A gas leakage risk grading method, device and terminal equipment

Abstract

This application provides a method, apparatus, and terminal equipment for classifying gas leakage risk levels, applicable to the field of data processing technology. The method includes: obtaining gas concentration prediction information based on meteorological information of the gas leakage environment, gas leakage source location information, and a gas concentration prediction model; performing correction and fusion processing on the gas concentration prediction information and gas concentration monitoring information to obtain corrected gas concentration information; and generating gas leakage risk level classification information based on the corrected gas concentration information and a preset gas leakage risk level classification threshold. This application improves the real-time performance and physical consistency of gas leakage concentration prediction calculations, accurately generates a dynamic risk level distribution of gas leakage across the entire chemical industrial park, and provides a reliable decision-making basis for emergency evacuation and safety management of gas leakage in chemical industrial parks.

Description

Technical Field

[0001] This application belongs to the field of data processing technology, and in particular relates to methods, devices and terminal equipment for classifying gas leakage risk levels. Background Technology

[0002] Toluene, a commonly used organic solvent in chemical industrial parks, is highly volatile, flammable, and toxic. Leaks can easily form toxic gas clouds, posing a serious threat to the lives of personnel and the ecological environment within the park. Chemical industrial parks are characterized by dense building clusters and complex terrain. The diffusion of toluene is influenced by a combination of factors, including wind speed and direction, temperature and humidity, surface roughness, building flow around structures, canyon effects, and wake recirculation, exhibiting strong nonlinear and spatiotemporally dynamic evolution characteristics.

[0003] Existing technologies typically employ single CFD numerical simulations or pure data-driven deep learning models to predict toluene leakage and diffusion. However, single CFD numerical simulations are extremely time-consuming and cannot meet the millisecond-level real-time response requirements for chemical leak emergencies. Summary of the Invention

[0004] In view of this, embodiments of this application provide a method, apparatus and terminal equipment for classifying gas leakage risk levels, aiming to solve the problems of low computational efficiency and difficulty in adapting to emergency real-time requirements in the prior art.

[0005] The first aspect of this application provides a method for classifying gas leakage risk levels, including:

[0006] Acquire multiple environmental meteorological information for gas leaks, multiple location information for gas leak sources, and multiple gas concentration monitoring information;

[0007] Based on the multiple gas leakage environmental meteorological information, multiple gas leakage source location information, and gas concentration prediction model, multiple gas concentration prediction information is obtained;

[0008] Based on the multiple gas concentration prediction information and multiple gas concentration monitoring information, correction and fusion processing are performed to obtain multiple gas concentration correction information;

[0009] Based on the multiple gas concentration correction information and multiple preset gas leakage risk level classification thresholds, multiple gas leakage risk level classification information is generated.

[0010] A second aspect of this application provides a gas leak risk level classification device, comprising:

[0011] The information acquisition module is used to acquire multiple gas leak environmental meteorological information, multiple gas leak source location information, and multiple gas concentration monitoring information;

[0012] The gas concentration prediction information generation module is used to obtain multiple gas concentration prediction information based on the multiple gas leakage environmental meteorological information, multiple gas leakage source location information and gas concentration prediction model;

[0013] The gas concentration correction information generation module is used to perform correction and fusion processing based on the multiple gas concentration prediction information and multiple gas concentration monitoring information to obtain multiple gas concentration correction information.

[0014] The gas leakage risk level classification information generation module is used to generate multiple gas leakage risk level classification information based on the multiple gas concentration correction information and multiple preset gas leakage risk level classification thresholds.

[0015] A third aspect of this application provides a terminal device, which includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the steps of the gas leakage risk level classification method described in the first aspect above.

[0016] A fourth aspect of this application provides a computer-readable storage medium, comprising: storing a computer program, wherein when executed by a processor, the computer program implements the steps of the gas leakage risk level classification method described in the first aspect above.

[0017] The beneficial effects of this application embodiment compared with the prior art are: this application improves the real-time calculation and physical consistency of gas leak concentration prediction, while maintaining good robustness in scenarios with abnormal sensor data and complex and changeable operating conditions, and is used to accurately generate the dynamic risk level distribution of gas leaks in the entire chemical industrial park, providing a reliable decision-making basis for emergency evacuation and safety management of gas leaks in chemical industrial parks. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram illustrating the implementation process of the gas leakage risk level classification method provided in Embodiment 1 of this application;

[0020] Figure 2 This is a schematic diagram illustrating the implementation process of the gas leakage risk level classification method provided in Embodiment 2 of this application;

[0021] Figure 3This is a schematic diagram illustrating the implementation process of the gas leakage risk level classification method provided in Embodiment 3 of this application;

[0022] Figure 4 This is a schematic diagram illustrating the implementation process of the gas leakage risk level classification method provided in Embodiment 4 of this application;

[0023] Figure 5 This is a schematic diagram illustrating the implementation process of the gas leakage risk level classification method provided in Embodiment 5 of this application;

[0024] Figure 6 This is a schematic diagram illustrating the implementation process of the gas leakage risk level classification method provided in Embodiment Six of this application;

[0025] Figure 7 This is a schematic diagram of the implementation process of the gas leakage risk level classification method provided in Embodiment 7 of this application;

[0026] Figure 8 This is a schematic diagram of the gas leakage risk level classification device provided in the embodiments of this application;

[0027] Figure 9 This is a schematic diagram of the terminal device provided in the embodiments of this application. Detailed Implementation

[0028] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0029] To illustrate the technical solution described in this application, specific embodiments are provided below.

[0030] Figure 1 A flowchart illustrating the implementation of the gas leakage risk level classification method provided in Embodiment 1 of this application is shown, and is described in detail below:

[0031] Step S101: Obtain multiple gas leak environmental meteorological information, multiple gas leak source location information, and multiple gas concentration monitoring information.

[0032] In this embodiment, the meteorological information of the gas leak environment may include wind speed information, wind direction information, humidity information and temperature information; the gas leak source location information may include three-dimensional spatial coordinate information and leak intensity information within the chemical industrial park; and the gas concentration monitoring information may be toluene concentration data collected in real time by sensors discretely deployed within the chemical industrial park.

[0033] Step S102: Based on the multiple gas leakage environmental meteorological information, multiple gas leakage source location information, and gas concentration prediction model, multiple gas concentration prediction information is obtained.

