquisition sources, with different data formats and types but interrelated, covering equipment data, site environment data and process flow data.

[0050] Specifically, the system builds an initial digital twin based on collected multi-source heterogeneous data, encompassing three core dimensions: geometry, physics, and behavior. In geometric modeling, CAD drawings of the site and equipment shape parameters are used to construct a 3D structural model consistent with the dimensions and spatial location of the physical equipment, ensuring accurate reproduction of the connections and layout distribution between devices. In physical modeling, basic physical attributes of the equipment, such as material, rated pressure, rated temperature, and maximum load, are entered to preliminarily define the physical operating patterns of the equipment under normal working conditions. In behavioral modeling, based on process flow data and equipment operating specifications, basic operating behaviors of the equipment are simulated, such as start-stop logic, parameter adjustment responses, and simple linkages with other equipment. The initial digital twin does not need to achieve extremely high precision, but it must ensure data integrity and the rationality of the model structure.

[0051] Optionally, the system can import pre-processed multi-source heterogeneous data, filter out data related to geometric modeling such as equipment dimensions, spatial coordinates, and connection relationships, and establish an equipment geometric parameter database. Using 3D modeling software, it can construct 3D geometric models of each piece of equipment based on the geometric parameter database, and then combine these models according to the actual layout of the site to form the overall geometric model of the site. It can also extract equipment physical attribute data and process behavior data, input them into the physical model module and behavior model module respectively, define basic physical rules and behavioral logic, and complete the integration of the initial digital twin. It is understood that other methods can also be used to construct the initial digital twin, such as combining preliminary point cloud data from laser scanning for rapid modeling; this is not limited here.

[0052] S202. Obtain the performance parameter deviation between the simulation output of the initial digital twin and the actual operating data of the physical entity.

[0053] Specifically, the system initiates the simulation of the initial digital twin, allowing it to simulate the operating state of the physical entity according to preset operating parameters and logic. It outputs corresponding simulation data, such as parameters like equipment operating temperature, pressure, flow rate, and vibration levels, as well as data on medium transmission efficiency and parameter adjustment response time in the process. Simultaneously, the system collects real-time operating data from the physical entity, ensuring consistency between the collected parameter types and time points and the simulation output data. Subsequently, the system compares the simulation output data with the actual operating data at the same time point and for the same parameter type, calculating the numerical differences to form a performance parameter deviation dataset. This deviation data includes not only the absolute and relative deviations of individual parameters but also the comprehensive deviations of multiple related parameters, comprehensively reflecting the gap between the initial digital twin and the physical equipment in terms of physical characteristics and operating patterns.

[0054] Optionally, the system can set a uniform data acquisition time interval (e.g., 100ms) to allow the initial digital twin to output simulation data at this interval and store it in the simulation database. Simultaneously, the system controls the sensors on the physical equipment to collect actual operating data at the same time interval and store it in the actual database. The system extracts data with the same timestamp and parameter name from the two databases, such as the compressor vibration value at time t1 and the tank pressure value at time t2. Through preset deviation calculation rules (e.g., absolute deviation = |simulation value - actual value|, relative deviation = |simulation value - actual value| / actual value × 100%), the system calculates the deviation value of each parameter and integrates them to form a performance parameter deviation dataset.

[0055] S203. Collect real-time acoustic resonance fingerprint data formed during the propagation of acoustic excitation signals based on multiple acoustic sensors deployed in the equipment.

[0056] Among them, acoustic excitation signal refers to sound wave signal with specific frequency and waveform actively emitted by acoustic exciter installed on the equipment; acoustic sensor refers to device installed on the equipment to collect vibration or sound wave data generated during the propagation of acoustic excitation signal; real-time acoustic resonance fingerprint data refers to unique data characteristics that can characterize the current physical structural integrity of the equipment when the acoustic excitation signal propagates in the physical structure of the equipment and resonates due to the structural characteristics of the equipment; physical structural integrity refers to whether there are structural defects such as cracks, loosening, and wear in the main structure, connection parts, and internal components of the equipment.

