Shape operation information display method and system based on multi-source information fusion

Abstract

The invention discloses a weather modification operation information display method and system based on multi-source information fusion, and relates to the technical field of meteorological information intelligent decision making, and the method comprises the steps: carrying out the reasoning analysis of a time sequence knowledge graph through a graph neural network, recognizing the cloud system development change, reasoning the evolution of cloud physical characteristics, and predicting the space-time range of an operation potential region, generating an accurate forecast conclusion and a figure operation suggestion; and converting the accurate forecast conclusion and the weather modification operation suggestion into a visual instruction set, and generating a visual comprehensive situation map through highlighting core weather indexes, displaying evolution paths by dynamic arrows and marking operation potential areas and operation corridors by color coverage. According to the method, reasoning analysis is carried out on the time sequence knowledge graph by using the graph neural network, the operation potential area is automatically identified from a historical mode, the dynamic evolution of the operation potential area is predicted, and the perspectiveness and the accuracy of figure operation decision making are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of meteorological information intelligent decision-making, and in particular to a human shadow operation information display method and system based on multi-source information fusion. BACKGROUND

[0002] Human shadow operation is a complex decision-making activity that highly depends on accurate meteorological information support. The current information display method in the field is mainly based on the integrated display of multi-source meteorological data. Typically, satellite, radar, and ground station observation data are fused with numerical prediction products through data assimilation technology, and are presented in a layered superimposed manner on a geographic information system platform. This method aims to provide a comprehensive meteorological situation background for command personnel to assist them in judging operation potential and safety window.

[0003] However, the existing method still has limitations in supporting the dynamic decision-making needs of human shadow operation. It focuses on the static display of meteorological elements, lacks automated identification and structured representation of cloud evolution and triggering mechanisms, resulting in a high dependence on human experience for judging operation timing and location. It also lacks quantitative analysis capabilities for causal relationships between information, making it difficult to automatically infer weather process evolution trends and operation effects, which restricts the intelligent level of operation scheme generation and evaluation. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a human shadow operation information display method based on multi-source information fusion to solve the problem of insufficient key event chain identification and dynamic operation scheme intelligent reasoning capabilities.

[0006] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the present application provides a human shadow operation information display method based on multi-source information fusion, which includes collecting and preprocessing multi-source meteorological observation data, and repairing and super-resolution reconstructing defective data through a generative adversarial network model to generate a multi-source time series dataset. Time series event extraction is performed on the multi-source time series dataset to identify cloud evolution, thermal and dynamic triggering, cloud system development, and operation action events related to human shadow operation as nodes, and a time series knowledge graph covering before operation to after operation is constructed. Graph neural network is used to analyze and reason the time series knowledge graph, identify cloud system development changes, and reason the evolution of cloud physical characteristics, predict the spatiotemporal range of operation potential area, generate accurate prediction conclusions and human shadow operation suggestions, and convert the accurate prediction conclusions and human shadow operation suggestions into a visual instruction set. The visual instruction set is used to highlight core weather indicators, dynamically display evolution paths with arrows, and mark operation potential areas and operation corridors with color overlays to generate a visual comprehensive situation map.

[0007] As a preferred scheme of the human shadow operation information display method based on multi-source information fusion, the preprocessing includes outlier rejection, format standardization, coordinate unification and space-time interpolation.

[0008] As a preferred scheme of the human shadow operation information display method based on multi-source information fusion, the multi-source time series data set is generated by the following steps, Defective data in the preprocessed multi-source meteorological observation data are identified and input into the adversarial network model for repair and super-resolution reconstruction to obtain a repaired and enhanced data set. The repaired and enhanced data set is fused with the effective data in the preprocessed multi-source meteorological observation data to generate a multi-source time series data set.

[0009] As a preferred scheme of the human shadow operation information display method based on multi-source information fusion, the multi-source time series data set is subjected to time series event extraction by the following steps, Variable data related to cloud microphysics, thermal dynamics and operation records are extracted from the multi-source time series data set, and entities, relationships and constraint rules in the human shadow operation knowledge ontology are loaded to generate a physical variable time series data set of operation. The physical variable time series data set of operation is subjected to event detection using a sliding window and a mutation point detection algorithm to identify cloud water content mutation points, energy jump points and wind shear inflection points, and a candidate segment set of operation decision events is generated. The candidate segment set of operation decision events is matched and fused with event patterns in the human shadow operation knowledge ontology to form a human shadow operation event instance set.

[0010] As a preferred scheme of the human shadow operation information display method based on multi-source information fusion, the time series knowledge graph covering the period from before operation to after operation is constructed by the following steps, Node types are defined for each human shadow operation event instance in the human shadow operation event instance set and time and space attributes are assigned to generate a time series knowledge graph node set. Based on the time series knowledge graph node set, time series relationships between nodes are mined, and operation rules are inferred to determine causal relationships to generate a candidate set of human shadow operation event relationships. The human shadow operation event instance nodes are taken as vertices, and the candidate set of human shadow operation event relationships is taken as edges, and a time axis and a spatial coordinate are embedded to construct a time series knowledge graph covering the period from before operation to after operation.

[0011] As a preferred scheme of the human shadow operation information display method based on multi-source information fusion, the cloud system development and change is identified by the following steps, The node attributes, edge relationships and timestamps in the time sequence knowledge graph are encoded and aligned to be converted into a time sequence graph tensor object; The time sequence graph tensor object is input into a graph neural network to obtain a low-dimensional vector representation of the node through forward propagation calculation; The attention weight of the graph neural network is analyzed, and the high-weight path subgraph connecting the thermal-dynamic trigger event and the workable cloud system-catalytic event is identified to obtain an evolution subgraph set.

