Nuclear accident action decision method and device, computer device and storage medium

Abstract

The application relates to a nuclear accident action decision method and device, computer equipment and a storage medium. The method comprises the following steps: obtaining nuclear accident simulation data, the nuclear accident simulation data comprising release source information, diffusion result information and environmental factor information at different time points on a nuclear accident simulation timeline; taking the time points on the nuclear accident simulation timeline as time points of a current prediction step in sequence, and obtaining position information of a disaster body object at the current prediction step; obtaining action decision information of the current prediction step based on the position information and the release source information, the diffusion result information and the environmental factor information at the time point corresponding to the current prediction step; and generating action decision information of the disaster body object in the nuclear accident according to the action decision information of all the prediction steps, so as to realize dynamic decision action prediction and provide dynamic decision support for a scenario in which multiple disaster bodies and decision events dynamically change in the nuclear accident.

Description

Technical Field

[0001] This application relates to the field of nuclear accident emergency decision-making, and in particular to a method, apparatus, computer equipment, and computer-readable storage medium (hereinafter referred to as storage medium) for nuclear accident action decision-making. Background Technology

[0002] Nuclear emergency response and decision support after a nuclear accident are essential safeguards and the last line of defense for guiding nuclear accident rescue and operations. In the event of a nuclear accident that could lead to the release of large amounts of radioactive contaminants, nuclear emergency response decisions involve many dynamically changing factors. How to quickly make emergency response decisions based on the real-time dynamic situation of the nuclear accident is a pressing technical problem that needs to be solved in this field. Summary of the Invention

[0003] Therefore, it is necessary to provide a nuclear accident action decision-making method, apparatus, computer equipment, and storage medium to address the aforementioned technical problems and obtain emergency response actions for affected objects.

[0004] Firstly, this application provides a nuclear accident action decision-making method, including:

[0005] Acquire nuclear accident simulation data, which includes release source information, diffusion result information, and environmental factor information at different time points on the nuclear accident simulation timeline;

[0006] The time points on the nuclear accident simulation timeline are sequentially used as the time points of the current prediction step to obtain the location information of the disaster-affected object in the current prediction step.

[0007] Based on the location information and the release source information, diffusion result information and environmental factor information at the time point corresponding to the current prediction step, obtain the action decision information for the current prediction step;

[0008] Based on the action decision information from all prediction steps, action decision information for the affected object in the nuclear accident is generated.

[0009] In some embodiments of this application, obtaining the location information of the disaster-affected object in the current prediction step includes:

[0010] Obtain the location information and action decision information of the disaster-affected object in the preceding prediction step;

[0011] Based on the location information and action decision information of the previous prediction step, the location information of the disaster-affected object in the current prediction step is calculated.

[0012] In some embodiments of this application, obtaining action decision information for the current prediction step based on the location information and the release source information, diffusion result information, and environmental factor information at the time point corresponding to the current prediction step includes:

[0013] Based on the location information, the target release source information, target diffusion result information, and target environmental factor information are determined from the release source information, diffusion result information, and environmental factor information at the time point corresponding to the current prediction step.

[0014] The target release source information, target diffusion result information, and target environmental factor information are input into the action decision model, and the probability values ​​corresponding to different decision actions in the current prediction step are obtained through the action decision model.

[0015] The target decision action is determined from different decision actions based on the probability value corresponding to each decision action.

[0016] In some embodiments of this application, before inputting the target release source information, target diffusion result information, and target environmental factor information into the action decision model, the method further includes:

[0017] Obtain nuclear accident simulation samples, which include release source samples, diffusion result samples, and environmental factor samples at different time points on the nuclear accident simulation timeline;

[0018] The time points on the nuclear accident simulation timeline are used sequentially as sample time points to determine the optimal decision-making action of the affected body sample at the sample time points.

[0019] Based on the release source information, diffusion result information, environmental factor information, and optimal decision action at the sample time points, the random forest model is trained to obtain the action decision model.

[0020] In some embodiments of this application, determining the optimal decision action for the disaster-affected sample at the sample time point includes:

[0021] Starting from the sample time point, the decision-action combinations are iterated from the sample time point to the last time point in the nuclear accident simulation timeline.

[0022] Based on the release source samples, diffusion result samples, and environmental factor samples at different time points, calculate the total expected dose of the disaster-affected body sample for each of the aforementioned decision action combinations;

[0023] Based on the total expected dose of each of the aforementioned decision action combinations, determine the optimal decision action combination from the decision action combinations;

[0024] The decision action at the sample time point in the optimal decision action combination is determined as the optimal decision action at the sample time point.

[0025] In some embodiments of this application, calculating the total expected dose of the affected body sample for each of the aforementioned decision-making action combinations based on release source samples, diffusion result samples, and environmental factor samples at different time points includes:

[0026] For the target decision action combination in the decision action combination, the time point from the sample time point to the last time point in the nuclear accident simulation timeline is taken as the time point of the current prediction step to obtain the location information of the disaster body sample in the current prediction step.

[0027] Determine the target decision action for the current prediction step from the combination of target decision actions, and calculate the expected dose of the disaster-affected sample when the target decision action is taken in the current prediction step based on the location information and the release source information, diffusion result information and environmental factor information at the time point corresponding to the current prediction step.

[0028] Based on the expected doses corresponding to all prediction steps, calculate the total expected dose of the target decision action combination taken by the disaster-affected sample.

[0029] In some embodiments of this application, calculating the total expected dose of the disaster-affected sample taking the target decision action combination based on the expected doses corresponding to all prediction steps includes:

[0030] The discount factor for each prediction step is determined according to the order of the prediction steps.

[0031] The expected dose of each prediction step is weighted and summed based on the discount factor of each prediction step to obtain the total expected dose of the target decision action combination.