[0034] In this embodiment, the gas concentration prediction model can be pre-set and can be a Physical Information Neural Network (PINN). Time coordinates, three-dimensional spatial coordinates, and wind speed and direction from meteorological information related to the gas leak environment can be used as input information for the PINN. Based on the fluid dynamics physical constraints and residual network structure built into the PINN, multiple gas concentration prediction information across the entire spatiotemporal dimensions of the chemical industrial park can be calculated and generated.

[0035] In this embodiment, the gas concentration prediction model can employ an 8-layer fully connected residual network structure. The input layer receives a 5-dimensional feature vector, including time coordinates, three-dimensional spatial coordinates, and two key meteorological parameters: wind speed and wind direction, covering the core spatiotemporal and environmental factors affecting toluene diffusion. The hidden layers alleviate the gradient degradation problem in deep network training through a residual connection mechanism. Let the 1st layer... Layer input is The traditional fully connected layer is transformed into The output in the residual block is corrected to This skip connection allows gradients to flow directly to shallower networks during backpropagation, effectively improving the model's ability to represent steep gradient regions in the concentration field and accurately capturing complex features such as heavy gas deposition and vortex diffusion. The output layer is a single-output node, directly outputting the predicted toluene mass concentration at the corresponding spatiotemporal point without additional post-processing steps, ensuring millisecond-level inference speed. This application uses the Swish self-gated activation function, mathematically expressed as follows: ,in For the Sigmoid function, It is a learnable smoothing parameter. Its derivative properties... This ensures that the gradient remains non-zero in the negative interval, preventing neuron failure and enhancing the ability to capture small hydrodynamic perturbations.

[0036] Step S103: Correct and fuse the multiple gas concentration prediction information and multiple gas concentration monitoring information to obtain multiple gas concentration correction information.

[0037] In this embodiment, an ensemble Kalman filter algorithm can be used to correct and fuse multiple gas concentration prediction information and multiple gas concentration monitoring information. First, the gas concentration monitoring information can be preprocessed by sequentially performing outlier removal, sliding window smoothing filtering, multiple interpolation missing value repair, and inverse distance weighted interpolation. Then, statistical optimal estimation is completed through ensemble initialization, prediction step, and analysis step to correct the deviation of the gas concentration prediction information, thereby generating multiple gas concentration correction information that conforms to the real gas diffusion law.

[0038] Step S104: Generate multiple gas leakage risk level classification information based on the multiple gas concentration correction information and multiple preset gas leakage risk level classification thresholds.

[0039] In this embodiment, the preset gas leakage risk level classification threshold can be preset manually and used to perform risk quantification and conversion of multiple gas concentration correction information according to the classification mapping rules, thereby generating multiple gas leakage risk level classification information corresponding to the spatiotemporal location.

[0040] In this embodiment, to transform the time-varying toluene concentration distribution into a quantitative indicator that can directly guide evacuation decisions, the concept of a dynamic risk field can be introduced, and the dynamic risk field can be formally defined. For any point in Euclidean space At any moment The comprehensive risk measurement, as we understand it, is not a direct mapping of concentration, but rather a reflection of the probability distribution of irreversible harm or even death in the human body under specific exposure conditions, combined with toxicological effect models. Specifically, the quantification process relies on the classic Probit model to construct a nonlinear mapping from the concentration field to the risk probability. This model assumes that the organism's tolerance to toxins follows a log-normal distribution, which can well describe the gradual process from no reaction at low concentrations to lethality at high concentrations. For toluene exposure, the probability unit... The formula for calculation is:

[0041]

[0042] in, For a moment The toluene concentration (in ppm), with the integral term representing the cumulative exposure dose. , , These are the characteristic constants for toluene toxicity.

[0043] Among them, probability unit Compared with actual lethal probability The transformation follows a standard normal distribution:

[0044]

[0045] in, This is the cumulative function of the standard normal distribution. This formula can be used to convert probability units into intuitive lethal probabilities.

[0046] To meet the efficiency requirements of real-time evacuation decision-making while taking into account the risk of acute poisoning from instantaneous high concentrations, a threshold-based risk mapping function is constructed. :

[0047]

[0048] The indicator is clearly defined: Occupational exposure limit (PC-STEL); Concentrations that immediately threaten life and health; This is a penalty coefficient used to reinforce risk warnings in high-concentration areas. This definition quantifies the degree of harm the physical field poses to the human body and provides a clear potential energy function and obstacle avoidance constraints for subsequent evacuation route planning.

[0049] The gas leak risk level classification method provided in this application improves the real-time performance and physical consistency of gas leak concentration prediction calculations, while maintaining good robustness in scenarios with abnormal sensor data and complex and variable operating conditions. It is used to accurately generate the dynamic risk level distribution of gas leaks across the entire chemical industrial park, providing a reliable decision-making basis for emergency evacuation and safety management of gas leaks in chemical industrial parks.

[0050] Figure 2 The flowchart illustrating the implementation of the gas leakage risk level classification method provided in Embodiment 2 of this application is shown. The difference between this method and Embodiment 1 is that the method for generating the gas concentration prediction model specifically includes:

[0051] Step S201: Obtain the gas concentration prediction model to be trained.

[0052] In this embodiment, the gas concentration prediction model to be trained can be pre-set by humans, and can be a physical information neural network (PINN). Specifically, it can be a physical information neural network with an 8-layer fully connected residual network structure, configured with Swish self-gated activation function and batch normalization layer to provide basic network architecture support for subsequent model training.

[0053] Step S202: Based on the preset gas diffusion flow representation equation, the preset gas diffusion turbulent kinetic energy calculation equation, the preset gas diffusion turbulent kinetic energy dissipation rate calculation equation, and the preset gas diffusion concentration calculation model, multiple gas diffusion concentration information is calculated.

[0054] In this embodiment, the preset gas diffusion flow representation equation information can be preset by humans, the preset gas diffusion turbulent kinetic energy calculation equation can be preset by humans, the preset gas diffusion turbulent kinetic energy dissipation rate calculation equation can be preset by humans, and the preset gas diffusion concentration calculation model can be preset by humans. These can correspond to the Reynolds-averaged Navier-Stokes governing equations, the standard turbulent kinetic energy transport equation, the standard turbulent kinetic energy dissipation rate transport equation, and the component transport equation in sequence. The above equations can be solved based on the modified urban canopy turbulence model, thereby calculating and generating multiple gas diffusion concentration information that conforms to the physical laws of heavy gas diffusion in chemical industrial parks.