[0057] Specifically, acoustic exciters pre-deployed on the system control equipment actively emit acoustic excitation signals according to a preset frequency range (e.g., 20Hz-20kHz) and transmission period (e.g., once every 5 minutes). As the acoustic excitation signals propagate through the equipment's physical structure, they interact with the equipment's structural characteristics (e.g., material, size, internal structure) to generate resonance, forming unique acoustic signal characteristics. Multiple acoustic sensors deployed at different locations on the equipment (e.g., critical connection points, wear-prone areas, structurally weak points) synchronously collect acoustic signal data during propagation, including information such as signal frequency, amplitude, and phase. The system preprocesses the collected acoustic signal data, removing environmental noise interference and extracting core features reflecting the equipment's structural state to form real-time acoustic resonance fingerprint data.

[0058] S204. The difference between the real-time acoustic resonance fingerprint data and the reference acoustic resonance fingerprint is used as part of the performance parameter deviation.

[0059] Among them, the reference acoustic resonance fingerprint refers to the acoustic resonance fingerprint data determined in advance by digital twin simulation or actual collection under the ideal health condition of the equipment, which serves as a reference standard for the integrity of the equipment structure.

[0060] Specifically, the system calls upon pre-stored benchmark acoustic resonance fingerprint data. This data is acquired when the equipment is in brand-new condition or when its structure is confirmed to be intact and in good performance, ensuring stability and representativeness. The system then compares the real-time acquired and processed acoustic resonance fingerprint data with the benchmark data feature-by-feature, including comparisons of core characteristic parameters such as peak frequency, resonant bandwidth, amplitude attenuation rate, and phase difference. The system calculates the difference for each characteristic parameter, such as peak frequency offset and amplitude attenuation rate difference, and then performs a weighted calculation to obtain a comprehensive difference value, fully reflecting the degree of deviation between the real-time acoustic resonance fingerprint and the benchmark. This difference value is incorporated into the performance parameter deviation dataset, together with previously obtained operating parameter deviations (such as temperature deviation and pressure deviation), to form a complete performance parameter deviation dataset. This ensures that the performance parameter deviation not only reflects deviations in the equipment's operating state but also reflects the health status of the equipment's physical structure, improving the comprehensiveness and accuracy of the deviation data.

[0061] S205. Establish a prior probability distribution model for performance degradation parameters based on the performance degradation patterns in the historical operating data and maintenance records of the equipment.

[0062] Historical operating data refers to various operating parameter records accumulated during the equipment's past operation, including normal operating data, abnormal fluctuation data, and parameter change data before a failure. Maintenance records refer to detailed records of operations such as inspection, maintenance, and parts replacement performed during equipment use, including maintenance time, reasons for maintenance, maintenance content, and performance recovery status after maintenance. Performance degradation law refers to the inherent law that equipment gradually declines in physical performance, operating accuracy, and reliability as usage time increases and operating load accumulates. Prior probability distribution model refers to a mathematical model pre-established based on existing historical data and empirical laws to describe the possible value range and probability distribution of performance degradation parameters.

[0063] Specifically, the system collects historical operating data and maintenance records of the equipment, filters and organizes the data, removing invalid and abnormal data, and retaining key data related to equipment performance degradation, such as equipment operating time, cumulative load, long-term trends of key parameters (e.g., vibration, pressure), frequency of failures, maintenance cycle, and comparison of performance parameters before and after maintenance. Then, data analysis methods are used to uncover performance degradation patterns. For example, the correlation between equipment operating time and vibration amplitude is analyzed to determine the trend of vibration amplitude growth with operating time; the relationship between cumulative load and equipment sealing performance is analyzed to clarify the rate of degradation of sealing performance with load accumulation; and maintenance records are used to determine the mitigation effect of different maintenance methods on performance degradation. Based on the uncovered performance degradation patterns, an appropriate probability distribution type (e.g., normal distribution, Weibull distribution, exponential distribution) is selected, and the parameters of the distribution model (e.g., mean, variance, shape parameter, scale parameter) are determined to establish a prior probability distribution model for performance degradation parameters. This model reflects the probable range of values ​​for performance degradation parameters without considering current performance parameter deviations.