[0012] As a preferred scheme of the human shadow operation information display method based on multi-source information fusion, the steps of generating the accurate prediction conclusion and the human shadow operation suggestion are as follows, The cloud system development evolution rule, the operation window threshold and the corridor constraint are decoded from the low-dimensional vector representation of the node in the evolution subgraph set to generate a fusion quantization rule set; According to the fusion quantization rule set, the end node representing the dangerous weather event is spatiotemporally extrapolated to obtain a work potential area spatiotemporal range, and the end node representing the workable cloud system is spatiotemporally extrapolated to obtain a work window time interval and a work corridor geographical range; The work window time interval and the work corridor geographical range are fused to generate the human shadow operation suggestion; According to the preset prediction level standard, the work potential area spatiotemporal range is matched with the rule and the severity is evaluated to generate the accurate prediction conclusion.

[0013] As a preferred scheme of the human shadow operation information display method based on multi-source information fusion, the steps of converting the accurate prediction conclusion and the human shadow operation suggestion into a visual instruction set are as follows, The work target cloud system index parameter, the evolution path and the work potential area polygon coordinates are extracted from the accurate prediction conclusion, and the visual element set is generated in combination with the human shadow operation suggestion; According to the preset visualization rule, the visual element set is mapped into specific commands executable by the graphic rendering engine to generate the visual instruction set.

[0014] As a preferred scheme of the human shadow operation information display method based on multi-source information fusion, the steps of generating the visual comprehensive situation map are as follows, The visual instruction set is executed to drive the graphic rendering engine to respectively create a work potential area color overlay layer, an evolution path dynamic arrow layer, an index parameter highlight layer, a work corridor semi-transparent pipeline layer and a work window countdown marker layer to generate a visual layer set; The data matching the work potential area spatiotemporal range is intercepted from the multi-source time sequence data set to render a multi-source data background layer; The visualization layer set is superimposed and fused with the multi-source data background layer to generate a visual comprehensive situation map.

[0015] In a second aspect, the present application provides a human shadow operation information display system based on multi-source information fusion, comprising a data acquisition and repair module, a graph construction module, an inference analysis module, and a visual display module. The data acquisition and repair module is configured to acquire multi-source meteorological observation data for preprocessing, and repair and super-resolution reconstruct defective data through an adversarial network model to generate a multi-source time series dataset. The graph construction module is configured to extract time series events from the multi-source time series dataset, identify cloud body evolution, thermal and dynamic triggering, cloud system development, and operation action events related to human shadow operation as nodes, and construct a time series knowledge graph covering before operation to after operation. The inference analysis module is configured to use a graph neural network to perform inference analysis on the time series knowledge graph, identify cloud system development and change, and infer the evolution of cloud physical characteristics, predict the spatiotemporal range of the operation potential area, and generate accurate prediction conclusions and human shadow operation suggestions. The visual display module is configured to convert the accurate prediction conclusions and human shadow operation suggestions into a visual instruction set, highlight the core weather indicators, dynamically display the evolution path through arrows, and mark the operation potential area and operation corridor through color coverage, and generate a visual comprehensive situation map.

[0016] The present application has the following advantages: by constructing a time series knowledge graph covering before operation to after operation, the weather evolution process and human shadow operation are fully chain-semanticized, providing an interpretable data basis for subsequent intelligent inference; by using a graph neural network to perform inference analysis on the time series knowledge graph, the operation potential area is automatically identified from historical patterns and its dynamic evolution is predicted, improving the forward-looking and accuracy of human shadow operation decision-making. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0018] Fig. 1 Flowchart of the human shadow operation information display method based on multi-source information fusion.

[0019] Fig. 2 Schematic diagram of the human shadow operation information display system based on multi-source information fusion.

[0020] Fig. 3A flowchart for generating a precise forecast conclusion.

[0021] Fig. 4 A flowchart for generating a visualized comprehensive situation map. DETAILED DESCRIPTION

[0022] In order to make the above objectives, features and advantages of the present application more obvious and understandable, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0023] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, so the present application is not limited to the specific embodiments disclosed below.

[0024] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is independent of or mutually exclusive with other embodiments.

[0025] REFERENCE Figs. 1-4 For one embodiment of the present application, the embodiment provides a human shadow operation information display method based on multi-source information fusion, comprising the following steps: S1, collecting multi-source meteorological observation data for preprocessing, and repairing and super-resolution reconstruction of defective data through a generative adversarial network model to generate a multi-source time series dataset; The multi-source meteorological observation data includes multi-channel cloud images, brightness temperature, cloud top temperature, precipitation, air pressure, wind speed, wind direction and ground temperature and humidity. It should be noted that the multi-channel cloud image is directly obtained by a plurality of spectral channel sensors of a meteorological satellite, and is used to display water vapor, cloud particles and surface features at different height layers; the brightness temperature is obtained by converting the radiation value detected by the satellite infrared and microwave channels through the Planck formula, and represents the radiation temperature of the cloud top or the ground at a specific waveband; the cloud top temperature is obtained by processing the brightness temperature data of the meteorological satellite infrared channel; the precipitation is directly measured by a ground rain gauge or obtained by inversion of weather radar reflectivity factor; the air pressure is directly measured by an air pressure sensor (such as a silicon capacitive air box barometer) in a ground meteorological station; the wind speed is directly measured by a wind speed sensor (such as an ultrasonic wind speed meter) of a ground meteorological station; the wind direction is directly measured by a wind vane of a ground meteorological station; and the ground temperature and humidity are directly measured by a temperature and humidity sensor of a ground meteorological station in a weather vane.

[0026] The preprocessing includes outlier rejection, format standardization, coordinate unification and spatio-temporal interpolation. It should be noted that the outlier rejection refers to filtering the multi-source meteorological observation data using statistical methods (such as the 3σ criterion) to remove error data points that deviate significantly from the reasonable range; the format standardization refers to converting the multi-source meteorological observation data into the same data format (such as NetCDF) to ensure consistent data field definitions and units; the coordinate unification refers to converting the spatial reference information of different data sources in the multi-source meteorological observation data to a unified geographic coordinate system (such as WGS-84) and projection method (such as Lambert projection) to align the spatial positions; the spatio-temporal interpolation refers to interpolating the time series data with inconsistent sampling frequencies and spatial positions in the multi-source meteorological observation data to a unified spatio-temporal grid using the Kriging method.