[0055] In this embodiment, it is understood that the gas in this application can be toluene. The diffusion of toluene after leakage is essentially a multi-component, non-isothermal turbulent flow and mass transfer process. The preset gas diffusion flow equations can be represented by a set of equations, which can be based on the Reynolds-averaged Navier-Stokes (RANS) equation framework and adopt standard... The turbulence model is closed to effectively capture the influence of turbulent fluctuations on gas diffusion. Specifically, the first step is to use the mass conservation equation for the mixed gas, i.e., the continuity equation, to describe the mass flow equilibrium of the fluid in space:

[0056]

[0057] in, Represents the density of the gas mixture. It is a velocity vector. Indicates time. In studies of heavy gas diffusion, density... It is a variable that varies with time and space, and is affected by both component concentration and temperature. This is fundamentally different from the assumption of approximately constant density commonly used in the diffusion of neutral gases.

[0058] Secondly, there is the momentum conservation equation, which describes the relationship between the forces acting on a fluid element and changes in its state of motion. In heavy gas diffusion, the gravity term plays a crucial role and is the core physical basis of the heavy gas effect.

[0059]

[0060] in, For static pressure, The effective viscosity is the sum of the laminar viscosity and the turbulent viscosity. It is the gravitational acceleration vector. The momentum source term. The gravity term on the right-hand side of the equation. This causes high-density toluene vapor clouds to exhibit significant negative buoyancy, leading to a tendency for the gas to diffuse close to the ground and resist atmospheric turbulence and entrainment. This is the most crucial characteristic that distinguishes heavy gas diffusion from neutral gas diffusion.

[0061] Then there is the energy conservation equation, which describes the thermodynamic processes within the system and is suitable for leakage scenarios involving phase changes or large temperature differences:

[0062]

[0063] in, The static enthalpy of the fluid. For turbulent Prandtl number, This is the energy source term. The purpose of this equation is to capture the indirect effects of temperature changes on the diffusion process, thereby improving the physical realism of the model.

[0064] Finally, there is the component transport equation, which describes the convective transport and diffusion of toluene components in the flow field:

[0065]

[0066] in This represents the mass fraction of toluene. The laminar diffusion coefficient is... For turbulent viscosity, For turbulent Schmidt number, This is the leakage source term. This equation fully describes the convective transport and turbulent diffusion effects that toluene experiences simultaneously in the flow field, and is the core equation for subsequent calculations of the concentration field distribution.

[0067] In this embodiment, the preset gas diffusion turbulent kinetic energy calculation equation and the preset gas diffusion turbulent kinetic energy dissipation rate calculation equation can be turbulent kinetic energy... and turbulent kinetic energy dissipation rate The transport equations can be derived by modifying turbulence models for urban canopies. It is understandable that flows within the atmospheric boundary layer are typically high Reynolds number turbulent flows, and in environments such as chemical industrial parks, they are subject to strong disturbances from dense building clusters, resulting in significant three-dimensional unsteady flow field characteristics. While Direct Numerical Simulation (DNS) and Large Eddy Simulation (LES) offer high accuracy in current mainstream turbulence simulation methods, their computational costs are extremely high, making it difficult to meet the requirements for constructing large-scale training datasets. The RANS method, by applying Reynolds averaging to the Navier-Stokes equations, solves for the time-averaged flow field and uses a turbulence model to close the Reynolds stress term, achieving higher computational efficiency while maintaining engineering accuracy. Therefore, this application adopts the standard... Two-equation models, due to their robustness and computational economy, have accumulated rich validation data in high Reynolds number flow simulations. To accurately describe the transport process of toluene under complex geometric boundaries, turbulent kinetic energy can be introduced. and turbulent kinetic energy dissipation rate Use the transport equations to close the system of equations:

[0068] Turbulent kinetic energy The equation is:

[0069]

[0070] Turbulent kinetic energy dissipation rate The equation is:

[0071]

[0072] in, Indicates turbulent viscosity. The turbulent kinetic energy term is generated by the average velocity gradient. This is the turbulent kinetic energy term generated by buoyancy. All are model constants.

[0073] Understandably, standards are required for urban canopy environments. The model suffers from overestimating the turbulent kinetic energy on the windward side of the building, leading to inaccurate predictions of the stagnation point pressure. Therefore, this application corrects the model constants, specifically adjusting... and The values ​​of [value] were determined, and source term corrections considering streamline bending and rotation effects were introduced to enhance the ability to capture flow around buildings and separated flows. Validation employed a combination of quantitative and qualitative methods. The coefficient of determination (R²) was used as the quantitative indicator, while qualitative analysis was conducted through visual comparison of concentration distribution. Validation results showed that R² was only around 0.7 under low wind speed conditions.

[0074] In this embodiment, the toluene mass fraction transport equation was also modified by adding a molecular diffusion coefficient term. This coefficient, based on the ideal gas equation of state and empirical formulas for diffusion coefficients, can be calculated using a Python script. Verification showed that the modified CFD model significantly improved performance across all test scenarios. All are greater than 0.85, with 80% of the scenarios being [missing information]. Greater than 0.9. Specifically, the performance comparison table before and after CFD model correction is as follows:

[0075]

[0076] The table above shows a performance comparison of the CFD model before and after correction. The data shows that the RMSE of the corrected CFD model under low wind speed conditions decreased from 0.15 to 0.07, an improvement of 53.3%. The index improved from 0.70 to 0.91. Under medium and high wind speeds, all indicators also showed significant improvement. On average, RMSE decreased by 36.4%. The improvement of 0.12 indicates that the corrected CFD model can maintain high prediction accuracy under various wind speed conditions.

[0077] To further verify the reliability of the model in real-world scenarios, this application conducted on-site monitoring experiments in three test scenarios during the experimental verification process, obtaining measured concentration data from 30 detection points, and comparing the data with the prediction results of the CFD model. The performance comparison table between the CFD model and the measured data is as follows:

[0078]

[0079] As shown in the table above, in scenarios 2, 5, and 8, the RMSE between predicted and measured values ​​was kept at a low level. The p-values ​​reached 0.88, 0.87, and 0.90 respectively, with all p-values ​​less than 0.05, indicating a significant consistency between the two models. The results fully demonstrate that the modified model has high prediction accuracy and can provide reliable support for subsequent ground truth generation.

