Nuclear radiation diffusion analysis method based on Gaussian mixture atmospheric diffusion model

Abstract

The invention provides a nuclear radiation diffusion analysis method based on a Gaussian mixture atmospheric diffusion model. The method comprises the steps that nuclear facility source item parameters are acquired; integrating meteorological, geography, remote sensing and radiation monitoring data by using a heterogeneous data fusion engine, extracting atmospheric turbulence, terrain and building interference characteristics and source item parameters, and forming a dynamic input parameter set; constructing an unsteady state diffusion model containing a multi-modal distribution function, processing features, obtaining a mixed distribution model, and correcting the mixed distribution model; predicting spatial and temporal distribution of radiation pollutants, and generating multi-scale diffusion field simulation; and optimizing a diffusion path, and generating a three-dimensional risk field. According to the method, multi-source data are integrated, an unsteady state diffusion model is constructed, Gaussian component weights are dynamically adjusted, the model is corrected by using real-time data, and a geographic information system is coupled to carry out spatial constraint optimization. And through multi-scale feature coupling, dynamic parameter adjustment and geographic space constraint correction, the depicting capability of the atmospheric diffusion model is remarkably improved, and the nuclear radiation diffusion prediction precision is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of nuclear radiation diffusion analysis, and in particular to a nuclear radiation diffusion analysis method based on a Gaussian mixture atmospheric diffusion model. Background Art

[0002] With the widespread application of nuclear energy technology and the inherent potential risks of nuclear accidents, accurate prediction and risk assessment of the atmospheric diffusion of radioactive materials have become crucial issues in nuclear safety. Traditional atmospheric diffusion models (such as the Gaussian plume model) have been widely used under simple meteorological conditions and homogeneous terrain. However, their limited adaptability to complex environmental factors leads to a significant decline in prediction accuracy during sudden nuclear accidents or non-steady-state meteorological conditions, making them unable to meet the sophisticated requirements of modern nuclear emergency management. Traditional Gaussian models assume a normal distribution of pollutant concentrations and fail to fully account for complex factors such as topographical undulation, building obstruction, and turbulent heterogeneity. In mountainous areas, urban canyons, or environments with significant diurnal temperature fluctuations, actual diffusion patterns can exhibit multimodal and asymmetric characteristics, resulting in significant deviations between the predictions of a single Gaussian model and the true diffusion trajectory. Furthermore, nuclear accidents can present dynamic characteristics such as multiple release points, intermittent emissions, or sudden changes in wind direction. Traditional models struggle to integrate multi-source data with spatiotemporally varying meteorological parameters in real time, leading to prediction lags and accumulated errors. Summary of the Invention

[0003] The present invention aims to at least solve the technical problem of low prediction accuracy of existing atmospheric diffusion models in the prior art, and particularly innovatively proposes a nuclear radiation diffusion analysis method based on a Gaussian mixture atmospheric diffusion model.

[0004] In order to achieve the above-mentioned object of the present invention, the present invention provides a nuclear radiation diffusion analysis method based on a Gaussian mixture atmospheric diffusion model, the method comprising:

[0005] S1. Obtain nuclear facility source item parameters;

[0006] S2. Performing spatiotemporal alignment and feature extraction on meteorological observation data, spatial geographic information data, satellite remote sensing data, and real-time radiation monitoring data through a heterogeneous data fusion engine to obtain a dynamic input parameter set including atmospheric turbulence characteristics, terrain barrier effect characteristics, building interference characteristics, and nuclear facility source term parameters;

[0007] S3. Constructing a non-steady-state diffusion model including a multimodal distribution function, and using an adaptive mixing mechanism based on the non-steady-state diffusion model to perform Gaussian mixing processing on the atmospheric turbulence characteristics, terrain barrier effect characteristics, building interference characteristics, and nuclear facility source term parameters in the dynamic input parameter set to obtain a mixed distribution model that describes the diffusion characteristics of nuclear radiation in the atmosphere, and using real-time radiation monitoring data to correct the mixed distribution model;

[0008] S4. Based on the modified mixed distribution model, combined with the nuclide decay characteristics and the evolution of meteorological conditions, the spatiotemporal distribution of radioactive pollutants is probabilistically predicted to generate multi-scale diffusion field simulation results;

[0009] S5. The coupled geographic information system performs spatial constraint analysis on the multi-scale diffusion field simulation results, modifies the diffusion path through the terrain-building synergy algorithm, and generates a three-dimensional risk field that conforms to the actual geographical characteristics.

[0010] The present invention demonstrates the following benefits: This method utilizes a heterogeneous data fusion engine to deeply integrate multi-source data, constructing a non-steady-state diffusion model encompassing multimodal distribution functions. This model employs an adaptive hybrid mechanism to dynamically adjust Gaussian component weights. Furthermore, the hybrid distribution model is closed-loop corrected using real-time radiation monitoring data. Finally, a terrain-building synergy algorithm, coupled with a geographic information system, is introduced to optimize diffusion paths through spatial constraints. This technological system significantly enhances the atmospheric diffusion model's ability to capture complex meteorological conditions, geographic barrier effects, and building interference through its triple mechanism of multi-scale feature coupling, dynamic parameter adaptive adjustment, and geographic spatial constraint correction, thereby significantly improving the accuracy of nuclear radiation diffusion predictions.