[0064] Optionally, the system can extract historical operating data and maintenance records of the equipment over the past 5 years to filter out key parameters related to performance degradation (such as operating time, cumulative load, vibration amplitude, and sealing performance parameters); perform time-series analysis on the key parameters, plot parameter change trend curves, and determine the degradation function of the parameters with time or load through curve fitting, such as linear degradation function or exponential degradation function; statistically analyze the parameter value distribution at different degradation stages, and calculate the mean, variance, and other statistics for each stage; based on the statistical results and degradation function characteristics, select the Weibull distribution as the probability distribution type of the performance degradation parameters, determine the shape and scale parameters of the distribution model through maximum likelihood estimation, and construct a prior probability distribution model.

[0065] S206. Construct a likelihood function between observed and predicted values ​​based on the simulation model of the performance parameter deviation and the initial digital twin.

[0066] Among them, the observed value refers to the performance parameter data (including real-time acoustic resonance fingerprint difference value) collected during the actual operation of the physical entity equipment, which is the measured data reflecting the true state of the equipment; the predicted value refers to the corresponding performance parameter data output by the initial digital twin through the simulation model, which is the simulation result of the virtual model on the state of the equipment; the likelihood function is a probability distribution function that describes the deviation of the current performance parameter between the observed value and the predicted value under the given performance degradation parameter, and is used to quantify the possibility of deviation occurring under a specific degradation parameter.

[0067] The system analyzes the impact of performance degradation parameters on predicted values ​​based on the initial digital twin simulation model. For example, as the equipment sealing performance degradation parameter increases, the predicted valve leakage in the simulation model will rise, leading to a change in the deviation between the observed and predicted leakage values. Based on this, assuming the deviation follows a specific probability distribution (such as a normal distribution, consistent with common patterns of random errors in engineering), a likelihood function is constructed using the performance degradation parameter as a variable. The core logic of the function is "the probability of the current performance parameter deviation occurring under a specific performance degradation parameter," that is, the likelihood of the observed and predicted values ​​deviating when the performance degradation parameter takes a certain value. During the construction process, the distribution parameters (such as variance) need to be calibrated in conjunction with the equipment's physical characteristics (such as material fatigue rate and component wear patterns).

[0068] In some embodiments, the system can identify the types of performance parameter deviations (such as operating parameter deviations and structural parameter deviations), assume based on engineering experience that all deviations follow a normal distribution, since the normal distribution can better fit the deviation patterns caused by random disturbances during equipment operation; extract sample pairs of "known performance degradation parameters - corresponding deviations" from historical data (such as "seal wear amount - leakage deviation" data in equipment maintenance records), and use the maximum likelihood estimation method to calculate the mean (usually set to 0, assuming no systematic error) and variance of the normal distribution; based on the initial digital twin simulation model, establish a mapping formula between performance degradation parameters and predicted values ​​(such as for every 0.1 mm increase in seal wear amount, the predicted leakage value increases by 0.05 m³ / h), substitute it into the normal distribution probability density function, and replace it with an expression with performance degradation parameters as variables to form a likelihood function.

[0069] S207. Based on the prior probability distribution model and the likelihood function, the posterior probability distribution is obtained, and then the performance degradation parameters of the equipment are determined.

[0070] The posterior probability distribution refers to the probability distribution of performance degradation parameters obtained by combining the prior probability distribution model with the likelihood function according to Bayes' theorem. It is a comprehensive probability distribution that integrates historical experience information (prior) and real-time observation information (likelihood). Bayes' theorem refers to the formula "posterior probability = prior probability × likelihood function / evidence factor" to realize the update from prior information to posterior information. The evidence factor is a normalization constant to ensure that the posterior probability distribution satisfies the condition that the sum of probabilities is 1.