[0032] Secondly, this application provides a nuclear accident action decision-making device, comprising:

[0033] The simulation data acquisition module is used to acquire nuclear accident simulation data, which includes release source information, diffusion result information, and environmental factor information at different time points on the nuclear accident simulation timeline.

[0034] The location information determination module is used to sequentially use the time points on the nuclear accident simulation timeline as the time points of the current prediction step to obtain the location information of the disaster-affected object in the current prediction step.

[0035] The action decision prediction module is used to obtain action decision information for the current prediction step based on the location information and the release source information, diffusion result information and environmental factor information at the time point corresponding to the current prediction step.

[0036] The action decision acquisition module is used to generate action decision information for the affected object in the nuclear accident based on the action decision information of all prediction steps.

[0037] Thirdly, this application also provides a computer device, which includes:

[0038] One or more processors; memory; and one or more applications, wherein one or more applications are stored in memory and configured to be executed by the processor to implement a nuclear accident action decision-making method.

[0039] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, the computer program being loaded by a processor to execute steps in a nuclear accident action decision-making method.

[0040] Fifthly, embodiments of this application provide a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method provided in the first aspect described above.

[0041] The aforementioned nuclear accident action decision-making method, apparatus, computer equipment, and storage medium acquire nuclear accident simulation data, including release source information, diffusion result information, and environmental factor information at different time points on the nuclear accident simulation timeline; sequentially use the time points on the nuclear accident simulation timeline as the time points of the current prediction step to obtain the location information of the affected object at the current prediction step; based on the location information and the release source information, diffusion result information, and environmental factor information at the corresponding time point of the current prediction step, obtain action decision information for the current prediction step; and generate action decision information for the affected object in the nuclear accident based on the action decision information of all prediction steps. This enables dynamic decision-making action prediction based on the real-time changing release source information, diffusion result information, and environmental factor information in the time and spatial dimensions of a nuclear accident, improving the accuracy of decision-making action prediction and providing dynamic decision support for scenarios with multiple affected bodies and dynamically changing decision events in nuclear accidents. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0043] Figure 1 This is a schematic diagram of a scenario illustrating the nuclear accident action decision-making method in the embodiments of this application;

[0044] Figure 2 This is a flowchart illustrating the nuclear accident action decision-making method in the embodiments of this application;

[0045] Figure 3 This is a flowchart illustrating the location information acquisition steps in an embodiment of this application;

[0046] Figure 4 This is a flowchart illustrating the training steps of another action decision model in an embodiment of this application;

[0047] Figure 5 This is a flowchart illustrating the steps for obtaining the optimal decision-making action of the disaster-affected sample at the sample time point in this embodiment of the application.

[0048] Figure 6 This is a schematic diagram of the structure of the nuclear accident action decision-making device in the embodiments of this application;

[0049] Figure 7 This is a schematic diagram of the structure of the computer device in the embodiments of this application. Detailed Implementation

[0050] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0051] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0052] In the description of this application, the word "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0053] In the embodiments of this application, it should also be noted that the nuclear accident action decision-making method provided in the embodiments of this application can be applied to, for example, Figure 1 The computer device 100 shown can be a terminal or a server. Specifically, nuclear accident simulation data can be input into the computer device 100, which uses the time points on the nuclear accident simulation timeline as the current prediction step time points to obtain the location information of the affected objects in the current prediction step. Then, based on the location information and the release source information, diffusion result information, and environmental factor information at the corresponding time point of the current prediction step, the action decision information for the current prediction step is obtained. Finally, based on the action decision information of all prediction steps, the action decision information of the affected objects in the nuclear accident is generated.

[0054] Those skilled in the art will understand that Figure 1 The application environment shown is merely one application scenario of the solution in this application and does not constitute a limitation on the application scenario of the solution in this application. Other application environments may include those that are more specific to this application. Figure 1 The number of computer devices shown is more or less, for example Figure 1 Only one computer device is shown; it is understood that this nuclear accident action decision-making system may also include one or more other servers, which are not specified here. Additionally, as... Figure 1 As shown, the nuclear accident action decision-making system may also include a memory for storing data, such as release source information, diffusion result information, and environmental factor information.

[0055] It should also be noted that, Figure 1The schematic diagram of the nuclear accident action decision-making system shown is merely an example. The nuclear accident action decision-making system and scenario described in the embodiments of the present invention are for the purpose of more clearly illustrating the technical solutions of the embodiments of the present invention, and do not constitute a limitation on the technical solutions provided by the embodiments of the present invention. As those skilled in the art will know, with the evolution of nuclear accident action decision-making systems and the emergence of new business scenarios, the technical solutions provided by the embodiments of the present invention are also applicable to similar technical problems.

[0056] See Figure 2 This application provides a nuclear accident action decision-making method, which is mainly applied to the above-mentioned... Figure 1 Taking a computer device as an example, the method includes steps S210 to S240, as follows:

[0057] S210, Acquire nuclear accident simulation data, which includes information on release sources, diffusion results, and environmental factors at different points in time on the nuclear accident simulation timeline.

[0058] Nuclear accident simulation data refers to source term-related data, environmental data, and pollutant migration and diffusion results at different time and spatial scales simulated based on source term-related data and environmental data after a nuclear accident. Examples include information on released nuclides and release source strength at the nuclear accident site, wind and meteorological fields within the evaluation area, and pollutant diffusion concentration and dose field results simulated using source term and meteorological information. Specifically, nuclear accident simulation data can be obtained through simulation calculations by nuclear accident consequence assessment and decision support systems (such as NACADOS and JRODOS).