[0011] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments with reference to the accompanying drawings, in which:

[0013] Figure 1 The present invention is a flow chart of a nuclear radiation diffusion analysis method based on a Gaussian mixture atmospheric diffusion model. DETAILED DESCRIPTION

[0014] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0015] like Figure 1 As shown, a nuclear radiation diffusion analysis method based on a Gaussian mixture atmospheric diffusion model comprises:

[0016] S1. Obtaining nuclear facility source item parameters;

[0017] In step S1 , in this embodiment, the nuclear facility source term parameters include the type of radionuclide, initial release rate, release height, etc.

[0018] S2. Performing spatiotemporal alignment and feature extraction on meteorological observation data, spatial geographic information data, satellite remote sensing data, and real-time radiation monitoring data through a heterogeneous data fusion engine to obtain a dynamic input parameter set including atmospheric turbulence characteristics, terrain barrier effect characteristics, building interference characteristics, and nuclear facility source term parameters;

[0019] In step S2, the meteorological observation data of this embodiment specifically include conventional meteorological elements such as wind speed, wind direction, temperature, air pressure, humidity, precipitation, visibility, and professional meteorological parameters such as atmospheric turbulence intensity, stability level, and mixing layer height.

[0020] Spatial geographic information data: specifically covers natural geographic information such as digital elevation model (DEM), topographic data, land use type, water system distribution, vegetation cover, as well as human geographic information such as building outlines, road networks, and administrative divisions.

[0021] Satellite remote sensing data: mainly includes satellite image data, such as visible light, infrared, multispectral, hyperspectral, etc., as well as satellite remote sensing inversion products, such as surface temperature, vegetation index, atmospheric water vapor content, etc.

[0022] Real-time radiation monitoring data: specifically includes real-time data from ground radiation monitoring stations, such as gamma dose rate, neutron count, etc., as well as data from mobile monitoring equipment, drone-mounted monitoring data, and emergency monitoring vehicle data.

[0023] S3. Constructing a non-steady-state diffusion model including a multimodal distribution function, and using an adaptive mixing mechanism based on the non-steady-state diffusion model to perform Gaussian mixing processing on the atmospheric turbulence characteristics, terrain barrier effect characteristics, building interference characteristics, and nuclear facility source term parameters in the dynamic input parameter set to obtain a mixed distribution model that describes the diffusion characteristics of nuclear radiation in the atmosphere, and using real-time radiation monitoring data to correct the mixed distribution model;

[0024] It should be noted that in step S3, the construction of the mixed distribution model begins with the acquisition of a dynamic input parameter set. Based on the theory of non-steady-state diffusion, the model constructs a preliminary multimodal distribution function. This function captures the diffusion characteristics of nuclear radiation under different meteorological conditions and geographical environments. Subsequently, the adaptive mixing mechanism comes into play, dynamically adjusting the mean, variance, and weight of the Gaussian components based on real-time changes in atmospheric turbulence characteristics, terrain barrier effects, building interference characteristics, and nuclear facility source parameters. This process ensures that the mixed distribution model can reflect the actual diffusion of nuclear radiation in real time.

[0025] To further improve the model's accuracy, this embodiment incorporates real-time radiation monitoring data to correct the mixed distribution model. By comparing and analyzing the nuclear radiation intensity values ​​predicted by the model, it is possible to identify areas with large prediction errors and fine-tune the model accordingly. This closed-loop correction mechanism not only improves the model's prediction accuracy but also enhances its adaptability to complex environments. Therefore, by constructing a non-steady-state diffusion model containing a multimodal distribution function and employing an adaptive mixing mechanism and real-time radiation monitoring data correction, this embodiment successfully achieves high-precision predictions of nuclear radiation diffusion characteristics.

[0026] S4. Based on the modified mixed distribution model, combined with the nuclide decay characteristics and the evolution of meteorological conditions, the spatiotemporal distribution of radioactive pollutants is probabilistically predicted to generate multi-scale diffusion field simulation results;

[0027] It should be noted that in step S4, during its implementation, the impact of nuclide decay characteristics on the diffusion of radioactive contaminants is first considered. Different nuclides have different decay constants and half-lives, and these characteristics determine how radiation intensity changes over time. During the simulation, this embodiment calculates the residual activity of various nuclides at each time point based on the nuclide types and initial release rates provided in the nuclear facility source term parameters, combined with the nuclide decay formula. This step ensures that the simulation results accurately reflect the impact of nuclide decay on radiation concentration.

[0028] Secondly, the evolution of meteorological conditions is integrated. Meteorological conditions are one of the key factors influencing the diffusion of radioactive pollutants. This embodiment utilizes the spatiotemporal variations of wind speed, wind direction, temperature, and air pressure in meteorological observation data to simulate the impact of meteorological conditions on the diffusion process. By constructing a meteorological condition evolution model, this embodiment can predict meteorological changes over a period of time in the future and incorporate these changes into the diffusion simulation. This step makes the simulation results more closely aligned with the actual meteorological environment, improving the accuracy of the prediction.

[0029] This example uses a probabilistic prediction method to simulate the spatiotemporal distribution of radioactive pollutants, taking into account the decay characteristics of nuclides and the evolution of meteorological conditions. Using Gaussian mixture models and other statistical methods, this example probabilistically predicts radiation concentrations, generating diffusion field simulation results within different confidence intervals. These results not only provide average radiation concentrations but also reflect the uncertainty of the concentration distribution, providing more comprehensive information for decision makers.

[0030] Finally, multi-scale diffusion field simulation results are generated. This embodiment adjusts the simulation scale and resolution to generate diffusion field simulation results at different scales (e.g., local, regional, and global). These results can reflect the diffusion of radioactive contaminants within different spatial ranges. Furthermore, multi-scale simulation results can verify and complement each other, improving the reliability and accuracy of the overall prediction.