[0027] Identify the defective data in the pre-processed multi-source meteorological observation data and input it into the adversarial network model for repair and super-resolution reconstruction to obtain the repaired and enhanced data set; Further, the pre-processed multi-source meteorological observation data is scanned pixel by pixel, and the sequence points with continuous missing exceeding a certain time length, the values exceeding the reasonable physical range (such as negative precipitation), and the abnormal points with sharp mutations with the surrounding data are marked as defective data; the multi-source meteorological observation data containing defective data and the valid data in the pre-processed multi-source meteorological observation data are input into the adversarial network model, the generator in the adversarial network model generates repair content according to the position of the defective data and combines the context information of the surrounding valid data, and simultaneously up-samples the data to a higher spatial resolution; the discriminator distinguishes the repaired and reconstructed data from the real complete high-resolution data, and through multiple iterations of the generator and the discriminator, outputs the repaired and enhanced data set that conforms to the physical law.

[0028] It should be noted that the pre-trained adversarial network model, in specific operations, uses complete and high-quality multi-source meteorological observation data in historical periods as training samples, randomly generates multiple types of defect patterns (such as random point missing, block missing, and noise addition) in the training samples to simulate real data defects, inputs the training samples with simulated defects into the generator of the adversarial network model, and the generator outputs the repaired data; the repaired data and the original complete high-resolution training samples are input into the discriminator of the adversarial network model, the discriminator learns to distinguish between real data and generated data, through repeated iterative adversarial process, minimizes the reconstruction error of the generated data and maximizes the discrimination ability of the discriminator, until the adversarial network model can stably output high-quality repair results, thereby completing the pre-training of the adversarial network model.

[0029] Fuse the repaired and enhanced data set with the valid data in the pre-processed multi-source meteorological observation data to generate a multi-source time series data set; Further, in the unified space-time grid framework, the effective data in the repaired enhancement dataset and the preprocessed multi-source meteorological observation data are accurately matched in space-time, ensuring that the latitude and longitude coordinates and timestamps of each data are completely aligned; a weighted fusion strategy based on data source quality and space-time integrity is adopted, for the grid points with valid observations in the preprocessed multi-source meteorological observation data, the values in the preprocessed multi-source meteorological observation data are preferentially retained, and for the grid points marked as defective and repaired by the adversarial network model, the interpolation values in the repaired enhancement dataset are adopted; all effective values from the repaired enhancement dataset and the preprocessed multi-source meteorological observation data are combined to generate a multi-source time series dataset.

[0030] S2, time series event extraction is performed on the multi-source time series dataset, cloud evolution, thermal and dynamic triggering, cloud system development and operation event related to human shadow operation are identified as nodes, and a time series knowledge graph covering before operation to after operation is constructed; From the multi-source time series dataset, variable data related to cloud microphysics, thermal dynamics and operation records are extracted, and entity, relationship and constraint rules in the human shadow operation knowledge ontology are loaded to generate an operation physical variable time series dataset; Further, according to the entity attributes defined in the human shadow operation knowledge ontology, corresponding physical variables are screened out from the multi-source time series dataset, including cloud microphysics (such as cloud water content, cloud ice content), thermal dynamics (such as convective available potential energy, wind shear) and operation records (such as catalyst spreading amount, operation platform position); the entity types (such as operable cloud system, operation window), entity relationships (such as "has operation condition", "belongs to operation type") and constraint rules (such as temperature threshold for catalytic operation, atmospheric stability condition for safe operation) defined in the human shadow operation knowledge ontology are loaded; the screened physical variables are mapped and associated with the entity types, entity relationships and physical constraint rules defined in the human shadow operation knowledge ontology, ensuring that each physical variable is given a clear meteorological meaning and operation context; the physical variables with operation semantic annotations that have completed mapping and association are integrated to generate an operation physical variable time series dataset containing cloud water content time series data, energy time series data (convective available potential energy) and wind shear time series data.

[0031] It should be noted that the human shadow operation knowledge ontology defines the core concepts, concept attributes, concept relationships and constraint rules in the field of weather modification; the constraint rules are logical judgment conditions (such as trigger conditions, safety thresholds and operation specifications) based on meteorological principles and operation experience; the temperature threshold refers to the critical value of cloud top temperature or cloud temperature required for catalytic operation, which is set based on the temperature condition required for the existence of supercooled water in cloud microphysical process, and the exemplary value range is ≤-4℃, in order to ensure that the catalytic operation effectively initiates ice crystal growth in a suitable thermal stratification; the atmospheric stability condition for safe operation refers to the maximum allowable wind speed to ensure the safety of human shadow operation, which is set based on the stability requirements of operation platform (such as aircraft, rocket) and the precision requirements of catalyst scattering, and the exemplary value range is KI<32, in order to prevent strong convective weather from affecting the safety of operation.

[0032] The event detection algorithm is used to detect the event detection algorithm, identify the cloud water content mutation point, energy jump point and wind shear inflection point, and generate the operation decision event candidate segment set; Further, a fixed time length sliding window is set, and the sliding window is sequentially moved along the time axis of the operation physical variable time series data set. The cloud water content time series data, energy time series data and vertical wind shear time series data in each sliding window are respectively applied to the mutation point detection algorithm to identify the points where the cloud water content value, energy value and wind shear value occur mutation, jump and trend turning point. Each identified mutation point, jump point and inflection point is taken as the center, and a fixed time interval is extended forward and backward to form an event segment. All identified event segments with cloud water content mutation point, energy jump point and wind shear inflection point as the core are collected to form the operation decision event candidate segment set.

[0033] The operation decision event candidate segment set is matched and fused with the event mode in the human shadow operation knowledge ontology, and the human shadow operation event instance set is clustered. Further, the event segment set of each job decision event candidate segment containing cloud water content mutation points, energy jump points and wind shear inflection points is matched with the event mode predefined in the human shadow job knowledge ontology in terms of semantics and space-time, to determine whether the characteristics of the event segment meet the triggering conditions of a specific event mode: when the cloud water content mutation point value in the event segment exceeds the cloud water content threshold, the energy jump point value exceeds the convective available potential energy threshold, and the value at the wind shear inflection point also exceeds the vertical wind shear threshold, it is determined that the event segment meets the triggering conditions of the specific event mode “with job catalysis conditions” defined in the human shadow job knowledge ontology; for multiple event segments that are close in space-time and meet the same event mode or multiple event modes, clustering is performed based on the continuity of the time stamp and the proximity of the spatial position, multiple event segments describing the same physical process are fused into a composite event and assigned a unique identifier, while recording the start and end time, spatial range and event segment type, to generate the human shadow job event instance set.