[0080] In this embodiment, the proper setting of the computational domain is a prerequisite for ensuring the correct application of boundary conditions and the convergence of results in the numerical simulation. During the verification process, a three-dimensional computational domain with dimensions of 1000m × 800m × 100m (length × width × height) can be constructed using a typical chemical industrial park as a prototype. This scale adheres to the principle of fully developing the flow field, ensuring that the wake region around the building complex can be completely resolved, while avoiding non-physical interference from boundary backflow on the core area of ​​interest. The computational domain covers the leakage source, the densely populated equipment area, and the surrounding open area, and can be equipped with 30 discrete monitoring points for data acquisition during subsequent model verification.

[0081] Step S203: Based on the multiple gas diffusion concentration information, train the gas concentration prediction model to be trained to obtain the gas concentration prediction model.

[0082] In this embodiment, a domain-adaptive transfer learning training method can be adopted. First, the pre-training of the gas concentration prediction model to be trained is completed in a simple diffusion scenario. Then, most of the underlying network parameters are frozen and the parameters are fine-tuned in a complex scenario in a chemical industrial park. Then, the Adam and L-BFGS hybrid optimizer is combined to complete iterative convergence, thereby training and generating a gas concentration prediction model with spatiotemporal prediction capabilities and physical conservation properties.

[0083] The gas leakage risk level classification method provided in this application significantly reduces the number of training iterations of the gas concentration prediction model, improves the convergence efficiency of the gas concentration prediction model, and enhances the fitting ability of the gas concentration prediction model to complex diffusion characteristics such as flow around building clusters in chemical industrial parks and heavy gas settling, thereby ensuring the accuracy and generalization ability of subsequent gas concentration prediction results.

[0084] Figure 3 The flowchart illustrating the implementation of the gas leakage risk level classification method provided in Embodiment 3 of this application is shown. The difference between this method and Embodiment 2 is that step S203 specifically includes:

[0085] Step S301: Based on the multiple gas diffusion concentration information and the gas concentration prediction model to be trained, calculate the gas diffusion concentration prediction composite loss information.

[0086] In this embodiment, the partial derivatives of the fluid dynamics equations can be solved by the automatic differentiation function of the gas concentration prediction model to be trained. Then, multiple basic loss components can be calculated by combining multiple gas diffusion concentration information. Finally, the multiple basic loss components are weighted and summed to obtain the gas diffusion concentration prediction composite loss information used to characterize the model fitting accuracy and physical compliance.

[0087] Step S302: Based on the multiple gas diffusion concentration information, the gas diffusion concentration prediction composite loss information, and the preset gas diffusion concentration prediction composite loss threshold, the gas concentration prediction model to be trained is trained to obtain the gas concentration prediction model.

[0088] In this embodiment, the preset gas diffusion concentration prediction composite loss threshold can be preset manually and used to continuously iterate and update the weights and bias parameters of the gas concentration prediction model to be trained. Then, training stops when the gas diffusion concentration prediction composite loss information converges to the range of the preset gas diffusion concentration prediction composite loss threshold, thereby generating a gas concentration prediction model that meets the accuracy and physical constraints requirements.

[0089] The gas leakage risk level classification method provided in this application realizes multi-dimensional constraint training of the model by constructing composite loss information for gas diffusion concentration prediction. This effectively avoids overfitting and local optima problems in the gas concentration prediction model to be trained, thereby improving the prediction stability and reliability of the gas concentration prediction model under unknown leakage conditions.

[0090] Figure 4 The flowchart illustrating the implementation of the gas leakage risk level classification method provided in Embodiment 4 of this application is shown. The difference between this method and Embodiment 3 above is that step S301 specifically includes:

[0091] Step S401: Based on the multiple gas diffusion concentration information and the gas concentration prediction model to be trained, calculate the gas diffusion concentration fitting loss information, the gas diffusion equation residual loss information, and the gas diffusion boundary condition loss information.

[0092] In this embodiment, multiple gas diffusion concentration information and the output value of the gas concentration prediction model to be trained can be used to perform error calculation to generate gas diffusion concentration fitting loss information. Then, based on the fluid dynamics control equation, the residual loss information of the gas diffusion equation is calculated by solving the residual. Finally, the gas diffusion boundary condition loss information is calculated by combining the no-slip and zero-flux constraints of the building wall.

[0093] Step S402: Based on the preset weighted coefficients for the fitting loss of gas diffusion concentration, the preset weighted coefficients for the residual loss of the gas diffusion equation, and the preset weighted coefficients for the loss of gas diffusion boundary conditions, the gas diffusion concentration fitting loss information, the gas diffusion equation residual loss information, and the gas diffusion boundary condition loss information are weighted and summed to calculate the gas diffusion concentration prediction composite loss information.

[0094] In this embodiment, the preset weighting coefficient for the gas diffusion concentration fitting loss can be preset manually, and can be set to 0.4. The preset weighting coefficient for the gas diffusion equation residual loss can be preset manually, and can be set to 0.5. The preset weighting coefficient for the gas diffusion boundary condition loss can be preset manually, and can be set to 0.1. Then, the weight proportions of the gas diffusion concentration fitting loss information, the gas diffusion equation residual loss information, and the gas diffusion boundary condition loss information are matched according to the corresponding weighting coefficients. Then, a weighted summation operation is performed to calculate the gas diffusion concentration prediction composite loss information that takes into account data fitting, physical constraints, and boundary compliance.

[0095] In this embodiment, the core difference between PINN and traditional deep learning lies in the design of its composite loss function. It not only focuses on data fitting error but also forces the network output to conform to fundamental laws of fluid mechanics through physical constraints, constructing a triple constraint system of "data fitting - physical consistency - boundary satisfaction." (Gas diffusion concentration prediction composite loss information) Loss information fitted from gas diffusion concentration Gas diffusion equation residual loss information Information on loss due to gas diffusion boundary conditions Weighted composition, the expression is:

[0096]

[0097] in For all network weights and bias parameters, , , The weighting coefficients are the preset weighting coefficients for the gas diffusion concentration fitting loss, the preset weighting coefficients for the gas diffusion equation residual loss, and the preset weighting coefficients for the gas diffusion boundary condition loss. These are used to balance the contributions of different constraints to the gradient. The optimal values ​​can be determined through orthogonal experiments: 0.4, 0.5, and 0.1, respectively.

[0098] Gas diffusion concentration fitting loss information The mean squared error between the model's predicted values ​​and the true CFD values ​​is used to ensure that the network accurately reproduces the concentration distribution at known sample points. The formula is:

[0099]

[0100] in The number of training samples, To predict the concentration of PINN, This is the true value of the CFD, and this loss guarantees the basic fitting ability of the model.