[0031] S5. The coupled geographic information system performs spatial constraint analysis on the multi-scale diffusion field simulation results, modifies the diffusion path through the terrain-building synergy algorithm, and generates a three-dimensional risk field that conforms to the actual geographical characteristics.

[0032] In step S5, it is important to note that the spatial analysis capabilities of the geographic information system are first used to compare and calibrate the multi-scale diffusion field simulation results with actual geospatial data. This step ensures that the simulation results accurately reflect the actual impact of geographical elements such as terrain, water systems, vegetation, and buildings on the spread of nuclear radiation. By comparing the simulation results with geospatial data, potential deviations or errors in the simulation can be identified and corrected, thereby improving the accuracy of the simulation.

[0033] Subsequently, a terrain-building synergy algorithm was used to modify the diffusion path. This algorithm comprehensively considers the effects of terrain undulation and building distribution on the diffusion path of nuclear radiation. Terrain undulation alters the direction and speed of airflow, thus affecting the diffusion of nuclear radiation; while the presence of buildings creates barriers, blocking or altering the propagation path of nuclear radiation. By introducing this algorithm, the diffusion process of nuclear radiation in complex geographical environments can be more accurately simulated, resulting in a more realistic three-dimensional risk field.

[0034] Based on the modified diffusion path, the 3D risk field is further visualized and analyzed using a geographic information system. By generating a 3D risk field map, the spread and risk distribution of nuclear radiation across different geographic spaces can be intuitively displayed. This provides important reference information for decision-makers, helping to formulate more scientific and reasonable emergency response measures and long-term risk assessment plans. Furthermore, through continuous iteration and optimization of model parameters, the prediction accuracy and practicality of the 3D risk field can be further improved.

[0035] The principle of the nuclear radiation diffusion analysis method based on the Gaussian mixture atmospheric diffusion model in this embodiment is as follows: First, the method uses a heterogeneous data fusion engine to perform spatiotemporal alignment and feature extraction on meteorological observations, spatial geographic information, satellite remote sensing, and real-time radiation monitoring data to form a dynamic input parameter set that includes key features such as atmospheric turbulence, terrain barrier effects, and building interference.

[0036] Next, based on these dynamic input parameters, a non-steady-state diffusion model with a multimodal distribution function was constructed. This model employs a Gaussian mixture processing mechanism, dynamically adjusting the weights of each Gaussian component to accurately describe the diffusion characteristics of nuclear radiation in the atmosphere. Furthermore, the model was refined using real-time radiation monitoring data, further improving its prediction accuracy.

[0037] After obtaining the modified mixed distribution model, the method combines the decay characteristics of nuclides with the evolution of meteorological conditions to probabilistically predict the spatiotemporal distribution of radioactive pollutants and generate multi-scale diffusion field simulation results. This step achieves a comprehensive simulation of the radiation diffusion process.

[0038] Finally, by coupling a geographic information system, we conduct a spatial constraint analysis on the multi-scale diffusion field simulation results. We use a network analysis algorithm to simulate the propagation path and velocity of radiation within the geographic network, and incorporate a terrain-building synergy algorithm to correct the diffusion path, ultimately generating a three-dimensional risk field that conforms to the actual geographic characteristics.

[0039] In summary, the nuclear radiation diffusion analysis method based on the Gaussian mixture atmospheric diffusion model of this embodiment can accurately simulate and predict the nuclear radiation diffusion process. In specific applications, this method can help decision makers quickly identify high-risk areas, formulate targeted emergency measures, and effectively mitigate the impact of nuclear radiation accidents on the environment and public health.

[0040] As an optional embodiment of the present invention, optionally, obtaining the dynamic input parameter set in step S2 includes:

[0041] S201. Extracting atmospheric turbulence features from meteorological observation data using convolutional neural networks and attention mechanisms.

[0042] In step S201, it should be noted that the convolutional neural network (CNN), through its convolutional and pooling layer structures, can effectively extract the spatial and temporal characteristics of atmospheric turbulence from meteorological observation data. These characteristics reflect the intensity, scale, and evolution of atmospheric turbulence and are crucial for simulating the diffusion of nuclear radiation in the atmosphere. Furthermore, the introduction of the attention mechanism enables the model to focus more on key features that have a significant impact on the diffusion of nuclear radiation, thereby improving the accuracy and efficiency of feature extraction.

[0043] The introduction of the attention mechanism into a convolutional neural network is achieved through the following method: First, a basic convolutional neural network architecture, consisting of multiple convolutional and pooling layers, is selected to extract feature maps from the input meteorological observation data. Next, an attention mechanism module is introduced within this convolutional neural network architecture. This module calculates the importance weight of each location in the feature map and reweights the feature map based on these weights, thereby highlighting features that are more critical for predicting nuclear radiation spread. In practice, either a channel-wise or spatial-wise attention mechanism can be used. When training the convolutional neural network with the attention mechanism, supervised learning is performed using a large amount of atmospheric turbulence observation data covering various meteorological conditions. By continuously adjusting the network parameters, the model is able to accurately identify and extract atmospheric turbulence features that are closely related to nuclear radiation spread. This process ensures the model's accuracy and generalization in practical applications.