[0059] The nuclear accident simulation data includes information on release sources, diffusion results, and environmental factors at different points in time on the nuclear accident simulation timeline. The nuclear accident simulation timeline refers to the timeline after a nuclear accident, in which different data are used to simulate and assess the migration and diffusion of pollutants in the environment. Release source information includes, but is not limited to, information on release source strength and release source nuclides. Diffusion result information refers to the information on the migration and diffusion results of pollutants simulated based on source term data and environmental data, including, but not limited to, information on concentration fields and dose fields corresponding to different geographic grids at each simulation time point. Environmental factor information includes information on ground pressure, wind direction, wind speed, topography, precipitation / snowfall, elevation, temperature, humidity, and traffic corresponding to different geographic grids.

[0060] It is understandable that the information at different time points includes information on the release source, the diffusion results, and environmental factors. For example, assuming the nuclear accident simulation timeline is 48 hours long and the time interval between time points is 1 hour, the nuclear accident simulation data includes release source information, diffusion results information, and environmental factor information at 1 hour, 2 hours, 3 hours, and so on, for a total of 48 time points.

[0061] S220: The time points on the nuclear accident simulation timeline are used sequentially as the time points of the current prediction step to obtain the location information of the disaster-affected object in the current prediction step.

[0062] The term "disaster victims" refers to the population in the area surrounding the nuclear accident site. Specifically, this can be achieved by dividing the population in the surrounding area into different disaster victims based on geographical location information. For example, people within the same geographical grid are identified as one disaster victim. It is understandable that the location information of different disaster victims will differ, and their distance from the nuclear accident site (i.e., the release point of the release source) will vary.

[0063] In this process, each prediction step corresponds one-to-one with a point in time on the nuclear accident simulation timeline. Within each prediction step, the decision-making actions taken by the affected entity at a given point in time are predicted. In the event of a nuclear accident, different decision-making actions should be taken for different stages of the accident, such as inaction, covert action, walking action, single-vehicle action, and car action, to reduce the expected dose received by the affected entity, avoid deterministic effects, and minimize stochastic effects. Therefore, the decision-making actions taken by the affected entity at each point in time can be predicted sequentially according to the nuclear accident simulation timeline.

[0064] Due to varying distances from the nuclear accident site, the diffusion outcomes and environmental factors of affected objects differ, influencing subsequent predictions of their actions and decisions. Therefore, after obtaining nuclear accident simulation data, the location information of affected objects at the current prediction step can be acquired first. Specifically, if the current prediction step is the first time point on the nuclear accident simulation timeline (i.e., the starting time point), the initial location information of the affected object is the location information at the current prediction step. If the current prediction step is any time point on the nuclear accident simulation timeline other than the first time point, the location information at the current prediction step can be determined based on the location information and decision-making actions of the affected object in previous prediction steps.

[0065] In one embodiment, obtaining the location information of the disaster-affected object in the current prediction step includes: obtaining the location information and action decision information of the disaster-affected object in the previous prediction step; and calculating the location information of the disaster-affected object in the current prediction step based on the location information and action decision information of the previous prediction step.

[0066] Here, the preceding prediction step is the prediction step before the current prediction step, and the time point corresponding to the preceding prediction step is the time point before the time point corresponding to the current prediction step. For example, if the time point corresponding to the current prediction step is time t, the time point corresponding to the preceding prediction step is time (t-1).

[0067] Among them, action decision information includes the decision actions taken by the disaster-affected object at the corresponding prediction step (time point), such as no action, covert action, walking action, bicycle action, and car action, and the location information that affects subsequent prediction steps.

[0068] For example, if the disaster-affected object's decision action in the previous prediction step was no action, then the location information of the disaster-affected object in the current prediction step is the same as that in the previous prediction step; if the disaster-affected object's decision action in the previous prediction step was walking, then the location information of the disaster-affected object in the current prediction step is different from that in the previous prediction step. The location information of the disaster-affected object in the current prediction step can be updated based on the action decision information and the location information of the disaster-affected object in the previous prediction step.

[0069] S230: Based on location information and release source information, diffusion result information and environmental factor information at the time point corresponding to the current prediction step, obtain action decision information for the current prediction step.

[0070] After obtaining the location information of the affected object in the current prediction step, the location information, along with the release source information, diffusion result information, and environmental factor information at the corresponding time point of the current prediction step, can be input into a pre-trained action decision model. The action decision model then predicts the action decision information for the current prediction step, such as the action the affected object should take in the current prediction step. Specifically, this action decision model is used to predict the action decision information of the affected object in the current prediction step based on the location information of the affected object and the release source information, diffusion result information, and environmental factor information at the corresponding time point of the current prediction step. The action decision model can be constructed based on supervised learning algorithms (such as deep neural network models, hidden Markov models, Bayesian network models, etc.) or reinforcement learning models.

[0071] Furthermore, as mentioned above, due to differences in distance from the nuclear accident site, the release source information, diffusion result information, and environmental factor information of the affected objects differ. Based on the location information of the affected object in the current prediction step, the target release source information, target diffusion result information, and target environmental factor information corresponding to that location information can be determined from the release source information, diffusion result information, and environmental factor information at the corresponding time. Then, based on the target release source information, target diffusion result information, and target environmental factor information, the action decision information for the current prediction step can be predicted. Specifically, in one embodiment, as... Figure 3 As shown, step S230 includes:

[0072] S310, Based on the location information, determine the target release source information, target diffusion result information, and target environmental factor information from the release source information, diffusion result information, and environmental factor information at the time point corresponding to the current prediction step.

[0073] Among them, the actual geographic space can be divided into geographic grids according to latitude and longitude, and the location information can specifically be the geographic grid where the disaster-affected object is located in the current prediction step.