[0101] Gas diffusion equation residual loss information This is the core of PINN. By minimizing the residuals of the Navier-Stokes equations and the component transport equations, it forces the network output to conform to physical laws such as mass conservation and momentum conservation. For the component transport equations, the residuals... Defined as:

[0102]

[0103] Where the time derivative and spatial derivative Accurate calculations using automatic differentiation techniques avoid truncation errors in numerical differences. The formula for the gas diffusion boundary condition loss information is:

[0104]

[0105] in To calculate the number of configuration points for random sampling within the domain, To correspond to the residuals of the momentum and pressure equations, this loss ensures that the model still follows physical laws in the data-free region.

[0106] Boundary condition loss The no-slip wall condition and zero-flux condition are transformed into soft constraints, and the penalty function method is used to incorporate them into the loss function.

[0107]

[0108] in The number of boundary sampling points. The wall normal vector ensures that toluene gas does not penetrate the building surface, and its velocity is zero at the wall. By jointly optimizing the triple loss, the PINN model finds the optimal balance between data-driven and physical mechanisms, ensuring both prediction accuracy and strong generalization ability, while avoiding overfitting and physical logic fallacies inherent in pure data models.

[0109] The gas leakage risk level classification method provided in this application accurately balances the training ratio of data fitting accuracy, fluid physics equation constraints, and spatial boundary constraints. This ensures that the trained model can accurately reproduce concentration distribution data and strictly follow the law of conservation of mass, momentum, and energy, effectively preventing the occurrence of gas concentration prediction results that are not based on physical laws.

[0110] Figure 5The flowchart illustrating the implementation of the gas leakage risk level classification method provided in Embodiment 5 of this application is shown. The difference between this method and Embodiment 1 above is that step S103 specifically includes:

[0111] Step S501: Supplement and correct the multiple gas concentration monitoring information to obtain multiple corrected gas concentration monitoring information.

[0112] In this embodiment, the three-standard-deviation criterion can be used to detect and remove outliers from multiple gas concentration monitoring information. Then, multiple interpolation methods are used to repair missing data caused by communication interruptions and equipment failures. High-frequency random noise is then filtered out by sliding window averaging filtering. Finally, the inverse distance weighted interpolation method is used to complete the spatial filling of sparse monitoring points, thereby generating multiple regular and reliable corrected gas concentration monitoring information.

[0113] In this embodiment, it is understood that data collected by field sensors is often affected by electromagnetic interference, equipment drift, and communication packet loss, resulting in problems such as outliers, missing values, high-frequency noise, and sparse spatiotemporal distribution. Therefore, data quality must be improved through a system preprocessing procedure. For outliers in the data, a 3- Criteria are used for detection and rejection. For time-series observations at a specific monitoring point... Calculate the mean with standard deviation If the data satisfies If a value is found to be outlier, it is marked as invalid. This criterion, based on the characteristics of a normal distribution, can effectively eliminate extreme random errors with a probability less than 0.3%, thus avoiding interference from outlier data with the assimilation results.

[0114] To eliminate high-frequency noise, a sliding window averaging filter algorithm is used for smoothing. The window size is defined as... Smoothed observations for:

[0115]

[0116] This operation is equivalent to a low-pass filter, preserving the concentration change trend while filtering out transient fluctuations, thus improving data stability. For missing values ​​caused by communication interruptions or equipment malfunctions, multiple interpolation (MICE) is used for repair. This method utilizes the correlation between spatially adjacent sensors at the same time point and the autocorrelation of the time series, iteratively predicting missing values ​​through a chain equation to ensure the spatiotemporal integrity of the data.

[0117] Considering the sparse spatial distribution of sensors, inverse distance weighted interpolation (IDW) is used to construct the initial observation background field. For non-observation points... Its interpolation concentration From the neighborhood The weighted average of the sensor values ​​is obtained as follows:

[0118]

[0119] in, For point To the sensor Euclidean distance, power exponent This ensures that closer sensors contribute a higher weight.

[0120] Step S502: The multiple gas concentration prediction information and the multiple corrected gas concentration monitoring information are fused together to obtain multiple gas concentration fusion information.

[0121] In this embodiment, the spatial mapping relationship between multiple gas concentration prediction information and multiple corrected gas concentration monitoring information can be established by relying on the statistical estimation characteristics of ensemble Kalman filtering. Then, the inherent bias of the model prediction can be corrected by using the measured data of the monitoring points. Finally, the observation information can be diffused to the blank area where no sensors are deployed by combining the spatial correlation characteristics of the flow field, thereby generating multiple gas concentration fusion information that takes into account both the model inference law and the field measured data.

[0122] In this embodiment, it can be understood that EnKF is based on Monte Carlo random sampling theory. By approximating the statistical characteristics of the probability density function through the evolution of a set of states, it avoids the storage and calculation of high-dimensional covariance matrices in traditional Kalman filtering and adapts to the high-dimensional characteristics of dynamic risk fields. Specifically, it can be divided into three stages: set initialization, prediction step, and analysis step, to achieve optimal fusion of model priors and observation data.

[0123] The first phase is set initialization. At the initial moment... Initial prediction field based on PINN model Generate by superimposing random perturbations containing The initial state set of each member The disturbance follows a pattern with a mean of 0 and a covariance matrix of . Normal distribution:

[0124]

[0125] Set size The initial error covariance matrix is ​​set to 50 to balance statistical convergence and computational efficiency. Determination of verification error based on the PINN model.

[0126] The second stage is the prediction step. It utilizes the state evolution operator of the PINN model. By performing independent time integration on each member of the set, we obtain... Prior prediction set at time :

[0127]

[0128] in, This represents the members of the analysis set from the previous time step. Based on the prior set, the prior mean is statistically calculated. Approximate value of the prior error covariance matrix:

[0129]

[0130] in Let be the set perturbation matrix. As an all-one vector, this approximation avoids the direct storage of high-dimensional matrices.

[0131] The third stage is the analysis step. When acquiring... Observational data at time Then, the set of observed perturbations is first generated. ,in To maintain the statistical properties of the observation noise, the Kalman gain matrix is ​​then calculated. :

[0132]

[0133] in The observation matrix corresponds to the observation operator. The matrix form is used. The Kalman gain is used to update each prior set member, resulting in the analysis set. :

[0134]

[0135] By analyzing the mean of the set This is the optimal state estimate after assimilation, used for dynamic risk field reconstruction. The core advantage of EnKF lies in transmitting spatial correlation through ensemble perturbations. Even with sparse sensor deployment, it can utilize the spatial correlation characteristics of hydrodynamics to propagate observation information to unobserved areas, effectively correcting the prediction bias of the PINN model and improving the global reliability of the risk field.