[0044] S202. Extracting terrain barrier effect features using a terrain recognition algorithm based on spatial geographic information data and satellite remote sensing data, and performing spatial correction on the terrain barrier effect features using a geographic information system;

[0045] In step S202, it should be noted that the terrain recognition algorithm, by analyzing spatial geographic information data and satellite remote sensing data, can identify topographic features such as undulating terrain, mountain ranges, and river valleys. These features have a significant impact on the diffusion path of nuclear radiation. After the terrain barrier effect features are extracted, the Geographic Information System (GIS) further performs spatial correction on these features to ensure their accuracy and reliability. This spatial correction process takes into account factors such as the geographic coordinate system, terrain projection method, and the accuracy of the terrain data, thereby improving the usability of the terrain barrier effect features in practical applications.

[0046] S203. Extract building interference features using a target detection algorithm in a deep learning framework based on satellite remote sensing data and real-time radiation monitoring data.

[0047] In step S203, it should be noted that the target detection algorithm accurately identifies key information such as building outlines, heights, and density by analyzing satellite remote sensing data and real-time radiation monitoring data. The introduction of a deep learning framework enables the algorithm to automatically learn and extract useful features from large amounts of complex data, thereby improving the accuracy and efficiency of building interference feature extraction. By comprehensively considering factors such as building location, shape, and density, the diffusion process of nuclear radiation in complex urban environments can be more accurately simulated.

[0048] In this embodiment, the deep learning framework is specifically a mainstream framework such as TensorFlow or PyTorch.

[0049] To extract building interference features, the target detection algorithm used is the YOLO (You Only Look Once) family of deep learning algorithms or the Faster R-CNN algorithm. The YOLO algorithm is known for its high speed and good accuracy, making it suitable for scenarios with high real-time requirements. Faster R-CNN, on the other hand, offers superior accuracy and is suitable for scenarios requiring higher precision. In implementation, the appropriate algorithm for building interference feature extraction is selected based on the characteristics of satellite remote sensing data and real-time radiation monitoring data, as well as the requirements of the actual application scenario. By fine-tuning the algorithm parameters and optimizing the feature extraction process, the accuracy and efficiency of building interference feature extraction can be further improved.

[0050] S204 , performing spatiotemporal processing on the extracted atmospheric turbulence characteristics, terrain barrier effect characteristics, building interference characteristics, and nuclear facility source item parameters based on a heterogeneous data fusion engine to obtain the dynamic input parameter set.

[0051] In step S204, it's important to note that the heterogeneous data fusion engine uses advanced algorithms and technologies to precisely align and fuse data from diverse sources and formats. Specifically, through three steps: time synchronization, spatial matching, and data fusion, the engine organically integrates the extracted atmospheric turbulence characteristics, terrain barrier effect characteristics, building interference characteristics, and nuclear facility source parameters, forming a comprehensive and accurate dynamic input parameter set.

[0052] As an optional embodiment of the present invention, optionally, constructing a non-steady-state diffusion model including a multimodal distribution function in step S3 includes:

[0053] S301. Introducing a time variable into the steady-state model and expanding the diffusion equation in the steady-state model from a static form to a dynamic form, thereby obtaining the basic framework of the unsteady-state diffusion model;

[0054] The steady-state model selected in this embodiment is the classic Gaussian plume model or Gaussian puff model.

[0055] S302. Constructing corresponding multimodal distribution functions for different types of nuclear radiation, using multiple Gaussian components in a Gaussian mixture model to weight the probability distribution characteristics of nuclear radiation under different diffusion conditions in the multimodal distribution functions, and embedding the multimodal distribution functions into the basic framework of the non-steady-state diffusion model;

[0056] S303, using the Kalman filter algorithm in combination with real-time radiation monitoring data to update the weight of the Gaussian component;

[0057] S304. Introducing a covariance matrix into the basic framework of the non-steady-state diffusion model, using the covariance matrix to describe the diffusion direction and diffusion scale of nuclear radiation diffusion, and updating the covariance matrix through an iterative optimization algorithm to obtain a non-steady-state diffusion model.

[0058] The principle behind constructing the aforementioned unsteady diffusion model is to first introduce a time variable into the steady-state model, expanding the static diffusion equation into a dynamic form and laying the foundation for the model. This step enables the model to describe the time-varying diffusion of nuclear radiation, more accurately resembling actual conditions. Next, corresponding multimodal distribution functions are constructed for different types of nuclear radiation. These functions utilize weighted Gaussian components from a Gaussian mixture model to represent the probability distribution characteristics of nuclear radiation under different diffusion conditions. By embedding the multimodal distribution functions into the basic framework of the unsteady diffusion model, the model can more accurately describe the complex diffusion process of nuclear radiation in the atmosphere, improving the model's ability to describe the diffusion characteristics of nuclear radiation. The Kalman filter algorithm is then used to update the weights of the Gaussian components in conjunction with real-time radiation monitoring data. The Kalman filter is an efficient recursive filter that can update state estimates based on new measurement data in the presence of noise and uncertainty. By incorporating this algorithm, the model can adjust the weights of the Gaussian components in real time based on radiation monitoring data, thereby improving the accuracy and real-time performance of predictions. Finally, the covariance matrix is ​​introduced into the basic framework of the non-steady-state diffusion model to describe the diffusion direction and scale of nuclear radiation. The covariance matrix reflects the correlation and variability between variables. By updating the covariance matrix through an iterative optimization algorithm, the parameters of the non-steady-state diffusion model can be further optimized. This step enables the model to more accurately simulate the diffusion of nuclear radiation in different directions, improving the accuracy and reliability of the overall prediction.

[0059] The mixed distribution model describing the diffusion characteristics of nuclear radiation in the atmosphere obtained in step S3 includes:

[0060] S305. Based on the adaptive hybrid mechanism, weights are assigned to the atmospheric turbulence characteristics, terrain barrier effect characteristics, building interference characteristics, and nuclear facility source term parameters in the dynamic input parameter set, and the weights of the Gaussian components are dynamically adjusted according to the degree of influence of each characteristic on nuclear radiation diffusion.