[0074] The information includes the release source nuclides and release intensity at the corresponding time points; the environmental factor information includes ground pressure, wind direction, wind speed, topography, precipitation / snowfall, elevation, temperature, humidity, and traffic information corresponding to different geographic grids; and the diffusion result information includes the concentration field and dose field information corresponding to different geographic grids at the corresponding time points.

[0075] After obtaining the location information of the affected object, specific information such as the concentration field and dose field of the geographic grid corresponding to the location information can be obtained, as well as the target environmental factors such as ground pressure, wind direction, wind speed, topography, precipitation / snowfall, altitude, temperature, humidity and traffic information of the geographic grid, and the target release source information such as the source term nuclide information and source strength information of the release process.

[0076] S320: Input the target release source information, target diffusion result information, and target environmental factor information into the action decision model, and obtain the probability value corresponding to taking different decision actions in the current prediction step through the action decision model.

[0077] S330, determine the target decision action from different decision actions based on the probability value corresponding to each decision action.

[0078] In this process, once the target decision action is determined, it can be used as the action decision information for the current prediction step; alternatively, the expected dose of the target decision action can be calculated based on the target release source information, target diffusion result information, and target environmental factor information, and then the target decision action and the expected dose corresponding to the target decision action can be used as the action decision information for the current prediction step.

[0079] The action decision model can be a classifier built based on machine learning, such as a random forest model. It's understood that the action decision model here refers to a pre-trained random forest model.

[0080] Specifically, the target release source information, target diffusion result information, and target environmental factor information are input into the action decision model. The action decision model outputs the probability values ​​corresponding to different decision actions in the current prediction step. Taking a trained random forest model as an example, the target release source information, target diffusion result information, and target environmental factor information are input into the action decision model built based on the random forest model. Each decision tree in the random forest model votes on its classification result, i.e., the decision action, based on the decision actions output by each decision tree. Then, the final output result of the random forest model, i.e., the probability value corresponding to different decision actions, is determined based on the decision actions output by each decision tree.

[0081] After obtaining the probability values ​​corresponding to each decision action, the target decision action is determined from the different decision actions. For example, the decision action with the highest probability value can be determined as the target decision action.

[0082] By constructing an action decision-making model using a random forest model, dynamic decision-making and action prediction can be made based on real-time changes in release source information, diffusion outcome information, and environmental factor information in the time and space dimensions of a nuclear accident, thereby improving the accuracy of decision-making and action prediction.

[0083] S240, Based on the action decision information of all prediction steps, generate action decision information for the affected objects in a nuclear accident.

[0084] After obtaining action decision information for all time points on the nuclear accident simulation timeline, action decision information for the affected objects during the nuclear accident can be generated.

[0085] The aforementioned nuclear accident action decision-making method acquires nuclear accident simulation data, including release source information, diffusion result information, and environmental factor information at different time points on the nuclear accident simulation timeline. It then sequentially uses each time point on the nuclear accident simulation timeline as the current prediction step's time point to obtain the location information of the affected object at that current prediction step. Based on the location information and the release source information, diffusion result information, and environmental factor information at the corresponding time point of the current prediction step, it obtains action decision information for the current prediction step. Based on the action decision information from all prediction steps, it generates action decision information for the affected object in the nuclear accident. This enables dynamic decision-making and action prediction based on real-time changes in release source information, diffusion result information, and environmental factor information in the time and spatial dimensions of a nuclear accident, improving the accuracy of decision-making and action prediction and providing dynamic decision support for scenarios with multiple affected objects and dynamically changing decision events in nuclear accidents.

[0086] To predict the action decision information of disaster-affected objects in the current prediction step, it is necessary to pre-train the random forest model to obtain the trained action decision model. Specifically, in one embodiment, such as... Figure 4 As shown, before inputting the target release source information, target diffusion result information, and target environmental factor information into the action decision model, the following steps are also included:

[0087] S410, Obtain nuclear accident simulation samples, which include release source samples, diffusion result samples, and environmental factor samples at different time points on the nuclear accident simulation timeline.

[0088] Among them, nuclear accident simulation samples can be generated through a nuclear accident consequence assessment and decision support system.

[0089] Specifically, the nuclear accident simulation samples include release source samples, diffusion result samples, and environmental factor samples at different time points on the nuclear accident simulation timeline. Among them, the release source samples include, but are not limited to, release source strength and release source nuclides; the diffusion result samples include, but are not limited to, concentration fields and dose fields corresponding to different geographic grids at the corresponding time points; the environmental factor samples include, but are not limited to, surface pressure, wind direction, wind speed, topography, precipitation / snowfall, altitude, temperature, humidity, and traffic information corresponding to different geographic grids.

[0090] S420 uses the time points on the nuclear accident simulation timeline as sample time points to determine the optimal decision-making action for the affected body sample at each sample time point.

[0091] The optimal decision action refers to the action taken by the affected sample at the sample time point that minimizes the overall expected dose or maximizes the overall avoidable dose received by the affected sample during the entire nuclear accident process.

[0092] Specifically, by iterating through the possible combinations of different decision-making actions taken by the affected body sample at various time points on the nuclear accident simulation timeline, and calculating the total expected dose when taking different decision-making actions at various time points on the nuclear accident simulation timeline, the total expected dose can be replaced by dose values ​​such as the total individual effective dose, total avoidable dose, or total residual dose of the affected body sample during the accident process, depending on the different simulation systems.

[0093] Furthermore, the decision-making actions taken by the disaster-stricken objects at the sample time point have a certain impact on the expected dose at the current time point and subsequent time points, but are irrelevant to the expected dose at time points before the sample time point. Therefore, in one embodiment, such as Figure 5 As shown, determining the optimal decision-making action for the disaster-affected sample at the sample time point includes:

[0094] S510 is a combination of decision-making actions that traverses from the sample time point to the last time point in the nuclear accident simulation timeline, starting from the sample time point.