[0071] Specifically, the system calls the prior probability distribution model of performance degradation parameters established in step S205 (reflecting the probability law of degradation parameters based on historical data) and the likelihood function constructed in step S206 (reflecting the probability law of degradation parameters based on real-time deviation). According to Bayes' theorem, the prior probability distribution is multiplied by the likelihood function to obtain the unnormalized posterior probability distribution. For example, if the probability of "0.2mm of seal wear" in the prior distribution is 20%, and the likelihood function shows that the probability of the current deviation occurring under this wear amount is extremely high, then multiplying will significantly increase the probability of this wear amount. Subsequently, the system calculates the evidence factor (i.e., the sum of the integrals of "prior probability × likelihood function" under all possible degradation parameters) to normalize the unnormalized posterior distribution, ensuring that it satisfies the basic properties of probability distribution. Finally, based on the normalized posterior probability distribution, the degradation parameter value with the highest probability density (i.e., the mode of the posterior distribution) is selected as the current performance degradation parameter of the equipment. At the same time, the confidence interval of the parameter is output (e.g., the wear amount is 0.18-0.22mm at 95% confidence level) to quantify the uncertainty of the parameter.

[0072] S208. Based on the performance degradation parameters, the physical model and behavioral model are modified to obtain a full-element digital twin.

[0073] Specifically, the system identifies the corresponding parameter in the initial physical model based on the type of performance degradation parameters (such as seal wear, material fatigue, and increased bearing clearance). For example, when the seal wear is 0.2 mm, the "valve sealing coefficient" in the physical model is corrected from the initial value of 0.95 to 0.7 (a decrease in the coefficient indicates a degradation in sealing performance); when the material fatigue reaches 30%, the "equipment compressive strength parameter" is corrected from the initial value of 10 MPa to 7 MPa. During the correction process, the physical laws of equipment failure must be considered (e.g., for every 0.1 mm increase in wear, the sealing coefficient decreases by 0.15) to ensure that the parameter adjustments conform to the actual performance degradation logic.

[0074] Subsequently, based on the revised physical model, the simulation rules for equipment operation behavior are adjusted. For example, after the sealing coefficient decreases, the "valve leakage calculation logic" in the behavior model needs to be updated synchronously to keep the deviation between the simulated leakage and the actual measured leakage within 3%. After the compressive strength decreases, the "equipment overload protection trigger threshold" in the behavior model needs to be lowered from the initial 10MPa to 7MPa to avoid the virtual model misjudging the equipment's safety status. After the correction is completed, the system ensures the model's fidelity through "real-time data comparison and verification": the revised twin simulation output is compared with the real-time data of the physical entity. If the deviation exceeds a preset threshold (e.g., 2%), the "degradation parameter fine-tuning - model correction - verification" process is repeated until the deviation meets the requirements, ultimately forming a full-element digital twin.

[0075] S209. Based on the fault type and fault state determined by the prediction results, as well as the scenario parameters of the initial accident scenario, construct an accident simulation scenario in the virtual environment.

[0076] Specifically, the system acquires the fault type (e.g., "internal leakage at the tank inlet valve"), key parameters of the fault status (e.g., leakage rate of 0.6 m³ / h, current valve opening / closing degree of 90%), and scenario parameters of the initial accident scenario (e.g., real-time wind speed of 2.5 m / s, wind direction of southwest, station ambient temperature of 25℃, and distance between adjacent valves of 5 m). The scenario is then constructed within the virtual environment of a full-element digital twin. The first step is to locate the faulty equipment and mark the fault point in the virtual model (e.g., highlight the inlet valve of the storage tank in red). The second step is to set the fault status parameters, setting the internal leakage parameter of the virtual valve to 0.6 m³ / h to simulate the initial rate of media leakage. The third step is to load scene parameters and recreate the current wind speed, wind direction (simulating the direction of gas diffusion through particle effects) and ambient temperature (affecting the gas evaporation rate) in the virtual environment, while matching the current process load (such as a tank feed flow rate of 20 m³ / h, which affects the pressure change inside the tank after the leak). The fourth step is to associate the status of surrounding devices to ensure that the initial status of adjacent devices in the virtual scene (such as outlet valves and combustible gas detectors) is consistent with that of the physical devices (such as outlet valves being fully open and detectors having normal sensitivity).

[0077] Once the system is built, it initiates scenario pre-simulation verification: simulating the initial 10 seconds of an accident (e.g., the spread of leaked gas reaches 3m), comparing the virtual diffusion rate with the theoretical calculation value (based on a fluid dynamics model). If the deviation exceeds 5%, the scenario parameters are adjusted (e.g., the gas diffusion coefficient is corrected) until the scenario conforms to actual physical laws, ensuring the reliability of subsequent drill results.

