Gas equipment protection method and device, server and storage medium

Abstract

The invention provides a gas equipment protection method and device, a server and a storage medium. The method comprises the steps of obtaining operation data in an SCADA database in real time; constructing a DBN model by taking the operation data as leaf nodes, the implicit risk factor as an intermediate node and the dynamic failure probability as a root node; selecting current observation data from the operation data based on a sliding window mechanism; in the DBN model, dynamic reasoning is executed in combination with current observation data, and a dynamic failure probability is output; and generating a graded early warning signal based on the dynamic failure probability, and transmitting the graded early warning signal back to the SCADA system in real time, so that the SCADA system controls the gas equipment to perform a protection action in a linkage manner based on the graded early warning signal. In the technical scheme provided by the invention, the relevance among the nodes in the model is deeply mined through the dynamic Bayesian network model, the real-time reasoning and updating of the dynamic failure probability are realized, and the real-time performance and timeliness of the dynamic failure probability are improved.

Description

Technical Field

[0001] This invention relates to the field of data acquisition and monitoring technology, specifically to a method, device, server, and storage medium for protecting gas equipment. Background Technology

[0002] In related technologies, natural gas, as a clean and high-calorific-value energy source, is receiving increasing attention from various countries. Vigorously developing the natural gas industry to achieve sound, rapid, and sustainable economic development has become an important indicator for evaluating comprehensive national strength in today's society. With the acceleration of urbanization and the increase in energy demand, natural gas, as a clean and efficient energy source, is being used more and more widely in various fields such as industry, residential use, and transportation. Gas stations, as important nodes in the natural gas supply network, undertake key tasks such as gas storage, allocation, and distribution. The safety of their operation is directly related to the safety of people's lives and property and social stability.

[0003] As a critical node in the urban gas transmission and distribution system, the failure of equipment (such as pipelines, valves, and pressure regulating devices) at gas stations can lead to serious safety accidents. In recent years, with the widespread application of Supervisory Control and Data Acquisition (SCADA) systems in the gas industry, the amount of real-time monitoring data has grown exponentially. However, how to utilize this data to achieve dynamic risk assessment remains a challenge for the industry.

[0004] In existing technologies, failure probabilities are typically calculated based on historical failure data (such as Weibull distribution and Poisson process). These methods assume that the failure probability is constant and cannot reflect real-time changes in operating conditions (such as pressure fluctuations and sudden changes in corrosion rates), resulting in a serious lag in risk assessment under sudden operating conditions. Summary of the Invention

[0005] In view of the above problems, embodiments of the present invention provide a gas equipment protection method, device, server and storage medium to solve the problem that the failure probability calculation method in the prior art cannot reflect real-time operating condition changes, resulting in a serious lag in risk assessment under sudden operating conditions.

[0006] In a first aspect, embodiments of the present invention provide a method for protecting gas equipment, the method comprising: Real-time acquisition of operational data from the SCADA database; The DBN model is constructed by using the running data as leaf nodes, the implicit risk factors as intermediate nodes, and the dynamic failure probability as the root node. Based on the sliding window mechanism, the current observation data is selected from the running data; In the DBN model, dynamic inference is performed based on the current observation data to output the dynamic failure probability; Based on the dynamic failure probability, a graded early warning signal is generated and transmitted back to the SCADA system in real time, so that the SCADA system can control the gas equipment to perform protective actions in conjunction with the graded early warning signal.

[0007] In one possible implementation, after constructing the DBN model, the runtime data is used as leaf nodes, implicit risk factors as intermediate nodes, and dynamic failure probabilities as root nodes. This further includes: Based on the online structural learning algorithm, newly added risk nodes are automatically identified according to the operational data; The parent node of the newly added risk node is determined from the DBN model; The newly added risk node is added to the DBN model based on the parent node to obtain the updated DBN model.

[0008] In one possible implementation, the online structure learning algorithm includes constraint testing, Bayesian online structure learning algorithm, incremental machine learning algorithm, or graph neural network algorithm.