[0095] S520 calculates the total expected dose of each decision-making action combination for the affected body sample based on release source samples, diffusion result samples, and environmental factor samples at different time points.

[0096] S530, based on the total expected dose of each decision action combination, determine the optimal decision action combination from the decision action combination.

[0097] S540, determine the decision action at the sample time point in the optimal decision action combination as the optimal decision action at the sample time point.

[0098] Among them, the decision action combination includes the decision actions taken by the affected object at each time point from the sample time point to the last time point in the nuclear accident simulation timeline; traversing the decision action combination from the sample time point to the last time point in the nuclear accident simulation timeline means obtaining all possible decision actions taken by the affected object during the period from the sample time point to the last time point in the nuclear accident simulation timeline.

[0099] For example, suppose the decision actions include decision action A, decision action B, and decision action C. There are two time points in the nuclear accident simulation timeline, from the sample time point to the last time point in the nuclear accident simulation timeline. In this case, the combination of decision actions from the sample time point to the last time point in the nuclear accident simulation timeline includes {A, A}, {A, B}, {A, C}, {B, A}, {B, B}, {B, C}, {C, A}, {C, B}, and {C, C}. Taking {A, B} as an example, it refers to the disaster-affected sample taking decision action A at the sample time point and the disaster-affected sample taking decision action B at the next time point after the sample time point.

[0100] After obtaining all decision action combinations, for any given decision action combination, the total expected dose of the affected sample when taking that decision action combination can be calculated based on the release source samples, diffusion result samples, and environmental factor samples from all time points from the sample time point to the last time point in the nuclear accident simulation timeline.

[0101] It is understandable that different decision-making actions taken by a disaster-affected object at a sample time point often affect the object's location information at the corresponding subsequent time point. Therefore, different decision-making actions taken by a disaster-affected object at a sample time point not only affect the expected dose at the current sample time point but also the expected dose at subsequent time points. For example, in the above example, for decision-making action combinations {A, A} and {B, A}, although both take decision-making action B at the second time point, the expected dose of the disaster-affected object at the second time point may not be the same.

[0102] Further, in one embodiment, based on the release source samples, diffusion result samples, and environmental factor samples at different time points, the total expected dose of the affected body sample taking each combination of decision actions is calculated. Specifically, this includes: for the target decision action combination in the decision action combination, taking the sample time point to the last time point in the nuclear accident simulation timeline as the time point of the current prediction step, and obtaining the location information of the affected body sample in the current prediction step; determining the target decision action of the current prediction step from the target decision action combination, and calculating the expected dose corresponding to the affected body sample taking the target decision action in the current prediction step based on the location information and the release source information, diffusion result information, and environmental factor information at the time point corresponding to the current prediction step; and calculating the total expected dose of the affected body sample taking the target decision action combination based on the expected doses corresponding to all prediction steps.

[0103] The location information of the disaster-affected sample in the current prediction step can be the initial location information of the disaster-affected sample, or it can be calculated based on the location information and decision actions of the previous prediction steps. For example, if the current prediction step is the starting time point (i.e., the sample time point) of the disaster-affected sample on the nuclear accident simulation timeline, the initial location information of the disaster-affected sample is the location information of the current prediction step; if the current prediction step is a time point other than the starting time point of the nuclear accident simulation timeline, the location information of the disaster-affected object in the current prediction step can be determined based on the location information and decision actions of the disaster-affected object in the previous prediction steps.

[0104] After determining the location information of the affected body sample in the current prediction step, the nuclear accident consequences assessment and decision support system can be used to calculate the expected dose of the affected body sample when taking targeted decision actions in the current prediction step, based on this location information and the release source information, diffusion result information, and environmental factor information at the corresponding time point in the current prediction step. The expected dose can be expressed using dose values ​​such as the individual effective dose or avoidable dose of the affected body sample in the current prediction step.

[0105] After obtaining the expected doses corresponding to all prediction steps, the total expected dose of the target decision action combination for the disaster-affected sample is calculated based on the expected doses corresponding to all prediction steps. Specifically, the discount coefficient of each prediction step can be determined according to the order of each prediction step. The expected doses of each prediction step are weighted and summed according to the discount coefficients of each prediction step to obtain the total expected dose of the target decision action combination.

[0106] The total expected dose for the target decision action combination can be calculated using the following formula:

[0107] V = maximum(R) t +γR t+l +γ 2 R t+2 +γ 3 R t+3 +γ 4 R t+4 +…), γ∈(0,1)

[0108] Where V represents the total expected dose of the target decision action combination; R represents the individual value at different time points, i.e., the expected dose corresponding to different prediction steps; and γ is the discount rate, representing the discount coefficient of the individual value at different time points over time. For the decision actions taken by the disaster-affected objects at the sample time points, more attention is paid to their immediate individual value, while the individual value at subsequent time points is de-emphasized, in order to improve the accuracy of the optimal decision action at the sample time points.

[0109] After obtaining the total expected dose for each decision action combination, the decision action combination that minimizes the expected radiation dose to the affected objects can be determined based on the total expected dose, and this combination can be identified as the optimal decision action combination. Alternatively, the decision action combination that maximizes the avoidable dose to the affected objects can be determined based on the total expected dose, and this combination can also be identified as the optimal decision action combination. After determining the optimal decision action combination, the decision actions at the sample time points within the optimal decision action combination can be identified as the optimal decision actions at those sample time points.

[0110] S430: Based on the release source information, diffusion result information, environmental factor information, and optimal decision action at the sample time point, the random forest model is trained to obtain the action decision model.