[0034] It should be noted that the event mode is a standard template representing a typical weather process or job scenario, which is predefined in the human shadow job knowledge ontology based on meteorological principles and artificial weather modification job experience; the cloud water content threshold is set based on the supercooled water condition required for catalytic job, and the exemplary value range is greater than 0.01 g / m³, and less than 0.01 g / m³, which is insufficient supercooled water in the cloud, and the catalytic job cannot trigger effective ice crystal growth and precipitation.

[0035] For each human shadow job event instance in the human shadow job event instance set, a node type is defined and a space-time attribute is assigned, to generate a time-series knowledge graph node set; Further, each human shadow job event instance in the human shadow job event instance set is traversed, and a node type (including “workable cloud system event node”, “dynamic trigger event node” and “catalytic job event node”) is defined for the human shadow job event instance according to the event mode matched by the human shadow job event instance when clustering; the start time, end time and spatial range coordinates of the human shadow job event instance affected by the human shadow job event instance are extracted from the event segment in the human shadow job event instance, as space-time attributes assigned to the corresponding human shadow job event instance node and assigned a globally unique node identifier, to generate a time-series knowledge graph node set.

[0036] Based on the time-series knowledge graph node set, the time-series relationship between nodes is mined, and the job rule is inferred to determine the causal relationship, to generate a human shadow job event relationship candidate set; Further, compare the time attributes of each pair of shadow operation event instance nodes in the time sequence knowledge graph node set, and judge the time sequence relationship (including "before", "after", and "overlap") between them; for the pair of shadow operation event instance nodes with time sequence relationship, load the constraint rules defined in the shadow operation knowledge ontology, which specifies the time sequence logic that must be met between events, and judge whether the shadow operation event instance node that occurs first can cause the shadow operation event instance node that occurs later (such as judging whether "operation catalytic condition node" can trigger "operable cloud system event node"); create a directed edge for each pair of shadow operation event instance nodes that satisfy the time sequence relationship and have a possible cause-effect relationship verified by the constraint rules, and the edge type is cause-effect relationship, and the generated shadow operation event relationship candidate set is collected.

[0037] Embedding the time axis and the spatial coordinate, construct the time sequence knowledge graph covering from before the operation to after the operation, taking the shadow operation event instance nodes as vertices and the shadow operation event relationship candidate set as edges. Further, set the shadow operation event instance nodes in the time sequence knowledge graph node set as vertices, and the node type and spatio-temporal attribute of each shadow operation event instance node as the attributes of the vertex; take each edge in the shadow operation event relationship candidate set as a directed edge connecting the corresponding vertices, and the attribute of the edge is the cause-effect relationship type; embed the time attribute (start time and end time) of each shadow operation event instance node into the vertex as the time axis attribute of the vertex, and the spatial range coordinate of each shadow operation event instance node into the vertex as the spatial coordinate attribute of the vertex; construct a graph structure through vertices and edges, including all shadow operation event instance nodes and cause-effect relationship edges from before the operation (such as power trigger event) to during the operation (such as catalytic operation event) to after the operation (such as precipitation enhancement event), forming a time sequence knowledge graph covering from before the operation to after the operation.

[0038] S3, use graph neural network to analyze the time sequence knowledge graph, identify cloud system development and change and infer the evolution of cloud microphysical characteristics, predict the spatio-temporal range of operation potential area, and generate accurate prediction conclusion and shadow operation suggestion; Feature encoding and alignment of node attributes, edge relationships and time stamps in the time sequence knowledge graph, converting into time sequence graph tensor object; Further, the attributes of each shadow job event instance node in the time sequence knowledge graph are numerically encoded, including node type and space-time attributes. The node type is converted into a vector using one-hot encoding, the spatial coordinates in the space-time attribute are converted into latitude and longitude numerical values, and the time attribute is converted into a relative timestamp relative to the start time of the job. The edge relationships in the time sequence knowledge graph are encoded, the causal relationship type is mapped to a specific integer index, and the absolute timestamp associated with each node and edge in the time sequence knowledge graph is standardized and converted into a unified relative time representation. The encoded node attribute vectors are combined into a node feature matrix, the encoded edge relationship index and the start and end node indices of the edge are combined into an edge index tensor, and the processed timestamp is combined into a time tensor. The node feature matrix, edge index tensor, edge relationship type tensor, and time tensor are aligned and encapsulated to generate a time sequence graph tensor object.

[0039] The time sequence graph tensor object is input into a graph neural network, and a low-dimensional vector representation of the node is obtained through forward propagation calculation. Further, the node feature matrix, edge index tensor, edge relationship type tensor, and time tensor in the time sequence graph tensor object are loaded into the input layer of the graph neural network. The graph neural network uses a message passing mechanism to pass and aggregate information between adjacent shadow job event instance nodes along the connection defined by the edge index tensor in the time sequence graph tensor object. During the aggregation process, different weights are assigned to different types of causal relationships based on the edge relationship type tensor, and the information within the time sequence neighborhood is integrated by weighting using the time tensor. After multiple iterations of message passing and feature transformation in the graph neural network, the features of each shadow job event instance node are updated and condensed into a low-dimensional vector representation of the node.