[0009] In one possible implementation, the step of performing dynamic inference in the DBN model based on the current observation data and outputting a dynamic failure probability includes: A set of posterior samples is generated based on the current observation data using the MCMC algorithm; Calculate the failure probability of each sample at the target time point, average all failure probabilities to obtain a Bayesian estimate; or, read the failure state of each sample and use the proportion of samples in the failure state to all posterior samples as the Bayesian estimate. The Bayesian estimation is the Bayesian estimation of the dynamic failure probability.

[0010] In one possible implementation, the DBN model employs an approximate inference algorithm for high-frequency nodes and an exact inference algorithm for low-frequency key nodes to achieve dynamic inference.

[0011] In one possible implementation, the method further includes: Retrieve environmental data from an external database; A knowledge graph model is constructed using the gas equipment, environmental data, implicit risk factors, and protective actions as nodes, and the relationships between nodes as connecting edges.

[0012] In one possible implementation, in the DBN model, the posterior probability of the previous sliding window is used as the prior probability of the current sliding window; or, the statistical parameters of the previous sliding window are used as the weak prior of the current sliding window.

[0013] Secondly, embodiments of the present invention provide a gas equipment protection device, the device comprising: The first acquisition module is used to acquire running data from the SCADA database in real time; The first building module is used to construct the DBN model by using the running data as leaf nodes, the implicit risk factors as intermediate nodes, and the dynamic failure probability as the root node. The selection module is used to select the current observation data from the running data based on a sliding window mechanism; The inference module is used to perform dynamic inference in the DBN model by combining the current observation data and output the dynamic failure probability. The generation module is used to generate graded early warning signals based on the dynamic failure probability, and to transmit the graded early warning signals back to the SCADA system in real time, so that the SCADA system can control the gas equipment to perform protective actions based on the graded early warning signals.

[0014] Thirdly, embodiments of the present invention provide a server, the server including one or more processors; a memory; and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions that, when executed by the server, cause the server to perform the gas equipment protection method as described in the first aspect or any possible implementation of the first aspect.

[0015] Fourthly, embodiments of the present invention provide a computer-readable storage medium comprising a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform the gas equipment protection method as described in the first aspect or any possible implementation thereof.

[0016] The technical solution provided in this invention utilizes a dynamic Bayesian network model to deeply explore the correlations between nodes in the model, enabling real-time reasoning and updating of dynamic failure probabilities, thus improving the real-time performance and timeliness of dynamic failure probabilities. Furthermore, it provides early warnings through tiered warning signals, ensuring the safe operation of gas equipment. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a gas equipment protection method based on a SCADA system, provided as an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram of the structure of a gas equipment protection device provided in an embodiment of the present invention.

[0019] Figure 3This is a schematic diagram of a server provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer and more understandable, the invention will be further described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0021] Figure 1 This is a flowchart illustrating a gas equipment protection method based on a SCADA system, as provided in an embodiment of the present invention. Figure 1 As shown, the method includes: Step 101: Obtain the running data from the SCADA database in real time.

[0022] In this embodiment of the invention, each step is executed by a server deployed at a gas station or a higher-level center. The server is equipped with a database and Bayesian computation software. The server includes a local server and a cloud server. The local server is used to process real-time tasks, while the cloud server is used to process non-real-time tasks, balancing real-time performance and stability. This ensures low latency and high reliability for real-time tasks while reducing resource costs for non-real-time tasks.

[0023] In this step, the SCADA database is a component of the SCADA system, which is deployed at the gas station. Operational data includes physical quantities (such as pressure, temperature, and flow rate), equipment status (such as valve opening and closing signals, vibration frequency, seal wear index, and leakage alarm signals), and environmental parameters (such as soil corrosion rate and humidity). After data acquisition, the operational data undergoes preprocessing, such as outlier filtering, data alignment, and data normalization.

[0024] Step 102: Construct the DBN model by using the running data as leaf nodes, the implicit risk factors as intermediate nodes, and the dynamic failure probability as the root node.

[0025] In this step, implicit risk factors refer to risk data that cannot be directly observed. For example, implicit risk factors include micro-leakage risk and corrosion risk.