[0111] Specifically, a training sample can be generated by combining the release source information, diffusion outcome information, environmental factor information, and optimal decision action at a given time point. Multiple training samples are obtained by repeating the above steps to build the training data for the random forest model. The training data includes samples of affected entities, release source samples at the given time point, diffusion outcome samples at the given time point, environmental factor samples at the given time point, and the optimal decision action of the affected entity samples at the given time point. Then, a random forest model is constructed and trained based on the training samples to obtain the action decision model.

[0112] The action decision model is constructed based on the random forest model, which makes the action decision model more generalizable, avoids overfitting during training, and can obtain the correlation and importance of information features such as release source information, diffusion result information, and environmental factor information.

[0113] The following section provides further explanation of nuclear accident action decision-making methods in the context of application scenarios:

[0114] (1) The process of making decisions on the nuclear accident is analyzed by using object-oriented analysis methods, and all possible issues that may require decision-making are traversed and a list of the problem domains of the entire event is established.

[0115] Taking action decision-making problems as an example, for action decisions where the disaster-affected population is the public, the decision-making problems include taking no action, covert action, walking action, bicycle action, and car action. More specifically, car action can be further subdivided into fast car action, medium-speed car action, and slow car action, to be applied to action decisions in different congestion scenarios (based on the population size of the disaster-affected population).

[0116] (2) A large number of nuclear accident emergency cases were generated by NACADOS system, and machine learning training datasets and validation datasets were constructed based on the nuclear accident simulation data in the nuclear accident emergency cases.

[0117] Specifically, a scenario response mechanism (i.e., a scenario library) is established based on the nuclear accident process on the simulated nuclear accident timeline. Each scenario library includes the hazard source, the affected entity, the external environment, and nuclear accident emergency response actions. Each scenario in the scenario library corresponds to one or more training data sets, which include information on the release source, diffusion outcomes, environmental factors, and optimal decision actions.

[0118] Specifically, the hazard-causing body corresponds to the release source information and diffusion result information, the hazard-bearing body corresponds to the hazard-bearing object, the external environment corresponds to environmental factor information, and the nuclear accident emergency response corresponds to the optimal decision action. Among them, the release source information includes the release source nuclide and source strength, the diffusion result information includes the concentration field and dose field, and the environmental factor information includes topographic information, population information, transportation information, key meteorological elements, location information of points of concern, information on concealable buildings, speed of movement information, and other relevant parameter information.

[0119] (3) Statistical analysis is performed based on the time steps in the scenario database to obtain the decision reference point at the time when countermeasures should be taken.

[0120] The decision reference point refers to a disaster-affected object at a certain location at the point in time when countermeasures should be taken.

[0121] (4) Obtain the decision-making problem at the decision reference point at that time, such as taking no action, covert action, walking action, bicycle action, and car action.

[0122] (5) Construct a random forest model based on the decision problem and train the input parameters release source information, diffusion result information, and environmental factor information, such as the concentration field, dose field, wind field, precipitation, terrain information, release source strength, release source nuclide, population information, traffic conditions (fuzzy value), building conditions that can be concealed (fuzzy value), and movement speed of the decision reference point at this time.

[0123] The machine learning training of random forest is used to obtain the decision classification probability results of the decision reference point, and the decision utility of different decision actions is obtained. The decision utility predicted after training is transformed into the decision individual value, which can be represented by the expected dose of the disaster object.

[0124] The decision-making value of the current decision reference point at each time step in the entire nuclear emergency process is iteratively cycled, and the total value and the intelligent optimization decision result of the reference point are obtained by statistically analyzing all decision-making values ​​according to different time weights.

[0125] (6) Then, using the same logic, perform the above steps (4) and (5) on the next decision reference point at the same time point. After iteration, the intelligent optimization decision result of the nuclear event can be obtained.

[0126] To better implement the nuclear accident action decision-making method provided in the embodiments of this application, based on the nuclear accident action decision-making method proposed in the embodiments of this application, this application also provides a nuclear accident action decision-making device, such as... Figure 6 As shown, the nuclear accident action decision-making device 600 includes:

[0127] The simulation data acquisition module 610 is used to acquire nuclear accident simulation data, which includes release source information, diffusion result information, and environmental factor information at different time points on the nuclear accident simulation timeline.

[0128] The location information determination module 620 is used to sequentially use the time points on the nuclear accident simulation timeline as the time points of the current prediction step to obtain the location information of the disaster-affected object in the current prediction step.

[0129] The action decision prediction module 630 is used to obtain action decision information for the current prediction step based on location information and release source information, diffusion result information and environmental factor information at the time point corresponding to the current prediction step.

[0130] The action decision acquisition module 640 is used to generate action decision information for the affected objects in a nuclear accident based on the action decision information of all prediction steps.

[0131] In some embodiments of this application, the location information determination module is specifically used to obtain the location information and action decision information of the disaster-affected object in the previous prediction step; and to calculate the location information of the disaster-affected object in the current prediction step based on the location information and action decision information of the previous prediction step.

[0132] In some embodiments of this application, the action decision prediction module is specifically used to determine the target release source information, target diffusion result information, and target environmental factor information from the release source information, diffusion result information, and environmental factor information at the time point corresponding to the current prediction step, based on the location information; input the target release source information, target diffusion result information, and target environmental factor information into the action decision model; obtain the probability values ​​corresponding to taking different decision actions in the current prediction step through the action decision model; and determine the target decision action from different decision actions based on the probability values ​​corresponding to each decision action.

[0133] In some embodiments of this application, the nuclear accident action decision-making device 500 further includes a model training module. The model training module is used to acquire nuclear accident simulation samples, which include release source samples, diffusion result samples, and environmental factor samples at different time points on the nuclear accident simulation timeline. The time points on the nuclear accident simulation timeline are used as sample time points in sequence to determine the optimal decision action of the affected body sample at the sample time point. Based on the release source information, diffusion result information, environmental factor information, and optimal decision action at the sample time point, the random forest model is trained to obtain the action decision-making model.