[0136] Step S503: Based on the preset gas concentration filtering rules, the multiple gas concentration fusion information is subjected to gain correction processing to obtain multiple gas concentration correction information.

[0137] In this embodiment, the preset gas concentration filtering rules can be manually preset or can be rules using Kalman filtering. Smoothing constraints can be set according to the spatiotemporal evolution characteristics of the flow field in the chemical industrial park, and then the fusion information of multiple gas concentrations can be smoothed in the whole domain and corrected in the temporal trend to eliminate local numerical mutations and unreasonable concentration jumps, thereby generating multiple gas concentration correction information that are spatiotemporally distributed and conform to the heavy gas diffusion law.

[0138] The gas leakage risk level classification method provided in this application realizes the deep integration of gas concentration prediction data and gas concentration actual monitoring data. Then, through global gain correction, it effectively reduces the interference caused by sensor measurement error and model inference error, thereby improving the spatial continuity and temporal stability of gas concentration correction information and adapting to the heavy gas concentration field reconstruction needs under the complex building complex of chemical industrial parks.

[0139] Figure 6 The flowchart illustrating the implementation of the gas leakage risk level classification method provided in Embodiment Six of this application is shown. The difference between this method and Embodiment Five is that step S502 specifically includes:

[0140] Step S601: Based on the preset gas concentration prediction perturbation distribution mean information and the preset gas concentration prediction perturbation distribution covariance information, multiple gas concentration prediction perturbation information are randomly generated.

[0141] In this embodiment, the preset gas concentration prediction perturbation distribution mean information can be preset by humans and set to 0. The preset gas concentration prediction perturbation distribution covariance information can also be preset by humans and set according to the verification error of the physical information neural network model. Then, multiple uncertain gas concentration prediction perturbation information can be randomly generated in the form of a normal distribution.

[0142] Step S602: Generate multiple perturbed gas concentration prediction information based on the multiple gas concentration prediction information and the multiple gas concentration prediction perturbation information.

[0143] In this embodiment, multiple gas concentration prediction perturbation information can be superimposed one by one onto the corresponding multiple gas concentration prediction information, thereby introducing a random representation of the model process error, and thus generating multiple perturbation gas concentration prediction information covering the range of uncertainty fluctuations.

[0144] Step S603: Based on the preset gas concentration prediction state evolution operator, time integration is performed according to the multiple perturbed gas concentration prediction information to obtain multiple prior gas concentration prediction information.

[0145] In this embodiment, the preset gas concentration prediction state evolution operator can be artificially preset, corresponding to the state prediction inference operator of the physical information neural network. Then, time dimension integration inference can be performed on each group of multiple perturbation gas concentration prediction information to generate multiple prior gas concentration prediction information updated at time.

[0146] Step S604: Generate multiple perturbed gas concentration monitoring information based on the multiple corrected gas concentration monitoring information and the multiple gas concentration prediction perturbation information.

[0147] In this embodiment, a perturbation component that conforms to the statistical characteristics of observation noise can be superimposed on multiple corrected gas concentration monitoring information to simulate the random error characteristics of real sensor measurements, thereby generating multiple perturbed gas concentration monitoring information that match the statistical distribution law.

[0148] Step S605: Based on the preset gas concentration gain update rule, the multiple prior gas concentration prediction information and the multiple perturbed gas concentration monitoring information are fused to obtain multiple gas concentration fusion information.

[0149] In this embodiment, the preset gas concentration gain update rule can be preset manually. It can rely on the Kalman gain matrix to complete the weighted update of prior information and post-perturbation monitoring information, and then use ensemble statistical features to correct the global concentration field deviation, thereby generating multiple gas concentration fusion information under the statistical optimal estimate.

[0150] In this embodiment, the real-time reconstruction of the dynamic risk field of toluene leakage faces the dual challenges of model prediction bias and environmental uncertainty. Neither a single numerical model nor sensor observations can independently provide a high-confidence risk distribution. Ensemble Kalman Filtering (EnKF) provides a reliable framework for accurate reconstruction of the dynamic risk field by fusing prior model predictions with real-time observation data and using statistically optimal estimation to correct state biases. Its core principle is to abstract the toluene diffusion process as a nonlinear stochastic dynamic system, quantifying the system evolution and observation process through state-space equations and observation models.

[0151] Define time system state vector ,in To calculate the total number of discrete grid points within the domain, the state vector contains the toluene concentration values ​​for all spatial points, directly corresponding to the core input parameters of the dynamic risk field. Based on the PINN fast prediction model constructed in Section 3.3, the state evolution equation is defined as:

[0152]

[0153] in, This is the prediction operator for the PINN model, with the previous time step as the input. External inputs such as meteorological parameters The output is the prior concentration field at the current moment; This represents process noise, characterizing the uncertainties introduced by the physical approximation, parameterization errors, and environmental disturbances in the PINN model. Let be the process noise covariance matrix.

[0154] The observation model is used to establish the mapping relationship between the system state and sensor observations. Time is defined. observation vector , The number of gas sensors deployed in the field is typically The observation equation is:

[0155]

[0156] in, For observation operators, a high-dimensional state space is mapped to a low-dimensional observation space. For sparsely deployed sensors, Dimensionality reduction is achieved by extracting the concentration values ​​of the corresponding grid points; The observation noise includes sensor measurement error, electromagnetic interference, and representativeness error. The noise covariance matrix was determined through sensor calibration experiments. Understandably, the construction of the state equations and observation models must ensure physical consistency, the evolution of the state vectors must strictly follow the fluid dynamics control equations, and the mapping relationship of the observation operators must correspond one-to-one with the actual deployment locations of the sensors. By introducing a noise term, the model can quantify various uncertainties, providing a statistical basis for subsequent assimilation updates. The core task of data assimilation is based on the observation sequence. Solving the state posterior probability density function To achieve the optimal estimation of the risk field.

[0157] The gas leakage risk level classification method provided in this application achieves optimal data fusion and maintains strong robustness under conditions of sparse sensor deployment and variable environmental disturbances, effectively mitigating the impact of model approximation errors and observation noise on the gas concentration fusion results.

[0158] Figure 7 The flowchart illustrating the implementation of the gas leakage risk level classification method provided in Embodiment Seven of this application is shown. The difference between this method and Embodiment One described above is that, after step S104, the method further includes:

[0159] Step S701: Perform spatial positioning processing based on the multiple gas leak risk level classification information to obtain multiple gas leak risk level positioning information.