[0026] In this embodiment of the invention, after step 102, the method further includes: automatically identifying newly added risk nodes based on running data using an online structural learning algorithm; determining the parent node of the newly added risk node from the DBN model; and adding the newly added risk node to the DBN model based on the parent node to obtain an updated DBN model. In the updated DBN model, the newly added risk node participates in the dynamic inference process of dynamic failure probability. For scenarios with high real-time requirements, an approximate inference algorithm is used to implement dynamic inference; for scenarios with high accuracy requirements, an exact inference algorithm is used to implement dynamic inference. The online structural learning algorithm includes constraint testing, Bayesian online structural learning, incremental machine learning, or graph neural network algorithms.

[0027] In the DBN model, the conditional probabilities between nodes are quantified using a Conditional Probability Table (CPT). The CPT includes single-time-slice conditional probabilities and cross-time-slice conditional probabilities. For example, in the DBN model, the leaf nodes represent temperature and pressure, the implicit risk factor is corrosion risk, and the root node represents dynamic failure probability. The single-time-slice conditional probability is the conditional probability of corrosion risk at time t, given that the temperature and pressure at time t are known; the cross-time-slice conditional probability is the conditional probability of corrosion risk at time t+1, given that the corrosion risk at time t is known.

[0028] It is important to note that after determining the parent node of the newly added risk node from the DBN model, the CPT is updated based on the parent node and the newly added risk node. Specifically, the update process involves adding the relevant parameters of the newly added risk node to the CPT and updating the conditional probabilities of existing nodes in the DBN model based on the newly added risk node.

[0029] Step 103: Select the current observation data from the running data based on the sliding window mechanism.

[0030] In this step, the sliding window can be a fixed-length sliding window or a weighted exponentially decaying window. When using an exponentially decaying window, the weight of older running data decreases, while the weight of newer running data increases accordingly. The length of the sliding window can be set based on actual needs. For example, the sliding window can select running data from the past N hours or the most recent M running data points. After selecting the current observation data, it is preprocessed, for example, through denoising, discretization, and standardization.

[0031] In existing technologies, CPT relies on historical data for training, with update cycles lasting weeks or even months. When SCADA systems generate large amounts of data per second, the model cannot respond dynamically, leading to inaccurate predictions. In this embodiment of the invention, when the server collects new operational data, a sliding window is triggered. This sliding window mechanism ensures that the latest operational data is retained within the sliding window, avoiding excessive stale data and the problem of untimely CPT updates, thus preventing potential safety or economic losses.

[0032] Step 104: In the DBN model, perform dynamic inference based on the current observation data and output the dynamic failure probability.

[0033] In this step, the maximum likelihood estimation method is used to generate an initial CPT based on historical operating data. When current observation data indicates that the relationships between certain nodes in the DBN model have become stronger or weaker, the initial CPT needs to be updated to ensure its timeliness, i.e., that the CPT reflects current data patterns. For example, if current observation data indicates that the gas equipment is operating in a high-temperature, low-pressure environment, leading to an increased risk of corrosion, then the initial CPT needs to be updated. It is important to note that when updating the initial CPT, it is necessary to determine which entries in the CPT are allowed to be updated over time and which remain fixed; not all entries in the initial CPT are updated.

[0034] In this embodiment of the invention, in the DBN model, the posterior probability of the previous sliding window is used as the prior probability of the current sliding window; or, the statistical parameters of the previous sliding window are used as the weak prior of the current sliding window, so as to combine the current observation data and the prior data for parameter estimation, thereby achieving memory retention and avoiding learning from scratch each time. If there are latent variables (such as latent risk factors), the posterior of the latent variables is also processed.

[0035] In this embodiment of the invention, the Bayesian Expectation-Maximization (BEM) algorithm is used to update the initial CPT in the DBN model. In the expectation step, the posterior probability expectations of the latent risk factors and device states are calculated. In the maximization step, the posterior probability expectations are combined with prior pseudo-counts to update the initial CPT. To balance robustness and fast response, prior smoothing, forgetting factor control, and local incremental update strategies are employed. Furthermore, drift detection, performance monitoring, and rollback mechanisms are implemented to ensure the reliability of online learning.