[0134] In some embodiments of this application, the model training module is further configured to: take the sample time point as the starting time point, traverse the decision action combinations from the sample time point to the last time point in the nuclear accident simulation timeline; calculate the total expected dose of each decision action combination for the affected body sample based on the release source sample, diffusion result sample, and environmental factor sample at different time points; determine the optimal decision action combination based on the total expected dose of each decision action combination; and determine the decision action at the sample time point in the optimal decision action combination as the optimal decision action at the sample time point.

[0135] In some embodiments of this application, the model training module is further configured to, for the target decision action combination in the decision action combination, sequentially use the sample time point to the last time point in the nuclear accident simulation timeline as the time point of the current prediction step, obtain the location information of the disaster-affected sample in the current prediction step; determine the target decision action of the current prediction step from the target decision action combination; calculate the expected dose corresponding to the disaster-affected sample when taking the target decision action in the current prediction step based on the location information and the release source information, diffusion result information and environmental factor information at the time point corresponding to the current prediction step; and calculate the total expected dose of the disaster-affected sample when taking the target decision action combination based on the expected doses corresponding to all prediction steps.

[0136] In some embodiments of this application, the model training module is further configured to determine the discount coefficient of each prediction step according to the order of each prediction step; and to perform a weighted summation of the expected doses of each prediction step according to the discount coefficients of each prediction step to obtain the total expected dose of the target decision action combination.

[0137] The aforementioned nuclear accident action decision-making device acquires nuclear accident simulation data, including release source information, diffusion result information, and environmental factor information at different time points on the nuclear accident simulation timeline. It sequentially uses these time points as the current prediction step's time point to obtain the location information of the affected object at the current prediction step. Based on the location information and the release source information, diffusion result information, and environmental factor information at the corresponding time point of the current prediction step, it obtains action decision information for the current prediction step. Based on the action decision information from all prediction steps, it generates action decision information for the affected object in the nuclear accident. This enables dynamic decision-making action prediction based on real-time changes in release source information, diffusion result information, and environmental factor information in the time and spatial dimensions of a nuclear accident, improving the accuracy of decision-making action prediction and providing dynamic decision support for scenarios with multiple affected bodies and dynamically changing decision events in nuclear accidents.

[0138] In some embodiments of this application, the nuclear accident action decision-making device 600 can be implemented as a computer program, which can be implemented in, for example... Figure 7 The computer device shown runs on this system. The computer device's memory can store the various program modules that make up the nuclear accident action decision-making device 600, for example, Figure 6 The simulation data acquisition module 610, location information determination module 620, action decision prediction module 630, and action decision acquisition module 640 are shown. The computer program comprised of these modules causes the processor to execute the steps in the nuclear accident action decision-making methods of the various embodiments of this application described in this specification.

[0139] For example, Figure 7 The computer device shown can be used as follows Figure 6 The simulation data acquisition module 610 in the nuclear accident action decision-making device 600 shown executes step S210. The computer device can execute step S220 via the location information determination module 620. The computer device can execute step S230 via the action decision prediction module 630. The computer device can execute step S240 via the action decision acquisition module 640. The computer device includes a processor, memory, and a network interface connected via a system bus. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used to communicate with external computer devices via a network connection. When the computer program is executed by the processor, it implements a nuclear accident action decision-making method.

[0140] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0141] In some embodiments of this application, a computer device is provided, including one or more processors; a memory; and one or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the processors as described in the nuclear accident action decision-making method. The steps of the nuclear accident action decision-making method here may be steps from the nuclear accident action decision-making methods of the various embodiments described above.

[0142] In some embodiments of this application, a computer-readable storage medium is provided, storing a computer program that is loaded by a processor, causing the processor to execute the steps of the above-described nuclear accident action decision-making method. The steps of the nuclear accident action decision-making method here can be the steps in the nuclear accident action decision-making methods of the various embodiments described above.

[0143] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0144] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0145] The foregoing has provided a detailed description of a nuclear accident action decision-making method, apparatus, computer device, and storage medium provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method of decision making for nuclear accident action, characterized by, The method includes: Acquire nuclear accident simulation data, which includes release source information, diffusion result information, and environmental factor information at different simulation time points on the nuclear accident simulation timeline; wherein, the nuclear accident simulation data includes: source term related data, environmental related data, and pollutant migration and diffusion result data simulated based on the source term related data and the environmental related data; the diffusion result information includes concentration field and dose field information corresponding to different geographic grids at each simulation time point; the nuclear accident simulation data is obtained by simulation calculation by a nuclear accident consequence assessment and decision support system; the nuclear accident consequence assessment and decision support system includes at least one of the following: NACADOS, JRODOS; The time points on the nuclear accident simulation timeline are sequentially used as the time points of the current prediction step, and the population in the area surrounding the nuclear accident site is divided according to geographical location information to determine different disaster-affected objects; among them, the population in the same geographical grid is determined as one disaster-affected object; the nuclear accident simulation timeline refers to the timeline of simulation assessment of the migration and diffusion of pollutants in the environment based on different data after the nuclear accident. Obtain the location information of the disaster-affected object in the current prediction step; wherein, if the current prediction step corresponds to the first time point on the nuclear accident simulation timeline, then determine the initial location information of the disaster-affected object as the location information of the current prediction step; if the current prediction step corresponds to a time point other than the first time point, then calculate the location information of the disaster-affected object in the current prediction step based on the location information of the disaster-affected object in the previous prediction step and the action decision information of the previous prediction step; Based on the location information, target release source information, target diffusion result information, and target environmental factor information corresponding to the location information are determined from the release source information, diffusion result information, and environmental factor information at the time point corresponding to the current prediction step. The target release source information, target diffusion result information, and target environmental factor information are input into the action decision model, and the action decision model outputs the probability values ​​corresponding to different decision actions in the current prediction step. Based on the probability values ​​corresponding to each decision action, a target decision action is selected from different decision actions. The action decision model is trained on a random forest model based on nuclear accident simulation samples and the optimal decision actions corresponding to the sample time points. The nuclear accident simulation samples include release source samples, diffusion result samples, and environmental factor samples at different time points on the nuclear accident simulation timeline. Based on the action decision information from all prediction steps, action decision information for the affected object in the nuclear accident is generated.