[0061] S306. Using multiple Gaussian components in a Gaussian mixture model to mix the weighted features, each Gaussian component represents the probability distribution of nuclear radiation under specific diffusion conditions;

[0062] S307. Continuously adjust the mean, variance, and weight of the Gaussian components through an iterative optimization algorithm until the mixed distribution model can accurately describe the diffusion characteristics of nuclear radiation in the atmosphere, thereby generating a final mixed distribution model.

[0063] In step S3, the mixed distribution model is modified using the real-time radiation monitoring data, including:

[0064] S308. Compare and analyze the nuclear radiation intensity values ​​in the real-time radiation monitoring data with the nuclear radiation intensity values ​​predicted by the mixed distribution model to identify areas with large prediction errors; for areas with large prediction errors, readjust the relevant feature weights in the dynamic input parameter set, or reconfigure the Gaussian components in the mixed distribution model.

[0065] In Steps S301 to S307 described above, in Step S3, a non-steady-state diffusion model including a multimodal distribution function is first constructed. Building on the traditional steady-state model, the present invention introduces a time variable, expanding the originally static diffusion equation into a dynamic form to describe the temporal variation of the nuclear radiation diffusion process, thereby obtaining the basic framework of the non-steady-state diffusion model. Corresponding multimodal distribution functions are constructed for different types of nuclear radiation. Within the multimodal distribution functions, multiple Gaussian components in a Gaussian mixture model are weighted to represent the probability distribution characteristics of nuclear radiation under different diffusion conditions. These multimodal distribution functions are then embedded within the basic framework of the non-steady-state diffusion model.

[0066] Next, based on the unsteady-state diffusion model, an adaptive mixing mechanism is employed to perform a Gaussian mixture process on the dynamic input parameter set, including atmospheric turbulence characteristics, terrain barrier effect characteristics, building interference characteristics, and nuclear facility source term parameters. This adaptive mixing mechanism dynamically adjusts the weights of each Gaussian component based on the degree of influence of each characteristic on nuclear radiation diffusion. For example, in areas with complex terrain, where terrain barrier effect characteristics have a greater impact on diffusion, the weights of the associated Gaussian components are correspondingly increased. These weighted features are then mixed using multiple Gaussian components in the Gaussian mixture model, each representing the probability distribution of nuclear radiation under specific diffusion conditions. This approach provides a more comprehensive description of the diffusion characteristics of nuclear radiation in complex environments.

[0067] During model construction and Gaussian mixture processing, an iterative optimization algorithm continuously adjusts the mean, variance, and weights of the Gaussian components until the mixture distribution model accurately describes the diffusion characteristics of nuclear radiation in the atmosphere, generating the final mixture distribution model. Simultaneously, the mixture distribution model is refined using real-time radiation monitoring data. A Kalman filter algorithm, combined with real-time radiation monitoring data, updates the weights of the Gaussian components, enabling the model to reflect the actual diffusion of nuclear radiation in real time. Furthermore, a covariance matrix is ​​introduced into the basic framework of the non-steady-state diffusion model to describe the diffusion direction and scale of nuclear radiation. This covariance matrix is ​​updated using an iterative optimization algorithm to further improve model accuracy.

[0068] Furthermore, a comparative analysis of nuclear radiation intensity values ​​from real-time radiation monitoring data and those predicted by the mixture distribution model was conducted to identify areas with large prediction errors. For these areas, the weights of relevant features in the dynamic input parameter set were readjusted, or the Gaussian components in the mixture distribution model were reconfigured to ensure that the mixture distribution model accurately describes the diffusion characteristics of nuclear radiation in the atmosphere.

[0069] As an optional embodiment of the present invention, optionally, the expression for probabilistically predicting the spatiotemporal distribution of radioactive pollutants in step S4 is:

[0070]

[0071] Among them, C pred (x,t) represents the predicted radiation concentration at location x and time t;

[0072] x represents the coordinates in three-dimensional space (longitude, latitude, altitude);

[0073] t represents the time variable, specifically the predicted time starting from the initial moment;

[0074] Ω represents the space of all possible parameter combinations;

[0075] M represents the number of Gaussian components in the Gaussian mixture model; each component represents a diffusion mode;

[0076] ω i (t) represents the weight of the i-th Gaussian component at time t;

[0077] N(x;μ i (t),∑ i (t)) represents the probability density function of the i-th Gaussian component;

[0078] μ i (t) represents the mean vector of the i-th Gaussian component at time t;

[0079] ∑ i (t) represents the covariance matrix of the i-th Gaussian component at time t;

[0080] exp() represents the exponential function;

[0081] U represents the number of nuclides;

[0082] λ j represents the decay constant of the jth nuclide;

[0083] t′ represents the time variable;

[0084] τ j represents the decay time constant of the jth nuclide;

[0085] M(x,t) represents the meteorological condition evolution function at location x and time t, describing the influence of meteorological parameters such as wind speed and turbulence on diffusion;

[0086] θ represents the parameter vector of the mixed distribution model, including the weights, means, covariance matrix elements of the Gaussian components, and meteorological condition parameters.