[0040] It should be noted that the pre-trained graph neural network is used. In specific operations, a historical time sequence knowledge graph containing labeled shadow job event instance nodes and causal relationship edges covering multiple typical weather scenarios is used, and the historical time sequence knowledge graph is converted into a time sequence graph tensor object and directly input into the graph neural network for forward propagation calculation to obtain a low-dimensional vector representation of the node. Link prediction is used as a self-supervised pre-training task, which randomly removes some causal relationship edges in the historical time sequence knowledge graph and requires the graph neural network to predict the possibility of the existence of the removed causal relationship edges based on the learned node representation. The loss between the prediction result and the true causal relationship edge is calculated, and the parameters of the graph neural network are iteratively optimized using the backpropagation algorithm. When the performance of the graph neural network on the link prediction task tends to be stable, the training is ended, and a graph neural network that has learned the knowledge in the shadow job field is obtained.

[0041] The attention weights of the graph neural network are analyzed to identify high-weight path subgraphs that connect thermodynamic-power trigger events and connect workable cloud systems-catalytic events, respectively, and obtain an evolution subgraph set. Further, the edge attention weight matrix is extracted from the last convolution layer of the graph neural network, which is an N x N sparse matrix (N is the total number of nodes), and each non-zero element represents the attention weight value of an edge. Each edge in the edge attention weight matrix is traversed, and if the attention weight of the edge is greater than or equal to the preset attention weight threshold, it is marked as a high weight edge. For all marked high weight edges, a connected subgraph extraction based on depth-first search is performed: starting from any high weight edge, recursively search for other high weight edges connected to the node until it cannot be expanded, thereby forming a connected path subgraph. The subgraph is classified according to the node types connected by the edges: if the edges in the path subgraph mainly connect "thermal trigger event" nodes and "dynamic trigger event" nodes, it is classified as a thermal-dynamic trigger event high weight path subgraph; if the edges in the path subgraph mainly connect "workable cloud system" nodes and "catalytic event" nodes, it is classified as a workable cloud system-catalytic event high weight path subgraph. All extracted and classified high weight path subgraphs are collected to form an evolution subgraph set.

[0042] It should be noted that the attention weight threshold is determined by grid search based on the performance (such as F1 score) of the link prediction task on the validation set reserved in the historical time series knowledge graph. The exemplary value range is 0.5-0.9. Below 0.5, too many low weight edges are retained, resulting in subgraphs containing a large number of irrelevant connections and reducing pattern recognition accuracy. Above 0.9, too many effective edges are filtered, resulting in important causal paths being broken and missing key evolution patterns.

[0043] From the node low-dimensional vector representation of the evolution subgraph set, the cloud system development evolution law, the work window threshold and the corridor constraint are decoded to generate a fusion quantitative rule set. Further, for the high-weight path subgraph of the thermal-dynamic trigger event in the evolution subgraph set, collect the low-dimensional vector representation of all "thermal trigger event" nodes and "dynamic trigger event" nodes, and obtain the attention weight of the edge connecting the nodes; according to the attention weight of the edge, the original convective available potential energy peak value corresponding to each "thermal trigger event" node is weighted and averaged to obtain the thermal condition threshold, and the original vertical wind shear peak value corresponding to each "dynamic trigger event" node is weighted and averaged to obtain the dynamic condition threshold, and the thermal condition threshold and the dynamic condition threshold directly constitute the quantitative judgment standard of the working catalytic condition, when the real-time observation of the convective available potential energy and the vertical wind shear value exceeds the two thresholds, it is determined that the working catalytic condition is met; the time difference sequence between the event nodes in the thermal-dynamic trigger event high-weight path subgraph is counted, and the mean and standard deviation are calculated as the time constraint of the evolution law, and the spatial distance sequence between the event nodes is counted, and the mean is calculated as the spatial scale of the evolution law; for the high-weight path subgraph of the workable cloud system-catalytic event, collect the low-dimensional vector representation of all "workable cloud system" nodes and the attention weight of the edge connected to the "catalytic event" node, and according to the attention weight, the original cloud water content peak value and the cloud top temperature extreme value corresponding to the "workable cloud system" node are weighted and averaged to obtain the work window threshold, the spatial offset vector of the "catalytic event" node relative to the connected "workable cloud system" node is counted, and the typical value of the direction and distance is calculated as the work corridor constraint; integrate all the trigger conditions, evolution laws, work window thresholds and corridor constraints obtained by decoding to generate a fused quantitative rule set.

[0044] The expression for obtaining the work corridor constraint is: ; Wherein, is the work corridor constraint; represents a point in geographical space, used to define the boundary range of the work corridor; is the path center line, representing the movement path of the workable cloud system node; is the lateral safety distance, used to define the width of the work corridor; represents the distance from the point to the path center line ; is a two-dimensional plane in Euclidean space, used to represent a continuous geographical coordinate space; It should be noted that before obtaining the work corridor constraint, the latitude and longitude coordinates (unit: degree) are converted into planar rectangular coordinates (unit: meter or kilometer) by map projection conversion (such as UTM or Lambert projection) to ensure the consistency of spatial dimensions before distance calculation and buffer analysis; the derivation process of the expression of the work corridor constraint is as follows: based on the spatio-temporal attributes (current position and movement vector) of the workable cloud system event node, the path center line is generated by linear extrapolation method, and the buffer analysis is performed on the path center line according to the constraint rules in the human shadow work knowledge ontology; the Euclidean distance between any point in geographic space and the path center line is calculated, the point set with Euclidean distance less than or equal to the lateral safety distance is screened, and the altitude constraint condition (such as the fan-shaped area with altitude between 3-5 kilometers) is superimposed to form the three-dimensional work corridor geographic range; It should be noted that the evolution law refers to the inherent mode of the time and space evolution of the thermal trigger event and the dynamic trigger event node in the strong convective weather process, which is quantified as the typical time difference between the successive occurrence of key events and the typical spatial scale of the propagation of event influence range; the work window threshold refers to the critical value that the key meteorological parameters must reach to determine whether to have the qualification to implement artificial weather modification work, which is quantified as cloud microphysics and thermal dynamic indexes (such as cloud water content needs to be ≥0.01 g / m³ and the temperature of the work area needs to be between -4℃-15℃); the corridor constraint refers to the spatial range limit that must be followed by the catalytic work operation to ensure the safety and effect of the work, which is limited by the safety range of each work point.