2. The method of claim 1, wherein, The release source information includes release source term nuclide information and release source strength information.

3. The method of claim 1, wherein, The location information is the geographic grid where the disaster-affected object is located, and the geographic grid is divided into actual geographic spaces according to latitude and longitude; the decision-making actions include at least no action, covert action, walking action, single vehicle action, and car action.

4. The method of claim 1, wherein, Before inputting the target release source information, the target diffusion result information, and the target environmental factor information into the action decision model, the method further includes: Obtain nuclear accident simulation samples, which include release source samples, diffusion result samples, and environmental factor samples at different time points on the nuclear accident simulation timeline; The time points on the nuclear accident simulation timeline are used sequentially as sample time points to determine the optimal decision-making action of the affected body sample at the sample time points. Based on the release source information, diffusion result information, environmental factor information, and optimal decision action at the sample time points, the random forest model is trained to obtain the action decision model; The action decision model is either a random forest classifier built based on machine learning or an action decision model built based on reinforcement learning. After inputting the target release source information, the target diffusion result information, and the target environmental factor information into the action decision model, the model outputs the probability values ​​corresponding to different decision actions in the current prediction step, and determines the target decision action based on the probability values ​​corresponding to each decision action.

5. The method of claim 4, wherein, The optimal decision-making action for determining the disaster-affected sample at the sample's time point includes: Starting from the sample time point, the decision-action combinations are iterated from the sample time point to the last time point in the nuclear accident simulation timeline. Based on the release source samples, diffusion result samples, and environmental factor samples at different time points, calculate the total expected dose of the disaster-affected body sample for each of the aforementioned decision action combinations; Based on the total expected dose of each of the aforementioned decision action combinations, the optimal decision action combination is determined from the decision action combinations; The decision action at the sample time point in the optimal decision action combination is then determined as the optimal decision action at that sample time point.

6. The method of claim 5, wherein, The calculation of the total expected dose of each of the aforementioned decision action combinations for the disaster-affected sample includes: The expected dose of the disaster-affected sample at each prediction step is determined, and the expected doses at each prediction step are accumulated to obtain the total expected dose.

7. The method of claim 6, wherein, The step of accumulating the expected doses corresponding to each prediction step to obtain the total expected dose includes: The discount factor for each prediction step is determined according to the order of the prediction steps. The total expected dose is obtained by weighting and summing the expected doses of each prediction step according to the discount factor of each prediction step.

8. A nuclear accident action decision device characterized by comprising: The device includes: The simulation data acquisition module is used to acquire nuclear accident simulation data, which includes release source information, diffusion result information, and environmental factor information at different simulation time points on the nuclear accident simulation timeline. The nuclear accident simulation data includes source term-related data, environmental-related data, and pollutant migration and diffusion result data simulated based on the source term-related data and the environmental-related data. The diffusion result information includes concentration field and dose field information corresponding to different geographic grids at each simulation time point. The nuclear accident simulation data is obtained through simulation calculations by a nuclear accident consequence assessment and decision support system. The nuclear accident consequence assessment and decision support system includes at least one of the following: NACADOS and JRODOS. The location information determination module is used to sequentially use the time points on the nuclear accident simulation timeline as the time points of the current prediction step, and to divide the population in the area surrounding the nuclear accident site according to the geographical location information, and determine them as different disaster-affected objects; the population in the same geographical grid is determined as one disaster-affected object; the nuclear accident simulation timeline refers to the timeline of simulation assessment of the migration and diffusion of pollutants in the environment based on different data after the nuclear accident. The location information determination module is further configured to obtain the location information of the disaster-affected object in the current prediction step; wherein, if the current prediction step corresponds to the first time point on the nuclear accident simulation timeline, the initial location information of the disaster-affected object is determined as the location information of the current prediction step; if the current prediction step corresponds to a time point other than the first time point, the location information of the disaster-affected object in the current prediction step is calculated based on the location information of the disaster-affected object in the previous prediction step and the action decision information of the previous prediction step. The action decision prediction module is used to determine, based on the location information, the target release source information, target diffusion result information, and target environmental factor information corresponding to the location information from the release source information, diffusion result information, and environmental factor information at the time point corresponding to the current prediction step; input the target release source information, the target diffusion result information, and the target environmental factor information into the action decision model, and output the probability values ​​corresponding to different decision actions for the current prediction step through the action decision model; select the target decision action from different decision actions based on the probability values ​​corresponding to each decision action; wherein, the action decision model is obtained by training a random forest model based on nuclear accident simulation samples and the optimal decision actions corresponding to the sample time points, and the nuclear accident simulation samples include release source samples, diffusion result samples, and environmental factor samples at different time points on the nuclear accident simulation timeline; The action decision acquisition module is used to generate action decision information for the affected object in the nuclear accident based on the action decision information of all prediction steps.

9. A computer device, comprising: The computer device includes: One or more processors; The memory; and one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the processor to implement the nuclear accident action decision method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program, which is loaded by a processor to execute the steps in the nuclear accident action decision-making method according to any one of claims 1 to 7.

Images