[0087] As an optional embodiment of the present invention, optionally, the expression for generating the multi-scale diffusion field simulation result in step S4 is:

[0088]

[0089] Among them, D multi (x, t) represents the value of the multi-scale diffusion field simulation result at position x and time t;

[0090] S represents the scale set, which can include micro, meso and macro scales;

[0091] Ω s It represents the computational domain corresponding to the scale s, which expands as the scale increases;

[0092] K s () represents the scale-dependent spatial kernel function, which realizes spatial scale conversion;

[0093] x and x′ represent coordinates in three-dimensional space;

[0094] T s () represents the scale-dependent time kernel function, which realizes the time scale conversion;

[0095] t and t′ represent time variables;

[0096] Represents the single-scale predicted concentration field, the prediction result at scale s.

[0097] As an optional embodiment of the present invention, optionally, in step S5, performing spatial constraint analysis on the multi-scale diffusion field simulation results by the coupled geographic information system includes:

[0098] S501, obtaining geospatial data through a geographic information system, and preprocessing the geospatial data and multi-scale diffusion field simulation results;

[0099] S502, spatially superimposing the pre-processed geospatial data and the multi-scale diffusion field simulation results, and then analyzing the impact of the geospatial data on the multi-scale diffusion field simulation results to obtain spatial image data;

[0100] S503. quantifying the degree of influence of different geographic elements in the geospatial data on the multi-scale diffusion field simulation results based on the spatial image data;

[0101] S504: establishing a buffer zone based on specific geographical elements and their influence on the multi-scale diffusion field simulation results, and analyzing the distribution of radiation concentration in the core of the buffer zone based on the multi-scale diffusion field simulation results of the buffer zone;

[0102] S505: Setting a nuclear radiation concentration threshold for the buffer zone, and identifying an area where the nuclear radiation concentration exceeds the nuclear radiation concentration threshold based on the nuclear radiation concentration threshold and the distribution of nuclear radiation concentration in the buffer zone;

[0103] S506. Establish a geographic network based on the geographic spatial data, couple the multi-scale diffusion field simulation results with the geographic network, simulate the diffusion process of nuclear radiation in the geographic network, obtain simulation results for each network node in the geographic network, and use the simulation results as attribute values ​​of the network nodes;

[0104] S507 , simulating the propagation path and speed of nuclear radiation through a network analysis algorithm based on the attribute values ​​of the network nodes.

[0105] It should be noted that the entire analysis process, from steps S501 to S50, fully utilizes the capabilities of a geographic information system (GIS). GIS not only provides detailed geospatial data, such as topography, landforms, and building distribution, but also provides a framework for spatially constrained analysis of multi-scale diffusion field simulation results. The geospatial data acquired through GIS can more accurately reflect the complex terrain and building layouts found in real-world environments, thereby improving the accuracy of nuclear radiation diffusion simulations.

[0106] In step S502, the preprocessed geospatial data and the multi-scale diffusion field simulation results are spatially overlaid. This process organically integrates environmental characteristics with the nuclear radiation diffusion simulation results. Through overlay analysis, the impact of geospatial data on the multi-scale diffusion field simulation results can be intuitively observed, further understanding the diffusion characteristics of nuclear radiation in different geographical environments.

[0107] Steps S503 to S505 further quantify the impact of geographic factors on the multi-scale diffusion field simulation results and establish a buffer zone to analyze the distribution of nuclear radiation concentration within the buffer zone. This step not only helps identify areas with excessive nuclear radiation concentration.

[0108] Finally, in steps S506 and S507, the diffusion of nuclear radiation within the geographic network is simulated by coupling the multiscale diffusion field simulation results with the geographic network. This process not only considers the diffusion of nuclear radiation in the natural environment but also fully accounts for the influence of geographic factors on the diffusion process, resulting in a more accurate and comprehensive simulation result.

[0109] The expression of the network analysis algorithm in step S507 is:

[0110]

[0111] Among them, I y (t+Δt) represents the nuclear radiation intensity of network node y at time t+Δt;

[0112] I y (t) represents the nuclear radiation intensity of network node y at time t;

[0113] represents the natural attenuation factor of network node y within the time step Δt;

[0114] λ y represents the nuclear radiation attenuation coefficient of network node y;

[0115] N y represents the set of network nodes connected to network node y;

[0116] D xy (t) represents the nuclear radiation propagation direction coefficient from network node x to network node y;

[0117] d xy Represents the distance between network node x and network node y;

[0118] ω xy (t) the efficiency coefficient of nuclear radiation propagation from network node x to network node y;

[0119] Ix (t) represents the nuclear radiation intensity of network node x at time t;

[0120] Δt represents the time step.

[0121] In step S5, the diffusion path is corrected by the terrain-building synergy algorithm to generate a three-dimensional risk field that conforms to the actual geographical characteristics, including:

[0122] S508: Analyze the propagation path and speed of the nuclear radiation and the areas in the buffer zone where the nuclear radiation concentration exceeds the nuclear radiation concentration threshold to obtain the location and range of the high-risk areas in the geographic network;

[0123] It should be noted that in step S508, the propagation path and speed of nuclear radiation and the areas in the buffer zone where the nuclear radiation concentration exceeds the nuclear radiation concentration threshold are comprehensively analyzed to determine the high-risk areas in the geographic network. Specifically, by integrating the radiation propagation path and speed information obtained by the network analysis algorithm and the high-concentration areas identified by the buffer zone analysis, the specific location and scope of the high-risk areas can be determined.

[0124] S509. The diffusion path of nuclear radiation is corrected by taking into account the terrain undulation factor through the terrain-building synergy algorithm, and a three-dimensional risk field is generated according to the corrected diffusion path of nuclear radiation and the location and range of high-risk areas in the geographic network.