[0045] According to the fusion quantization rule set, the end nodes representing dangerous weather events are spatio-temporally extrapolated to obtain the spatio-temporal range of the work potential area, and the end nodes representing workable cloud systems are spatio-temporally extrapolated to obtain the work window time interval and the work corridor geographic range. Further, the end node with the latest occurrence in the time dimension in the evolution subgraph set and the node type of "dangerous weather event" (such as heavy precipitation event) is identified, the evolution law (i.e. the moving speed and direction of the event node influence range) decoded according to the fusion quantization rule set is used, and the linear extrapolation method is used to extrapolate the current spatial range of the end node along the moving direction according to the moving speed, so as to obtain the spatial position in a future period of time, that is, the operation potential area space-time range; the end node with the node type of "workable cloud system" is identified, if the physical properties (such as cloud water content and cloud top temperature) of the end node at the current time reach or exceed the operation window threshold value at the same time, it is determined that the operation catalytic condition is met, and the time period from the current time to the end of the life period of the end node is determined as the operation window time interval, if any property does not meet the standard, it is determined that the condition is not met, and the operation window is not generated; for the end node meeting the condition, the vector extrapolation method is used, the current position of the end node is taken as the path starting point, the moving speed and the operation effective time length in the fusion quantization rule set are used to calculate the ending point, the starting point and the ending point are connected to form a center line, and based on the center line, the lateral safety distance given in the fusion quantization rule set is extended to both sides, and a strip-shaped area with the center line as the central axis and the total width of 2 times the lateral safety distance is generated as the operation corridor geographical range.

[0046] The operation window time interval and the operation corridor geographical range are fused to generate a human shadow operation suggestion. Further, the operation window time interval and the operation corridor geographical range are aligned in the space-time dimension; each time of the operation window time interval is bound with the spatial position of the operation corridor geographical range at the corresponding time to generate a continuous spatial area that changes dynamically with time as an operation space-time domain; the operation space-time domain is given a preset operation action parameter to generate a human shadow operation suggestion.

[0047] It should be noted that the operation action parameter includes the catalyst type, the catalyst amount per unit time, and the operation platform cruising height, and based on the historical operation effect evaluation data, the median of the parameters corresponding to the optimal operation effect is determined by statistics.

[0048] According to the preset prediction level standard, the operation potential area space-time range is matched with the rule and the severity is evaluated to generate a precise prediction conclusion. Further, according to the preset prediction level standard, the intensity of the predicted weather phenomenon (such as the maximum wind speed in the next 1 hour, the cumulative precipitation in the next 3 hours) in each space-time grid in the space-time range of the operation potential area is determined one by one: if the predicted hourly maximum wind speed of the space-time grid reaches or exceeds the wind speed threshold of the gale blue prediction, it is determined that the space-time grid reaches the gale blue prediction level, and if the predicted 3-hour cumulative precipitation of the space-time grid reaches the rainfall threshold of the rainstorm orange prediction at the same time, it is determined that the space-time grid is a higher level of rainstorm orange prediction according to the higher principle; the proportion of the number of space-time grids reaching the orange and above prediction levels in the total number of space-time grids in the space-time range of the operation potential area is calculated, and the duration of the high-level prediction continuously covering the same geographical area is calculated; if the proportion exceeds the preset spatial coverage threshold and the duration exceeds the preset time range, the overall risk is evaluated as high severity, and the spatial boundary, risk type, prediction level distribution and severity of the whole operation potential area space-time range are integrated to generate a precise prediction conclusion.

[0049] S4, converting the precise prediction conclusion and human shadow operation suggestion into a set of visual instructions, highlighting the core weather indicators, dynamically displaying the evolution path with arrows, and marking the operation potential area and operation corridor with color coverage to generate a visual comprehensive situation map; From the precise prediction conclusion, the operation target cloud system index parameters, evolution path and operation potential area polygon coordinates are extracted, and a set of visual elements is generated in combination with the human shadow operation suggestion; Further, the key meteorological element value field in the precise prediction conclusion is read, and the "maximum wind speed value" and "cumulative precipitation value" are directly extracted as operation target cloud system index parameters, and the data segment (usually composed of a series of geographical coordinate points arranged in time sequence) in the precise prediction conclusion for describing the movement trajectory of the operation catalytic phenomenon is parsed. The geographical coordinate points are connected in time sequence to form the evolution path, and the geographical coordinate set in the precise prediction conclusion for defining the boundary of the operation potential area is located as the operation potential area polygon coordinates; the operation target cloud system index parameters, evolution path, operation potential area polygon coordinates and operation window time interval, operation corridor geographical boundary coordinates and operation action parameters in the human shadow operation suggestion are integrated to form a set of visual elements.

[0050] According to the preset visualization rule, the set of visual elements is mapped to specific commands executable by the graphic rendering engine to generate a set of visual instructions; Further, according to the preset visualization rule, the job potential area polygon coordinates are mapped to a red translucent filled drawing command, the evolution path coordinate sequence is mapped to a drawing command of a blue gradient color line with an arrow, the job target cloud system index parameter is mapped to a rendering command of a specific position text label, the job corridor geographical boundary coordinates are mapped to a green dashed line border drawing command, and the job window time interval is mapped to a time axis highlight segment marking command. All the instructions generated for different visualization elements are combined in the rendering order to generate a visualization instruction set.

[0051] It should be noted that the visualization rule is a specification set by defining the mapping relationship between data attributes and visual channels according to the meteorological industry mapping standard and the human-computer interaction cognitive principle.