[0160] In this embodiment, the coordinate matching and regional labeling of multiple gas leakage risk level classification information can be performed based on the spatial coordinate system of the three-dimensional computing domain of the chemical industrial park. Then, the specific spatial locations of high-risk hotspot areas, leeward retention areas of buildings, and street canyon diffusion zones can be located, thereby generating multiple gas leakage risk level location information with accurate geographic coordinates.

[0161] In this embodiment, gas concentration values ​​can be reconstructed into a continuous dynamic risk field through spatial interpolation, providing a continuous spatial potential energy function and obstacle avoidance constraints for evacuation decisions. Specifically, the reconstruction mechanism constructs a mapping Φ from discrete state vectors to a continuous risk field function R(x,y,z,t), fusing the assimilation increment and the PINN prediction field to ensure the physical consistency and spatiotemporal continuity of the reconstruction results. Specifically, this can be achieved by first calculating the assimilation increment based on the difference between the analysis set and the prior set. This increment reflects the correction amount of the observed data to the model prediction. Using the Kriging interpolation method, the discrete assimilation increment is extended into a continuous error correction field. Kriging interpolation, based on the covariance function of spatial points, can make full use of the spatial correlation of the concentration field, ensuring the smoothness and rationality of the correction field.

[0162] Next, the error correction field is superimposed on the original prediction field of PINN to obtain the reconstructed concentration field:

[0163]

[0164] Based on dynamic risk field mapping function By combining the Probit toxicology model, the reconstructed concentration field is transformed into a real-time risk field:

[0165]

[0166] Achieving a three-level transformation of "concentration reconstruction - toxicological effects - risk probability" ensures that the risk field can reflect real-time observation information and conform to the hazard patterns of human exposure.

[0167] Step S702: Generate multiple gas leak emergency evacuation decision information based on the multiple gas leak risk level location information and the preset gas leak emergency evacuation decision rules.

[0168] In this embodiment, the preset gas leak emergency evacuation decision rules can be preset by humans. Based on the toluene toxicological hazard threshold and risk level zoning results, personnel avoidance routes and park control areas can be formulated. Then, by combining multiple gas leak risk level location information, high-risk areas, medium-risk areas and low-risk areas can be divided, thereby generating multiple gas leak emergency evacuation decision information including evacuation routes, restricted areas and personnel placement guidelines.

[0169] The gas leakage risk level classification method provided in this application enables precise spatial positioning of risk areas and automatically generates emergency evacuation-related decision information, providing intuitive and practical decision support for real-time emergency response and personnel safety evacuation in toluene leakage accidents in chemical industrial parks.

[0170] Corresponding to the method in the above embodiments, Figure 8 A structural block diagram of the gas leakage risk level classification device provided in the embodiments of this application is shown. For ease of explanation, only the parts related to the embodiments of this application are shown. Figure 8 The gas leak risk level classification device in the example can be the implementing entity of the gas leak risk level classification method provided in the aforementioned embodiment 1.

[0171] Reference Figure 8 The gas leak risk level classification device includes:

[0172] The information acquisition module 810 is used to acquire multiple gas leak environmental meteorological information, multiple gas leak source location information, and multiple gas concentration monitoring information;

[0173] The gas concentration prediction information generation module 820 is used to obtain multiple gas concentration prediction information based on the multiple gas leakage environmental meteorological information, multiple gas leakage source location information and gas concentration prediction model;

[0174] The gas concentration correction information generation module 830 is used to perform correction and fusion processing based on the multiple gas concentration prediction information and multiple gas concentration monitoring information to obtain multiple gas concentration correction information.

[0175] The gas leakage risk level classification information generation module 840 is used to generate multiple gas leakage risk level classification information based on the multiple gas concentration correction information and multiple preset gas leakage risk level classification thresholds.

[0176] The process by which each module in the gas leak risk level classification device provided in this application performs its respective function can be found in the foregoing. Figure 1 The description of Embodiment 1 shown will not be repeated here.

[0177] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0178] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0179] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0180] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0181] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance. It should also be understood that although the terms "first," "second," etc., are used in the text to describe various elements in some embodiments of this application, these elements should not be limited by these terms. These terms are merely used to distinguish one element from another. For example, a first table may be named a second table, and similarly, a second table may be named a first table, without departing from the scope of the various described embodiments. Both the first table and the second table are tables, but they are not the same table.

[0182] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0183] The gas leakage risk level classification method provided in this application embodiment can be applied to terminal devices such as mobile phones, tablets, wearable devices, vehicle-mounted devices, augmented reality / virtual reality devices, laptops, super mobile personal computers, netbooks, and personal digital assistants. This application embodiment does not impose any restrictions on the specific type of terminal device.

[0184] For example, the terminal device may be a station in a WLAN, a cellular phone, a cordless phone, a session initiation protocol phone, a wireless local loop station, a personal digital processing device, a handheld device with wireless communication capabilities, a computing device or other processing device connected to a wireless modem, an in-vehicle device, a vehicle networking terminal, a computer, a laptop computer, a handheld communication device, a handheld computing device, a satellite wireless device, a wireless modem card, a set-top box, a user premises equipment, and / or other devices for communication over a wireless system, as well as next-generation communication systems, such as mobile terminals in 5G networks or mobile terminals in future evolved public terrestrial mobile networks, etc.

[0185] Figure 9 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. For example... Figure 9 As shown, the terminal device 9 of this embodiment includes: at least one processor 90 ( Figure 9 (Only one is shown in the image), a memory 91, which stores a computer program 92 that can run on the processor 90. When the processor 90 executes the computer program 92, it implements the steps in the various embodiments of the gas leak risk level classification methods described above, for example... Figure 1 Steps S101 to S104 are shown. Alternatively, when the processor 90 executes the computer program 92, it implements the functions of each module / unit in the above-described device embodiments, for example... Figure 8 The functions of modules 810 to 840 are shown.

[0186] The terminal device 9 can be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor 90 and a memory 91. Those skilled in the art will understand that... Figure 9 This is merely an example of terminal device 9 and does not constitute a limitation on terminal device 9. It may include more or fewer components than shown, or combine certain components, or different components. For example, the terminal device may also include input transmission devices, network access devices, buses, etc.