[0036] In this embodiment of the invention, the updated CPT is smoothed to avoid zero probability and the magnitude of each update is limited to prevent drastic disturbances from outliers. If an anomaly or concept drift is detected in the current observation data, the update rate is temporarily reduced or the update is paused, and an alarm message is generated.

[0037] In this embodiment of the invention, the dynamic inference process of the DBN model is as follows: Using the Markov Chain Monte Carlo Method (MCMC), a set of posterior samples is generated based on the current observation data. The failure probability of each sample at the target time point is calculated, and the average of all failure probabilities is used to obtain a Bayesian estimate. Alternatively, the failure state of each sample is read, and the proportion of samples in a failed state to all posterior samples is used as the Bayesian estimate. This Bayesian estimate is the Bayesian estimate of the dynamic failure probability. Optionally, uncertainty information such as confidence intervals can be obtained through the posterior samples to evaluate the accuracy of the failure probability. Commonly used MCMC algorithms include Gibbs sampling and Metropolis-Hastings. Alternatively, a hierarchical strategy can be adopted, using Metropolis sampling for the running data and Gibbs sampling for the hidden states.

[0038] In existing technologies, traditional Bayesian network inference algorithms (such as connection tree algorithms) struggle to handle high-dimensional real-time data and cannot meet the requirements for real-time early warning. In this embodiment of the invention, for high-dimensional CPT (Continuous Predictive Time), only local sub-tables strongly correlated with the current observation data are updated, saving significant computational resources. This ensures that CPT meets timeliness requirements while also maintaining CPT update efficiency.

[0039] In the DBN model, approximate inference algorithms (such as variational Bayesian algorithms) are used for high-frequency nodes (e.g., pressure), while exact inference algorithms are used for low-frequency critical nodes (e.g., corrosion risk), balancing inference speed and accuracy. For example, exact inference algorithms include Junction-Tree (clustered tree / likelihood propagation), Variable Elimination / Forward-Backward algorithms. Furthermore, in continuous linear-Gaussian scenarios, a Kalman filter + RTS smoother scheme is used to achieve exact inference; in mixed scenarios (discrete-continuous), a conditional linear-Gaussian exact propagation scheme on the clustered tree is used to achieve exact inference. In practical applications, the specific exact inference algorithm is selected based on the type (discrete / continuous / mixed), structure (chain / tree / dense), and computational resources; this embodiment of the invention does not limit this selection.

[0040] Step 105: Based on the dynamic failure probability, generate graded early warning signals and transmit the graded early warning signals back to the SCADA system in real time, so that the SCADA system can control the gas equipment to perform protective actions based on the graded early warning signals.

[0041] In this step, there is a correspondence between the dynamic failure probability and the graded early warning signals. Different graded early warning signals correspond to different dynamic failure probability ranges. The graded early warning signals include low-risk, medium-risk, high-risk, and extremely high-risk signals. For example, a dynamic failure probability of 0-0.5 corresponds to a low-risk early warning signal; 0.5-0.7 corresponds to a medium-risk signal; 0.7-0.85 corresponds to a high-risk signal; and 0.85-1 corresponds to an extremely high-risk signal. The correspondence between the dynamic failure probability and the graded early warning signals can be adjusted based on actual needs, and this embodiment of the invention does not limit this. It should be noted that after updating the DBN model, the graded early warning strategy and the linkage control strategy need to be updated accordingly to improve the accuracy of the early warning and the safety of the gas equipment.

[0042] For example, protective actions include closing valves. When the DBN model predicts an increase in leakage risk and the dynamic failure probability is 0.8, the server generates a high-risk warning signal and sends it to the SCADA system. Based on the high-risk warning signal, the SCADA system controls the valves upstream and downstream of the leak point to form a local isolation and prevent the risk from spreading.

[0043] For example, protective actions include pressure relief actions. When the DBN model predicts an increase in overpressure risk and the dynamic failure probability is 0.8, the server generates a high-risk warning signal and sends it to the SCADA system. Based on the high-risk warning signal, the SCADA system controls the partial opening of the pressure relief valve, allowing some gas to be discharged into the safety vent pipe or flare system, thereby reducing the internal pressure.