[0125] It should be noted that in step S509, the terrain-building synergy algorithm is used to correct the diffusion path of nuclear radiation to generate a three-dimensional risk field that conforms to actual geographical characteristics. This algorithm comprehensively considers the impact of terrain undulation and building distribution on radiation diffusion, and fine-tunes the initially generated diffusion path by introducing a terrain correction factor and a building shielding coefficient. The terrain correction factor reflects the obstruction or guidance effect of terrain undulation on radiation diffusion, while the building shielding coefficient describes the blocking effect of buildings on radiation propagation. By combining these factors, the algorithm can more accurately simulate the diffusion process of radiation in complex geographical environments.

[0126] Based on the revised diffusion paths and the locations and extent of high-risk areas, a three-dimensional risk field is generated. This risk field provides a three-dimensional visualization of the spatial spread and risk distribution of nuclear radiation. This 3D risk field allows decision-makers to clearly identify high-risk areas, assess the risk levels of different areas, and develop targeted emergency response measures and long-term risk assessment plans accordingly.

[0127] This embodiment corrects the diffusion path through the terrain-building synergy algorithm and generates a three-dimensional risk field that conforms to actual geographical characteristics, significantly improving the accuracy and practicality of nuclear radiation diffusion analysis.

[0128] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and alterations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. A method for analyzing nuclear radiation diffusion based on a Gaussian mixture atmospheric diffusion model, characterized in that: The method comprises: S1. Obtaining nuclear facility source item parameters; S2. Performing spatiotemporal alignment and feature extraction on meteorological observation data, spatial geographic information data, satellite remote sensing data, and real-time radiation monitoring data through a heterogeneous data fusion engine to obtain a dynamic input parameter set including atmospheric turbulence characteristics, terrain barrier effect characteristics, building interference characteristics, and nuclear facility source term parameters; S3. Constructing a non-steady-state diffusion model including a multimodal distribution function, and using an adaptive mixing mechanism based on the non-steady-state diffusion model to perform Gaussian mixing processing on the atmospheric turbulence characteristics, terrain barrier effect characteristics, building interference characteristics, and nuclear facility source term parameters in the dynamic input parameter set to obtain a mixed distribution model that describes the diffusion characteristics of nuclear radiation in the atmosphere, and using real-time radiation monitoring data to correct the mixed distribution model; S4. Based on the modified mixed distribution model, combined with the nuclide decay characteristics and the evolution of meteorological conditions, the spatiotemporal distribution of radioactive pollutants is probabilistically predicted to generate multi-scale diffusion field simulation results; S5. The coupled geographic information system performs spatial constraint analysis on the multi-scale diffusion field simulation results, modifies the diffusion path through the terrain-building synergy algorithm, and generates a three-dimensional risk field that conforms to the actual geographical characteristics.

2. The method for analyzing nuclear radiation diffusion based on the Gaussian mixture atmospheric diffusion model according to claim 1, wherein: Obtaining the dynamic input parameter set in step S2 includes: S201. Extracting atmospheric turbulence features from meteorological observation data using convolutional neural networks and attention mechanisms. S202. Extracting terrain barrier effect features using a terrain recognition algorithm based on spatial geographic information data and satellite remote sensing data, and performing spatial correction on the terrain barrier effect features using a geographic information system; S203. Extract building interference features using a target detection algorithm in a deep learning framework based on satellite remote sensing data and real-time radiation monitoring data. S204 , performing spatiotemporal processing on the extracted atmospheric turbulence characteristics, terrain barrier effect characteristics, building interference characteristics, and nuclear facility source item parameters based on a heterogeneous data fusion engine to obtain the dynamic input parameter set.

3. The method for analyzing nuclear radiation diffusion based on the Gaussian mixture atmospheric diffusion model according to claim 1, wherein: Constructing a non-steady-state diffusion model including a multimodal distribution function in step S3 includes: S301. Introducing a time variable into the steady-state model and expanding the diffusion equation in the steady-state model from a static form to a dynamic form, thereby obtaining the basic framework of the unsteady-state diffusion model; S302. Constructing corresponding multimodal distribution functions for different types of nuclear radiation, using multiple Gaussian components in a Gaussian mixture model to weight the probability distribution characteristics of nuclear radiation under different diffusion conditions in the multimodal distribution functions, and embedding the multimodal distribution functions into the basic framework of the non-steady-state diffusion model; S303, using the Kalman filter algorithm in combination with real-time radiation monitoring data to update the weight of the Gaussian component; S304. Introducing a covariance matrix into the basic framework of the non-steady-state diffusion model, using the covariance matrix to describe the diffusion direction and diffusion scale of nuclear radiation diffusion, and updating the covariance matrix through an iterative optimization algorithm to obtain a non-steady-state diffusion model.

4. The method for analyzing nuclear radiation diffusion based on the Gaussian mixture atmospheric diffusion model according to claim 3, wherein: The mixed distribution model describing the diffusion characteristics of nuclear radiation in the atmosphere obtained in step S3 includes: S305. Based on the adaptive hybrid mechanism, weights are assigned to the atmospheric turbulence characteristics, terrain barrier effect characteristics, building interference characteristics, and nuclear facility source term parameters in the dynamic input parameter set, and the weights of the Gaussian components are dynamically adjusted according to the degree of influence of each characteristic on nuclear radiation diffusion. S306. Using multiple Gaussian components in a Gaussian mixture model to mix the weighted features, each Gaussian component represents the probability distribution of nuclear radiation under specific diffusion conditions; S307. Continuously adjust the mean, variance, and weight of the Gaussian components through an iterative optimization algorithm until the mixed distribution model can accurately describe the diffusion characteristics of nuclear radiation in the atmosphere, thereby generating a final mixed distribution model.