[0052] The visualization instruction set is executed to drive the graphics rendering engine to respectively create a job potential area color overlay layer, an evolution path dynamic arrow layer, an index parameter highlight layer, a job corridor translucent pipeline layer, and a job window countdown marking layer, and generate a visualization layer set. Further, the instructions in the visualization instruction set about drawing the job potential area color overlay layer are parsed, a new layer is created by calling the underlying graphics application program interface, and filling rendering is performed according to the job potential area polygon coordinates and color parameters in the instructions to generate the job potential area color overlay layer. At the same time, the instructions about drawing the evolution path dynamic arrow layer, the index parameter highlight layer, the job corridor translucent pipeline layer, and the job window countdown marking layer are parsed in turn. For each type of instruction, the graphics rendering engine creates a new layer, and generates the evolution path dynamic arrow layer according to the evolution path coordinate sequence and arrow style in the instruction, generates the index parameter highlight layer according to the job target cloud system index parameter and position in the instruction, generates the job corridor translucent pipeline layer according to the job corridor geographical boundary coordinates and transparency in the instruction, and generates the job window countdown marking layer according to the job window time interval and marking style in the instruction. The rendered job potential area color overlay layer, evolution path dynamic arrow layer, index parameter highlight layer, job corridor translucent pipeline layer, and job window countdown marking layer are collected as a visualization layer set.

[0053] The data matching the spatio-temporal range of the job potential area is intercepted from the multi-source time series data set, and a multi-source data background layer is rendered and generated. Further, the spatiotemporal range of the operation potential area in the accurate forecast conclusion is taken as a screening condition to extract a data subset (including meteorological radar reflectivity, satellite cloud brightness temperature, and ground station observation data) completely overlapping or intersecting the spatiotemporal range of the operation potential area in time and space from a multi-source time series data set, to perform standardization rendering processing on the extracted multi-source data subset, to render the meteorological radar reflectivity data into a color spot map, to render the satellite cloud brightness temperature data into a grayscale map, and to render the ground station observation data into a site symbol filling map; and to combine the rendered multi-source data subset into a composite image according to a preset layer superimposition order to generate a multi-source data background layer.

[0054] It should be noted that the layer superimposition order is set according to the perspective relationship of the data sources (for example, the geographic information is the bottom layer, the satellite cloud image covers it, and the radar echo is the top layer) and the visual importance.

[0055] The visualization layer set and the multi-source data background layer are superimposed and fused to generate a visual comprehensive situation map. Further, the multi-source data background layer is loaded as a bottom basic image, the operation potential area color overlay layer, the evolution path dynamic arrow layer, the index parameter highlight layer, the operation corridor semi-transparent pipeline layer, and the operation window countdown marker layer are sequentially superimposed on the multi-source data background layer according to the definition order of the layers in the visualization layer set, and in the superimposition process, the geographic coordinates of each layer are strictly aligned with the multi-source data background layer; in the fusion process, the operation potential area color overlay layer adopts a semi-transparent mixing mode to ensure that the display of the bottom multi-source data background layer is not affected, the operation corridor semi-transparent pipeline layer adopts a brightening mixing mode to highlight the display, the dynamic arrow layer and the countdown marker layer adopt a superimposition mode to ensure that the dynamic effect is clear and visible, and a visual comprehensive situation map is generated.

[0056] The embodiment also provides a human shadow operation information display system based on multi-source information fusion, which comprises a data acquisition and repair module, a graph construction module, an inference analysis module, and a visualization display module. The data acquisition and repair module is used for collecting multi-source meteorological observation data for preprocessing, and repairing and super-resolution reconstructing defective data through an adversarial network model to generate a multi-source time series data set. The graph construction module is used for extracting time series events from the multi-source time series data set, identifying cloud body evolution, thermal and dynamic triggering, cloud system development, and operation action events related to human shadow operation as nodes, and constructing a time series knowledge graph covering before operation to after operation. The inference analysis module is used for performing inference analysis on the time series knowledge graph by using a graph neural network, identifying cloud system development and change and inferring evolution of cloud microphysical characteristics, predicting the spatiotemporal range of the operation potential area, and generating an accurate forecast conclusion and human shadow operation suggestion. The visualization display module is used for converting the accurate prediction conclusion and the human shadow operation suggestion into a visual instruction set, and generating a visual comprehensive situation map by highlighting the core weather index, dynamically displaying the evolution path through an arrow, and marking the operation potential area and operation corridor through color overlay.

[0057] The embodiment also provides a computer device suitable for the case of the human shadow operation information display method based on multi-source information fusion, which comprises a memory and a processor; the memory is used for storing computer executable instructions, and the processor is used for executing the computer executable instructions to realize the human shadow operation information display method based on multi-source information fusion proposed in the above embodiment.

[0058] The computer device can be a terminal, and the computer device comprises a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is used for providing computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used for wired or wireless communication with external terminals. The wireless communication can be realized through WIFI, an operator network, NFC (Near Field Communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0059] The embodiment also provides a storage medium having a computer program stored thereon, the program being executed by a processor to realize the human shadow operation information display method based on multi-source information fusion proposed in the above embodiment. The storage medium can be realized by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0060] To sum up, the application realizes the full-chain semantic representation of the weather evolution process and the human shadow operation by constructing the time sequence knowledge graph covering before operation to after operation, and provides an interpretable data basis for subsequent intelligent reasoning; by using the graph neural network to analyze the time sequence knowledge graph, the operation potential area is automatically identified from the historical mode and its dynamic evolution is predicted, and the forward-looking and accuracy of the human shadow operation decision are improved.

[0061] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A method for displaying information on weather modification operations based on multi-source information fusion, characterized in that: include, Multi-source meteorological observation data are collected and preprocessed, and defective data are repaired and super-resolution reconstructed using an adversarial network model to generate a multi-source time series dataset. Temporal events are extracted from multi-source temporal datasets to identify cloud evolution, thermal and dynamic triggering, cloud system development and operation action events related to cloud weathering operations and use them as nodes to construct a temporal knowledge graph covering the period from before to after the operation. By using graph neural networks to perform reasoning analysis on time-series knowledge graphs, we can identify the development and changes of cloud systems and infer the evolution of cloud physical characteristics, predict the spatiotemporal range of potential areas for operations, and generate accurate forecast conclusions and suggestions for artificial weather modification operations. The accurate forecast conclusions and suggestions for artificial weather modification operations are transformed into a set of visual instructions. By highlighting key weather indicators, displaying evolution paths with dynamic arrows, and marking potential operation areas and operation corridors with color coverage, a comprehensive visual situation map is generated.

2. The method for displaying artificial weathering operation information based on multi-source information fusion as described in claim 1, characterized in that: The preprocessing includes outlier removal, format standardization, coordinate unification, and spatiotemporal interpolation.