[0078] S210. Obtain at least one preset processing scheme, each preset processing scheme including a processing operation sequence and an operation timing.

[0079] Specifically, the system matches a library of preset handling solutions based on the core characteristics of the accident simulation scenario (fault type, fault status, scenario parameters). For example, for the scenario of "internal leakage of the inlet valve of the storage tank (leakage rate 0.6 m³ / h) + southwest wind 2.5 m / s", the system selects three relevant preset solutions from the solution library: Solution 1 (close the inlet valve first → then close the outlet valve → start the combustible gas detection), Solution 2 (close the inlet and outlet valves simultaneously → start the fire sprinkler → isolate the surrounding area), and Solution 3 (start the fire sprinkler first → then close the inlet valve → evacuate personnel).

[0080] Each plan includes a complete sequence of operations and timing. For example, in Plan 2, the operation sequence is: "1. Remotely close the tank inlet valve; 2. Remotely close the tank outlet valve; 3. Start the tank area fire sprinkler system; 4. Set up an isolation zone to prevent personnel from entering." The timing is: "Operation 1 and Operation 2 are started simultaneously, and the closure must be completed within 10 seconds; Operation 3 is started within 5 seconds after the inlet and outlet valves are closed; Operation 4 is executed within 10 seconds after Operation 3 is started."

[0081] After acquiring the solutions, the system verifies their completeness and adaptability: it checks whether the operation sequence of each solution covers the core links of "fault control → risk isolation → personnel protection" and whether the operation timing meets the equipment's response capability (e.g., the actual shortest time for remote valve closure is 8 seconds, and the timing set in the solution to 10 seconds meets the requirements). If a solution has missing operations (e.g., it does not include leak detection) or unreasonable timing (e.g., it requires closing the valve within 5 seconds, which exceeds the equipment's capability), it is marked as "to be optimized" and excluded from this exercise to ensure that the solutions participating in the exercise have basic feasibility.

[0082] S211. The accident development process is simulated by sequentially executing the handling operation sequence of each preset handling plan in the accident simulation scenario, and the exercise results corresponding to each preset handling plan are obtained.

[0083] Specifically, the system allocates an independent simulation thread for each preset handling scheme to avoid interference between simulations of multiple schemes. Taking a certain preset handling scheme as an example, the system executes operations step by step in the accident simulation scenario according to the sequence and timing of the handling operations: First, the first handling operation is triggered (such as "remotely closing the tank inlet valve"), and the start time of the operation and the equipment response delay (such as the time from receiving the command to the valve starting to close) are recorded simultaneously; Second, during the operation execution, the virtual scenario status is updated in real time—for example, during the valve closing process, as the leakage rate gradually decreases from the initial value, the system calculates the change in the diffusion range of the leaked gas through a fluid dynamics model, adjusts the diffusion trajectory in combination with wind speed and wind direction parameters, and monitors the changes in the values ​​of surrounding combustible gas detectors (such as leakage rate). After the reduction, the detector value gradually returns from the excessive state); the third step is to execute subsequent operations according to the operation sequence (such as "starting the fire sprinkler 5 seconds after the inlet valve is closed"), and continuously track the dynamic changes of key accident indicators (such as accident control time, total leakage, and scope of impact); the fourth step is to summarize the key data of the entire process after the simulation ends (usually ending when the accident is effectively controlled, such as when the leakage drops below the safety threshold), including operation execution time, accident status parameters at each time point, equipment resource consumption (such as fire sprinkler water consumption), etc., to form the exercise results corresponding to the scheme, and at the same time generate a visual video record of the simulation process.

[0084] S212. Extract quantitative evaluation indicators from each exercise result, and calculate the exercise score corresponding to each preset processing plan based on the quantitative evaluation indicators.