[0044] In this embodiment of the invention, the method further includes: acquiring environmental data from an external database; and constructing a knowledge graph model using gas equipment, environmental data, implicit risk factors, and protective actions as nodes, and the relationships between nodes as connecting edges. Gas equipment includes gas pipeline sections, valves, storage tanks, and pump stations, such as pipeline section P-101 and valve V-21. Environmental data includes Geographic Information System (GIS) data, meteorological data (such as wind speed and rainfall), and construction data (such as construction area). Implicit risk factors include corrosion risk, leakage risk, and micro-explosion risk. Protective actions include valve opening and closing actions, pressure regulation actions, and pressure relief actions. Relationships between nodes include causal relationships, influence relationships, or temporal relationships. Correspondingly, connecting edges include causal edges, influence edges, and time-series edges. Causal edges represent the impact of objective causes on equipment status, such as heavy rainfall directly accelerating corrosion. Influence edges represent the impact of operational behaviors on equipment status, such as closing valves causing a drop in pipeline pressure. Time-series edges represent the change in node status over time, such as the corrosion risk from time t to time t+1.

[0045] The technical solution provided in this invention utilizes a dynamic Bayesian network model to deeply explore the correlations between nodes in the model, enabling real-time reasoning and updating of dynamic failure probabilities, thus improving the real-time performance and timeliness of dynamic failure probabilities. Furthermore, it provides early warnings through tiered warning signals, ensuring the safe operation of gas equipment.

[0046] Figure 2 This is a schematic diagram of the structure of a gas equipment protection device provided in an embodiment of the present invention, as shown below. Figure 2 As shown, the device includes a first acquisition module 11, a first construction module 12, a selection module 13, an inference module 14, and a generation module 15. The first acquisition module 11 is used to acquire operational data from the SCADA database in real time; the first construction module 12 is used to construct a DBN model using the operational data as leaf nodes, implicit risk factors as intermediate nodes, and dynamic failure probability as the root node; the selection module 13 is used to select currently observed data from the operational data based on a sliding window mechanism; the inference module 14 is used to perform dynamic inference in the DBN model, combining the current observed data, and output the dynamic failure probability; the generation module 15 is used to generate graded early warning signals based on the dynamic failure probability and transmit the graded early warning signals back to the SCADA system in real time, so that the SCADA system can control the gas equipment to perform protective actions based on the graded early warning signals.

[0047] In this embodiment of the invention, the device further includes an identification module 16, a determination module 17, and an update module 18. The identification module 16 is used to automatically identify newly added risk nodes based on online structure learning algorithms and running data; the determination module 17 is used to determine the parent node of the newly added risk node from the DBN model; the update module 18 is used to add the newly added risk node to the DBN model based on the parent node to obtain an updated DBN model.

[0048] In this embodiment of the invention, the online structure learning algorithm includes constraint testing, Bayesian online structure learning algorithm, incremental machine learning algorithm, or graph neural network algorithm.

[0049] In this embodiment of the invention, the inference module 14 includes a generation submodule and an estimation submodule. The generation submodule is used to generate a set of posterior samples based on the current observation data using the MCMC algorithm; the estimation submodule is used to calculate the failure probability of each sample at the target time point, average all failure probabilities to obtain a Bayesian estimate; or, it reads the failure state of each sample and uses the proportion of samples in the failure state to all posterior samples as the Bayesian estimate; the Bayesian estimate is the dynamic failure probability estimate.

[0050] In this embodiment of the invention, in the DBN model, an approximate inference algorithm is used for high-frequency nodes, and an exact inference algorithm is used for low-frequency key nodes to achieve dynamic inference.

[0051] In this embodiment of the invention, the device further includes a second acquisition module 19 and a second construction module 20. The second acquisition module 19 is used to acquire environmental data from an external database; the second construction module 20 is used to construct a knowledge graph model using gas equipment, environmental data, implicit risk factors, and protective actions as nodes, and the relationships between nodes as connecting edges.

[0052] In this embodiment of the invention, in the DBN model, the posterior probability of the previous sliding window is used as the prior probability of the current sliding window; or, the statistical parameters of the previous sliding window are used as the weak prior of the current sliding window.