3. The method for displaying artificial weathering operation information based on multi-source information fusion as described in claim 2, characterized in that: The steps for generating the multi-source time-series dataset are as follows: Defective data in preprocessed multi-source meteorological observation data is identified and input into an adversarial network model for repair and super-resolution reconstruction to obtain a repaired and enhanced dataset; The repaired and enhanced dataset is fused with the effective data from the preprocessed multi-source meteorological observation data to generate a multi-source time-series dataset.

4. The method for displaying artificial weathering operation information based on multi-source information fusion as described in claim 3, characterized in that: The steps for extracting time-series events from multi-source time-series datasets are as follows: Extract variable data related to cloud microphysics, thermodynamics, and operation records from multi-source time-series datasets, and load entities, relationships, and constraint rules from the cloud weathering operation knowledge ontology to generate a time-series dataset of operation physical variables. Using a sliding window and abrupt change detection algorithm, event detection is performed on the time series dataset of operational physical variables to identify abrupt changes in cloud water content, energy jump points, and wind shear inflection points, generating a candidate fragment set of operational decision events. The candidate fragment set of operation decision events is matched and fused with the event patterns in the knowledge ontology of weather modification operations, and clustered to form a set of weather modification operation event instances.

5. The method for displaying artificial weathering operation information based on multi-source information fusion as described in claim 4, characterized in that: The steps for constructing a time-series knowledge graph covering the period from before to after the task are as follows. Define a node type and assign spatiotemporal attributes to each shadow operation event instance in the shadow operation event instance set to generate a temporal knowledge graph node set. Based on the node set of the temporal knowledge graph, the temporal relationship between nodes is mined, and the causal relationship is inferred by combining the operation rules to generate a candidate set of events related to the weather modification operation; Using event instance nodes of weather modification operations as vertices and candidate sets of event relationships of weather modification operations as edges, and embedding time axes and spatial coordinates, a temporal knowledge graph covering the period from before to after the operation is constructed.

6. The method for displaying information on weather modification operations based on multi-source information fusion as described in claim 5, characterized in that: The steps for identifying the development and changes of cloud systems are as follows: The node attributes, edge relationships, and timestamps in the temporal knowledge graph are feature-encoded and aligned, and then transformed into temporal graph tensor objects. The temporal graph tensor object is input into the graph neural network, and the low-dimensional vector representation of the node is obtained through forward propagation. By analyzing the attention weights of the graph neural network, high-weight path subgraphs connecting thermal-dynamic triggering events and working cloud systems-catalytic events are identified, and a set of evolutionary subgraphs is obtained.

7. The method for displaying information on weather modification operations based on multi-source information fusion as described in claim 6, characterized in that: The steps for generating accurate forecast conclusions and suggestions for weather modification operations are as follows. Decode the evolutionary patterns of cloud systems, operational window thresholds, and corridor constraints from the low-dimensional vector representations of nodes in the evolutionary subgraph set to generate a fusion-quantified rule set; Based on the fusion quantification rule set, the spatiotemporal range of the operational potential area is obtained by extrapolating the terminal nodes representing hazardous weather events in time and space. The operational window time interval and the geographical range of the operational corridor are obtained by extrapolating the terminal nodes representing operational cloud systems in time and space. By integrating the time interval of the operation window with the geographical range of the operation corridor, suggestions for artificial weather modification operations are generated. Based on the preset forecast level standards, the spatiotemporal range of the potential operational area is matched with rules and the severity is assessed to generate accurate forecast conclusions.

8. The method for displaying artificial weathering operation information based on multi-source information fusion as described in claim 7, characterized in that: The steps for converting accurate forecast conclusions and artificial weather modification operation suggestions into a visual instruction set are as follows. Extract target cloud system parameters, evolution paths, and polygon coordinates of potential operation areas from accurate forecast conclusions, and generate a set of visual elements by combining artificial weathering operation suggestions; Based on preset visualization rules, the set of visualization elements is mapped to specific commands that the graphics rendering engine can execute, generating a visualization instruction set.

9. The method for displaying information on weather modification operations based on multi-source information fusion as described in claim 8, characterized in that: The steps for generating the visualized comprehensive situation map are as follows: Execute the visualization instruction set to drive the graphics rendering engine to create a color overlay layer for the potential area of ​​the task, a dynamic arrow layer for the evolution path, a highlight layer for the indicator parameters, a semi-transparent pipe layer for the task corridor, and a countdown marker layer for the task window, thereby generating a set of visualization layers. Data matching the spatiotemporal range of the operational potential area is extracted from the multi-source time-series dataset and rendered to generate a multi-source data background layer; By overlaying and merging the visualization layer set with the multi-source data background layer, a comprehensive visualization situation map is generated.

10. A system for displaying information on weather modification operations based on multi-source information fusion, comprising the method for displaying information on weather modification operations based on multi-source information fusion as described in any one of claims 1 to 9, characterized in that: include, Data acquisition and repair module, map construction module, reasoning and analysis module, visualization display module; The data acquisition and repair module is used to collect multi-source meteorological observation data for preprocessing, and to repair and super-resolution reconstruct defective data through an adversarial network model to generate multi-source time series datasets. The graph construction module is used to extract time-series events from multi-source time-series datasets, identify cloud evolution, thermal and dynamic triggering, cloud system development and operation action events related to the weathering operation and use them as nodes to build a time-series knowledge graph covering the period from before to after the operation. The reasoning and analysis module is used to perform reasoning and analysis on the time-series knowledge graph using graph neural networks, identify the development and changes of cloud systems and reason about the evolution of cloud physical characteristics, predict the spatiotemporal range of potential operation areas, and generate accurate forecast conclusions and suggestions for artificial weather modification operations. The visualization module is used to transform accurate forecast conclusions and suggestions for artificial weather modification operations into a set of visual instructions. It generates a comprehensive visual situation map by highlighting key weather indicators, displaying evolution paths with dynamic arrows, and marking potential operation areas and operation corridors with color overlays.

Images