[0085] Specifically, the system constructs the scope and standards for extracting quantitative evaluation indicators: core indicators include accident control time (the time from the execution of the first response operation to the accident meeting the "control standard", such as leakage amount ≤0.01m³ / h), total leakage amount (the total leakage volume of gas during the simulation process), and impact range (the maximum diffusion radius of the accident or the area of ​​the affected area); auxiliary indicators include the success rate of response operations (such as the percentage of successful operations such as valve closure and sprinkler activation) and response costs (such as the economic cost of fire water consumption and equipment energy consumption).

[0086] During the extraction process, the system cleans the raw data in the exercise results, removing outliers (such as instantaneous data fluctuations during the simulation). Then, the system calculates the exercise score according to preset scoring rules: First, it determines the weight of each indicator (e.g., accident control time 40%, total leakage 30%, impact range 20%, operation success rate 10%), and the weight setting must be combined with the safety priorities of the gas station (e.g., "shortest control time" is more important than "lowest cost"); Second, it standardizes each indicator (e.g., converting accident control time into a normalized score of "1 - actual time / baseline time", where the baseline time is the average control time of similar accidents in the industry) to avoid the influence of differences in indicator dimensions on the score; Third, it multiplies the standardized indicator score by its corresponding weight and sums the results to obtain the exercise score for the preset processing scheme, which is usually based on a percentage system.

[0087] S213. Select the preset handling plan with the highest exercise score as the optimal handling plan for the initial accident scenario.

[0088] Specifically, the system sorts the scores of all preset processing schemes in descending order to determine the ranking of each scheme (e.g., scheme B scores 92 points, scheme A scores 85 points, and scheme C scores 78 points). After sorting, the system needs to perform secondary verification on the highest-scoring solution (e.g., Solution B) to ensure its optimality: verify the accuracy of the score calculation, review the extraction and calculation process of the quantitative evaluation indicators for the solution (e.g., whether there are errors in indicator data or biases in weight application), and avoid misselection of the solution due to calculation errors; verify the adaptability of the solution to the initial accident scenario, confirm that the solution's handling operations (e.g., "simultaneously close the inlet and outlet valves + start the fire sprinkler") can cover the core risks of the initial scenario (e.g., "the leakage spreads rapidly in windy weather," and fire sprinklers can suppress the spread), and that the operation sequence is consistent with the actual response capability of the on-site equipment (e.g., the valve closing time is 10 seconds, the shortest measured closing time of the equipment is 8 seconds, and the sequence is reasonable); evaluate the feasibility of the solution, check whether there are operational conflicts in the solution (e.g., "whether operating multiple valves simultaneously exceeds the capabilities of maintenance personnel"), and whether special equipment support is required (e.g., whether it is necessary to call a backup fire pump), to ensure that the solution can be implemented in the actual scenario.

[0089] After the secondary verification is passed, the system officially identifies the solution as the optimal handling solution for the initial accident scenario and generates a detailed solution report, including operation steps, timing requirements, risk warnings (such as "after closing the valve, the sealing performance should be checked to prevent secondary leakage"), which is then pushed to the mobile devices and central control screens of the site maintenance personnel.

[0090] S214. Based on the historical execution data of operators who are scheduled to execute the optimal handling plan, the reserved time margin of each handling operation in the optimal handling plan is adjusted to obtain a personalized handling plan.

[0091] Specifically, the system obtains the identity information (such as employee ID) of the scheduled operator and retrieves their historical execution data from the personnel operation database. For example, the historical data of operator A is "average execution time of 12 seconds for similar valve closing operations, with a success rate of 95%; average execution time of 8 seconds for starting fire sprinklers, with a success rate of 100%".

[0092] Subsequently, the system analyzes the current status of the reserved time margin for each operation in the optimal disposal plan: for example, the reserved time margin for "closing the inlet valve" in the plan is 10 seconds (theoretical operation time 8 seconds + buffer time 2 seconds), and the reserved time margin for "starting the fire sprinkler" is 8 seconds (theoretical operation time 6 seconds + buffer time 2 seconds).