[0053] The technical solution provided in this invention utilizes a dynamic Bayesian network model to deeply explore the correlations between nodes in the model, enabling real-time reasoning and updating of dynamic failure probabilities, thus improving the real-time performance and timeliness of dynamic failure probabilities. Furthermore, it provides early warnings through tiered warning signals, ensuring the safe operation of gas equipment.

[0054] This invention provides a protection system for gas equipment, comprising gas equipment, a SCADA system, and a server. The SCADA system monitors and controls the gas equipment, collecting its operational data. The server acquires operational data from the SCADA database in real time. A DBN model is constructed using the operational data as leaf nodes, implicit risk factors as intermediate nodes, and dynamic failure probability as the root node. Based on a sliding window mechanism, currently observed data is selected from the operational data. In the DBN model, dynamic inference is performed using the currently observed data to output the dynamic failure probability. Based on the dynamic failure probability, a graded early warning signal is generated and transmitted back to the SCADA system in real time, allowing the SCADA system to control the gas equipment to perform protective actions in conjunction with the graded early warning signal.

[0055] In this embodiment of the invention, the server includes a local server and a cloud server. The local server is used to handle real-time tasks, and the cloud server is used to handle non-real-time tasks. For example, non-real-time tasks include historical data archiving and cleaning, model retraining, statistical analysis, and strategy optimization analysis. Historical data archiving and cleaning refers to uploading historical data accumulated by the SCADA system and the local server to the cloud server, where the cloud server performs data cleaning, such as outlier detection, missing value completion, and normalization. Model retraining refers to periodically retraining the DBN model to improve its performance and adaptability. Statistical analysis refers to periodically performing statistical analysis and visualization of the long-term operating status and failure probability of gas equipment. Examples include trend analysis of corrosion risk changes with the seasons and pressure distribution map analysis. Strategy optimization analysis refers to using cross-site or multi-equipment data aggregated by the cloud server to provide strategy optimization suggestions to the local server, such as adjusting warning thresholds or updating the Bayesian network structure.

[0056] In this embodiment of the invention, a Database Management System (DBMS) is deployed on the server to achieve structured data storage. For example, the DBMS stores a SCADA raw data table, a risk factor probability table, a dynamic failure probability table, and a device and threshold configuration table. The SCADA raw data table stores raw SCADA data, including a primary key, device number, sensor number, parameter name (pressure, temperature, flow rate, etc.), collected values, units (MPa, ℃, m³ / h, etc.), and a collection timestamp. The risk factor probability table stores a primary key, device number, implicit risk factors, the probability of occurrence of the implicit risk factors, the sliding window number, and a calculation timestamp. The dynamic failure probability table stores a primary key, device number, dynamic failure probability, graded warning signals (extremely high / high / medium / low), protection actions, and a calculation timestamp. The device and threshold configuration table stores a device number, device type (valve, pipe section, storage tank, etc.), the dynamic failure probability range corresponding to each graded warning signal, protection actions, and configuration update time. Different gas devices can be configured with different warning strategies and protection action strategies, supporting flexible configuration.

[0057] The technical solution provided in this invention utilizes a dynamic Bayesian network model to deeply explore the correlations between nodes in the model, enabling real-time reasoning and updating of dynamic failure probabilities, thus improving the real-time performance and timeliness of dynamic failure probabilities. Furthermore, it provides early warnings through tiered warning signals, ensuring the safe operation of gas equipment.

[0058] This invention provides a computer-readable storage medium including a stored program, wherein, when the program runs, it controls the device where the computer-readable storage medium is located to execute the steps of the embodiments of the gas equipment protection method described above. For a detailed description, please refer to the embodiments of the gas equipment protection method described above.

[0059] Figure 3 A schematic diagram of a server provided in an embodiment of the present invention, as shown below. Figure 3 As shown, the server includes a processor 301, a memory 302, and a computer program 303 stored in the memory 302 and executable on the processor 301. When the processor 301 executes the computer program 303, it implements the steps in the various method embodiments described above. Alternatively, when the processor 301 executes the computer program 303, it implements the functions of each module / unit in the various device embodiments described above.