5. The method for analyzing nuclear radiation diffusion based on a Gaussian mixture atmospheric diffusion model according to claim 1, 3 or 4, wherein: In step S3, the mixed distribution model is modified using the real-time radiation monitoring data, including: S308. Compare and analyze the nuclear radiation intensity values ​​in the real-time radiation monitoring data with the nuclear radiation intensity values ​​predicted by the mixed distribution model to identify areas with large prediction errors; for areas with large prediction errors, readjust the relevant feature weights in the dynamic input parameter set, or reconfigure the Gaussian components in the mixed distribution model.

6. The method for analyzing nuclear radiation diffusion based on the Gaussian mixture atmospheric diffusion model according to claim 1, wherein: The expression for probabilistic prediction of the spatiotemporal distribution of radioactive pollutants in step S4 is: Among them, C pred (x, t) represents the predicted radiation concentration at position x and time t, x represents the coordinate in three-dimensional space, t represents the time variable, Ω represents the space of all possible parameter combinations, M represents the number of Gaussian components in the Gaussian mixture model, each component represents a diffusion mode, ω i (t) represents the weight of the i-th Gaussian component at time t, N(x; μ i (t),Σ i (t)) represents the probability density function of the i-th Gaussian component, μ i (t) represents the mean vector of the i-th Gaussian component at time t, ∑ i (t) represents the covariance matrix of the i-th Gaussian component at time t, exp() represents the exponential function, U represents the number of nuclides, and λ j represents the decay constant of the jth nuclide, t′ represents the time variable, τ j represents the decay time constant of the jth nuclide, M(x,t) represents the meteorological condition evolution function at position x and time t, and θ represents the parameter vector of the mixture distribution model.

7. The method for analyzing nuclear radiation diffusion based on a Gaussian mixture atmospheric diffusion model according to claim 1 or 6, wherein: The expression for generating the multi-scale diffusion field simulation result in step S4 is: Among them, D multi (x, t) represents the value of the multi-scale diffusion field simulation result at position x and time t, S represents the scale set, Ω s represents the computational domain corresponding to scale s, K s () represents the scale-dependent spatial kernel function, x and x′ represent the coordinates in three-dimensional space, T s () represents the scale-dependent time kernel function, t and t′ represent time variables, Represents the single-scale predicted concentration field, the prediction result at scale s.

8. The method for analyzing nuclear radiation diffusion based on the Gaussian mixture atmospheric diffusion model according to claim 1, wherein: In step S5, the spatial constraint analysis of the multi-scale diffusion field simulation results by the coupled geographic information system includes: S501, obtaining geospatial data through a geographic information system, and preprocessing the geospatial data and multi-scale diffusion field simulation results; S502, spatially superimposing the pre-processed geospatial data and the multi-scale diffusion field simulation results, and then analyzing the impact of the geospatial data on the multi-scale diffusion field simulation results to obtain spatial image data; S503. quantifying the degree of influence of different geographic elements in the geospatial data on the multi-scale diffusion field simulation results based on the spatial image data; S504: establishing a buffer zone based on specific geographical elements and their influence on the multi-scale diffusion field simulation results, and analyzing the distribution of radiation concentration in the core of the buffer zone based on the multi-scale diffusion field simulation results of the buffer zone; S505: Setting a nuclear radiation concentration threshold for the buffer zone, and identifying an area where the nuclear radiation concentration exceeds the nuclear radiation concentration threshold based on the nuclear radiation concentration threshold and the distribution of nuclear radiation concentration in the buffer zone; S506. Establish a geographic network based on the geographic spatial data, couple the multi-scale diffusion field simulation results with the geographic network, simulate the diffusion process of nuclear radiation in the geographic network, obtain simulation results for each network node in the geographic network, and use the simulation results as attribute values ​​of the network nodes; S507 , simulating the propagation path and speed of nuclear radiation through a network analysis algorithm based on the attribute values ​​of the network nodes.

9. The method for analyzing nuclear radiation diffusion based on the Gaussian mixture atmospheric diffusion model according to claim 8, wherein: The expression of the network analysis algorithm in step S507 is: Among them, I y (t+Δt) represents the nuclear radiation intensity of network node y at time t+Δt, I y (t) represents the nuclear radiation intensity of network node y at time t, represents the natural attenuation factor of network node y within the time step Δt, λ y Represents the nuclear radiation attenuation coefficient of network node y, N y represents the set of network nodes connected to network node y, D xy (t) represents the nuclear radiation propagation direction coefficient from network node x to network node y, d xy Represents the distance between network node x and network node y, ω xy (t) Nuclear radiation transmission efficiency coefficient from network node x to network node y, I x (t) represents the nuclear radiation intensity of network node x at time t, and Δt represents the time step.

10. The method for analyzing nuclear radiation diffusion based on the Gaussian mixture atmospheric diffusion model according to claim 8, wherein: In step S5, the diffusion path is corrected by the terrain-building synergy algorithm to generate a three-dimensional risk field that conforms to the actual geographical characteristics, including: S508: Analyze the propagation path and speed of the nuclear radiation and the areas in the buffer zone where the nuclear radiation concentration exceeds the nuclear radiation concentration threshold to obtain the location and range of the high-risk areas in the geographic network; S509. The diffusion path of nuclear radiation is corrected by taking into account the terrain undulation factor through the terrain-building synergy algorithm, and a three-dimensional risk field is generated according to the corrected diffusion path of nuclear radiation and the location and range of high-risk areas in the geographic network.

Images