[0093] Next, the system adjusts the time margin based on historical execution data: First, it compares the operator's historical average execution time with the theoretical operation time of the plan. If the historical average time is longer than the theoretical time (e.g., operator A's valve closing average is 12 seconds > theoretical 8 seconds), the margin is adjusted by "historical average time + basic buffer time (e.g., 3 seconds)" (e.g., the valve closing margin is adjusted to 12 + 3 = 15 seconds) to avoid operational errors due to insufficient time. If the historical average time is shorter than the theoretical time (e.g., operator A's sprinkler start-up average is 8 seconds > theoretical 6 seconds, it still needs to be adjusted according to the actual average time), the buffer time is appropriately compressed, but a minimum safety buffer (e.g., at least 2 seconds) must be retained. Second, the margin is adjusted based on the operation success rate. If the operator's success rate for a certain type of operation is low (e.g., <90%), an additional buffer time is added (e.g., an additional 2 seconds) to allow time for correcting operational errors. If the success rate is high (e.g., ≥98%), the buffer time can be maintained or slightly reduced. After the adjustment is completed, the system generates a personalized handling plan, which specifies the adjustment margin for each operation (e.g., "Close the inlet valve: complete within 15 seconds, start the fire sprinkler: complete within 10 seconds") and marks the basis for the adjustment (e.g., "Based on operator A's average valve closing time of 12 seconds, the margin is increased by 5 seconds").

[0094] The intelligent safety management and control system for gas stations based on digital twins, as described in this invention, is applied to electronic devices. Figure 3 A schematic diagram of the architecture of an electronic device suitable for implementing embodiments of the present invention is shown.

[0095] It should be noted that, Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0096] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by instructions (computer programs), or by instructions (computer programs) controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. The electronic device of this embodiment includes a storage medium and a processor, wherein the storage medium stores multiple instructions that can be loaded by the processor to execute any step of the method provided in the embodiments of the present invention.

[0097] Specifically, the storage medium and the processor are electrically connected directly or indirectly to enable data transmission or interaction. For example, these components can be electrically connected to each other via one or more signal lines. The storage medium stores computer-executable instructions that implement data access control methods, including at least one software functional module that can be stored in the storage medium in the form of software or firmware. The processor executes various functional applications and data processing by running the software program and module stored in the storage medium. The storage medium can be, but is not limited to, Random Access Memory (RAM), Read-Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The storage medium stores the program, and the processor executes the program after receiving the execution instructions.

[0098] Furthermore, the software programs and modules within the aforementioned storage medium may also include an operating system, which may include various software components and / or drivers for managing system tasks (e.g., memory management, storage device control, power management, etc.) and may communicate with various hardware or software components to provide an operating environment for other software components. The processor may be an integrated circuit chip with signal processing capabilities. The aforementioned processor may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc., which can implement or execute the methods, steps, and logic flowcharts disclosed in this embodiment. The general-purpose processor may be a microprocessor or any conventional processor.

[0099] Since the instructions stored in the storage medium can execute the steps in any of the methods provided in the embodiments of the present invention, the beneficial effects of any of the methods provided in the embodiments of the present invention can be achieved, as detailed in the preceding embodiments, and will not be repeated here.

[0100] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A digital twin-based intelligent safety management and control method for a gas station, applied to a digital twin-based intelligent safety management and control system for a gas station, characterized in that, The method comprises: Real-time collection of multi-source heterogeneous data in the gas station, the multi-source heterogeneous data comprising device data, station environment data, and process flow data; Construction of a full-factor digital twin of the gas station based on the multi-source heterogeneous data, the full-factor digital twin integrating a geometric model, a physical model, and a behavior model of the device; Extraction of multi-dimensional feature data from the digital twin, the multi-dimensional feature data comprising historical operation data, real-time operation data, station environment data, and process flow data, input into a pre-trained long short-term memory network model to predict a future state of the device, to obtain a prediction result; Simulation of a development process of at least one preset handling scheme in a virtual environment corresponding to the digital twin based on the prediction result and a preset initial accident scenario, to obtain an exercise result corresponding to each preset handling scheme; Determination of an optimal disposal scheme corresponding to the initial accident scenario based on an exercise score corresponding to each exercise result.