[0060] Server 3 can be an electronic device such as a desktop computer, laptop, handheld computer, or cloud server. Server 3 may include, but is not limited to, processor 301 and memory 302. Those skilled in the art will understand that... Figure 3This is merely an example of server 3 and does not constitute a limitation on server 3. It may include more or fewer components than shown in the illustration, or different components.

[0061] The processor 301 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0062] The memory 302 can be an internal storage unit of the server 3, such as a hard drive or memory of the server 3. The memory 302 can also be an external storage device of the server 3, such as a plug-in hard drive, SmartMedia Card (SMC), Secure Digital (SD) card, or FlashCard. The memory 302 can also include both internal and external storage units of the server 3. The memory 302 is used to store computer programs and other programs and data required by electronic devices.

[0063] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments 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 as a software functional unit.

[0064] The above are merely preferred embodiments 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 method for protecting gas equipment, characterized in that, The method includes: Real-time acquisition of operational data from the SCADA database; The DBN model is constructed by using the running data as leaf nodes, the implicit risk factors as intermediate nodes, and the dynamic failure probability as the root node. Based on the sliding window mechanism, the current observation data is selected from the running data; In the DBN model, dynamic inference is performed based on the current observation data to output the dynamic failure probability; Based on the dynamic failure probability, a graded early warning signal is generated and transmitted back to the SCADA system in real time, so that the SCADA system can control the gas equipment to perform protective actions in conjunction with the graded early warning signal.

2. The method according to claim 1, characterized in that, Using the operational data as leaf nodes, implicit risk factors as intermediate nodes, and dynamic failure probability as the root node, after constructing the DBN model, the following is also included: Based on the online structural learning algorithm, newly added risk nodes are automatically identified according to the operational data; The parent node of the newly added risk node is determined from the DBN model; The newly added risk node is added to the DBN model based on the parent node to obtain the updated DBN model.

3. The method according to claim 2, characterized in that, The online structure learning algorithm includes constraint testing, Bayesian online structure learning algorithm, incremental machine learning algorithm, or graph neural network algorithm.

4. The method according to claim 1, characterized in that, In the DBN model, dynamic inference is performed based on the current observation data to output a dynamic failure probability, including: A set of posterior samples is generated based on the current observation data using the MCMC algorithm; Calculate the failure probability of each sample at the target time point, average all failure probabilities to obtain a Bayesian estimate; or, read the failure state of each sample and use the proportion of samples in the failure state to all posterior samples as the Bayesian estimate. The Bayesian estimation is the Bayesian estimation of the dynamic failure probability.

5. The method according to claim 1, characterized in that, In the DBN model, an approximate inference algorithm is used for high-frequency nodes, while an exact inference algorithm is used for low-frequency key nodes to achieve dynamic inference.

6. The method according to claim 1, characterized in that, The method further includes: Retrieve environmental data from an external database; A knowledge graph model is constructed using the gas equipment, environmental data, implicit risk factors, and protective actions as nodes, and the relationships between nodes as connecting edges.

7. The method according to claim 1, characterized in that, In the DBN model, the posterior probability of the previous sliding window is used as the prior probability of the current sliding window; or, the statistical parameters of the previous sliding window are used as the weak prior of the current sliding window.

8. A gas equipment protection device, characterized in that, The device includes: The first acquisition module is used to acquire running data from the SCADA database in real time; The first building module is used to construct the DBN model by using the running data as leaf nodes, the implicit risk factors as intermediate nodes, and the dynamic failure probability as the root node. The selection module is used to select the current observation data from the running data based on a sliding window mechanism; The inference module is used to perform dynamic inference in the DBN model by combining the current observation data and output the dynamic failure probability. The generation module is used to generate graded early warning signals based on the dynamic failure probability, and to transmit the graded early warning signals back to the SCADA system in real time, so that the SCADA system can control the gas equipment to perform protective actions based on the graded early warning signals.

9. A server, characterized in that, The server includes one or more processors; a memory; and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions that, when executed by the server, cause the server to perform the gas equipment protection method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform the gas equipment protection method as described in any one of claims 1 to 7.

Images