[0187] The processor 90 may be a central processing unit, or it may be other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0188] In some embodiments, the memory 91 may be an internal storage unit of the terminal device 9, such as a hard disk or memory of the terminal device 9. The memory 91 may also be an external storage device of the terminal device 9, such as a plug-in hard disk, smart memory card, secure digital card, flash memory card, etc., equipped on the terminal device 9. Furthermore, the memory 91 may include both internal and external storage units of the terminal device 9. The memory 91 is used to store operating systems, applications, bootloaders, data, and other programs, such as the program code of computer programs. The memory 91 can also be used to temporarily store data that has been sent or will be sent.

[0189] Furthermore, the functional units in the various embodiments of this application 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.

[0190] This application also provides a terminal device, which includes at least one memory, at least one processor, and a computer program stored in the at least one memory and executable on the at least one processor. When the processor executes the computer program, it causes the terminal device to implement the steps in any of the above method embodiments.

[0191] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.

[0192] This application provides a computer program product that, when run on a terminal device, enables the terminal device to implement the steps described in the various method embodiments above.

[0193] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory, a random access memory, an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0194] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0195] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0196] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0197] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for classifying gas leakage risk levels, characterized in that, include: Acquire multiple environmental meteorological information for gas leaks, multiple location information for gas leak sources, and multiple gas concentration monitoring information; Based on the multiple gas leakage environmental meteorological information, multiple gas leakage source location information, and gas concentration prediction model, multiple gas concentration prediction information is obtained; Based on the multiple gas concentration prediction information and multiple gas concentration monitoring information, correction and fusion processing are performed to obtain multiple gas concentration correction information; Based on the multiple gas concentration correction information and multiple preset gas leakage risk level classification thresholds, multiple gas leakage risk level classification information is generated.

2. The gas leakage risk level classification method as described in claim 1, characterized in that, The gas concentration prediction model is obtained through the following steps: Obtain the gas concentration prediction model to be trained; Based on the preset gas diffusion flow representation equation, the preset gas diffusion turbulent kinetic energy calculation equation, the preset gas diffusion turbulent kinetic energy dissipation rate calculation equation, and the preset gas diffusion concentration calculation model, multiple gas diffusion concentration information is calculated. Based on the multiple gas diffusion concentration information, the gas concentration prediction model to be trained is trained to obtain the gas concentration prediction model.

3. The gas leakage risk level classification method as described in claim 2, characterized in that, The step of training the gas concentration prediction model based on the multiple gas diffusion concentration information to obtain the gas concentration prediction model specifically includes: Based on the multiple gas diffusion concentration information and the gas concentration prediction model to be trained, the composite loss information for gas diffusion concentration prediction is calculated. Based on the multiple gas diffusion concentration information, the gas diffusion concentration prediction composite loss information, and the preset gas diffusion concentration prediction composite loss threshold, the gas concentration prediction model to be trained is trained to obtain the gas concentration prediction model.

4. The gas leakage risk level classification method as described in claim 3, characterized in that, The step of calculating the composite loss information for gas diffusion concentration prediction based on the multiple gas diffusion concentration information and the gas concentration prediction model to be trained specifically includes: Based on the multiple gas diffusion concentration information and the gas concentration prediction model to be trained, the gas diffusion concentration fitting loss information, the gas diffusion equation residual loss information, and the gas diffusion boundary condition loss information are calculated. Based on preset weighted coefficients for gas diffusion concentration fitting loss, preset weighted coefficients for gas diffusion equation residual loss, and preset weighted coefficients for gas diffusion boundary condition loss, the gas diffusion concentration fitting loss information, gas diffusion equation residual loss information, and gas diffusion boundary condition loss information are weighted and summed to calculate the gas diffusion concentration prediction composite loss information.

5. The gas leakage risk level classification method as described in claim 1, characterized in that, The step of correcting and fusing the multiple gas concentration prediction information and multiple gas concentration monitoring information to obtain multiple gas concentration correction information specifically includes: The multiple gas concentration monitoring information is supplemented and corrected to obtain multiple corrected gas concentration monitoring information; The multiple gas concentration prediction information and the multiple corrected gas concentration monitoring information are fused together to obtain multiple gas concentration fusion information. Based on preset gas concentration filtering rules, gain correction processing is performed on the multiple gas concentration fusion information to obtain multiple gas concentration correction information.

6. The gas leakage risk level classification method as described in claim 5, characterized in that, The step of fusing the multiple gas concentration prediction information and the multiple corrected gas concentration monitoring information to obtain multiple gas concentration fusion information specifically includes: Based on the preset gas concentration prediction perturbation distribution mean information and the preset gas concentration prediction perturbation distribution covariance information, multiple gas concentration prediction perturbation information are randomly generated. Based on the multiple gas concentration prediction information and the multiple gas concentration prediction perturbation information, multiple perturbed gas concentration prediction information are generated. Based on the preset gas concentration prediction state evolution operator, time integration is performed according to the multiple perturbed gas concentration prediction information to obtain multiple prior gas concentration prediction information. Based on the multiple corrected gas concentration monitoring information and the multiple gas concentration prediction perturbation information, multiple perturbed gas concentration monitoring information are generated; Based on the preset gas concentration gain update rule, the multiple prior gas concentration prediction information and the multiple perturbed gas concentration monitoring information are fused to obtain multiple gas concentration fusion information.

7. The gas leakage risk level classification method as described in claim 1, characterized in that... After the step of generating multiple gas leak risk level classification information based on the multiple gas concentration correction information and multiple preset gas leak risk level classification thresholds, the method further includes: Spatial positioning processing is performed based on the multiple gas leak risk level classification information to obtain multiple gas leak risk level positioning information; Based on the multiple gas leak risk level location information and the preset gas leak emergency evacuation decision rules, multiple gas leak emergency evacuation decision information is generated.

8. A gas leak risk level classification device, characterized in that, include: The information acquisition module is used to acquire multiple gas leak environmental meteorological information, multiple gas leak source location information, and multiple gas concentration monitoring information; The gas concentration prediction information generation module is used to obtain multiple gas concentration prediction information based on the multiple gas leakage environmental meteorological information, multiple gas leakage source location information and gas concentration prediction model; The gas concentration correction information generation module is used to perform correction and fusion processing based on the multiple gas concentration prediction information and multiple gas concentration monitoring information to obtain multiple gas concentration correction information. The gas leakage risk level classification information generation module is used to generate multiple gas leakage risk level classification information based on the multiple gas concentration correction information and multiple preset gas leakage risk level classification thresholds.

9. A terminal device, characterized in that, The terminal device includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.

Images