2. The method of claim 1, wherein, The step of constructing the full-factor digital twin of the gas station based on the multi-source heterogeneous data comprises: Construction of an initial digital twin of the device based on the multi-source heterogeneous data, the initial digital twin comprising a geometric model, a physical model, and a behavior model; Obtaining a performance parameter deviation between a simulation output of the initial digital twin and actual operation data of the physical entity; Inversion calculation of a performance degradation parameter of the device based on a Bayesian inference algorithm and the performance parameter deviation; Correction of the physical model and the behavior model based on the performance degradation parameter, to obtain the full-factor digital twin.

3. The method of claim 2, wherein, The step of inversion calculation of the performance degradation parameter of the device based on the Bayesian inference algorithm and the performance parameter deviation comprises: Establishment of a prior probability distribution model of the performance degradation parameter based on a performance degradation law in historical operation data and maintenance records of the device; Construction of a likelihood function between an observation value and a prediction value based on the performance parameter deviation and a simulation model of the initial digital twin, the likelihood function being used to represent a probability of producing the performance parameter deviation under a given performance degradation parameter; Multiplication of the prior probability distribution model and the likelihood function according to Bayes' theorem, to obtain a posterior probability distribution of the performance degradation parameter; Determination of the performance degradation parameter of the device based on the posterior probability distribution.

4. The method of claim 1, wherein, The step of simulating the development process of at least one preset handling scheme in the virtual environment corresponding to the digital twin based on the prediction result and the preset initial accident scenario comprises: Determination of a fault type and a fault state of a faulty device according to the prediction result; Construction of an accident simulation scenario in the virtual environment according to the fault type, the fault state, and scenario parameters of the initial accident scenario; Obtaining of at least one preset handling scheme, each preset handling scheme comprising a disposal operation sequence and an operation timing; Simulation of an accident development process by sequentially executing the disposal operation sequence of each preset handling scheme in the accident simulation scenario, to obtain an exercise result corresponding to each preset handling scheme.

5. The method of claim 1, wherein, The step of determining the optimal disposal scheme corresponding to the initial accident scene based on the exercise score corresponding to each exercise result specifically comprises: extracting a quantitative evaluation index from each exercise result, the quantitative evaluation index including an accident control time, a total leakage amount, and an influence range; calculating an exercise score corresponding to each preset disposal scheme based on the quantitative evaluation index; selecting a preset disposal scheme with the highest exercise score as the optimal disposal scheme corresponding to the initial accident scene.

6. The method of claim 5, wherein, After the step of selecting a preset disposal scheme with the highest exercise score as the optimal disposal scheme corresponding to the initial accident scene, the method further comprises: obtaining historical execution data of an operator scheduled to execute the optimal disposal scheme, the historical execution data including a historical average execution time and an operation success rate; adjusting a reserved time margin of each disposal operation in the optimal disposal scheme based on the historical average execution time and the operation success rate to obtain a personalized disposal scheme.

7. The method of claim 2, wherein, After the step of obtaining the performance parameter deviation between the simulation output of the initial digital twin and the actual operation data of the physical entity, the method further comprises: setting at least one acoustic exciter for actively emitting a preset acoustic excitation signal on the equipment; collecting real-time acoustic resonance fingerprint data formed by the acoustic excitation signal in the propagation process according to a plurality of acoustic sensors arranged on the equipment, the real-time acoustic resonance fingerprint data being used to represent the current physical structural integrity of the equipment; taking a difference value between the real-time acoustic resonance fingerprint data and a reference acoustic resonance fingerprint as a part of the performance parameter deviation, the reference acoustic resonance fingerprint being determined in the digital twin in an ideal healthy state in advance.

8. A gas field station intelligent safety management and control system based on digital twinning, characterized in that, The system comprises one or more processors and a memory; The memory is coupled to the one or more processors, and the memory is configured to store computer program codes, the computer program codes comprising computer instructions, and the one or more processors are configured to invoke the computer instructions to enable the system to perform the method according to any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instructions are run on the digital-twin-based gas field station intelligent safety management and control system, the system is enabled to perform the method according to any one of claims 1-7.

10. A computer program product, characterised in that, When the computer program product is run on the digital-twin-based gas field station intelligent safety management and control system, the system is enabled to perform the method according to any one of claims 1-7.

Images