A method and system for estimating the influence area of a rocket artificial precipitation enhancement

Abstract

The application discloses a rocket artificial rain enhancement influence area estimation method and system, comprising the following steps: A1, collecting rocket rain enhancement operation parameter data, regional wind field data and atmospheric water vapor distribution data and respectively preprocessing; A2, respectively extracting structured diffusion characteristics, regional wind field characteristics and water vapor distribution characteristics; A3, cross-modal fusion, obtaining multi-modal fusion characteristics and catalyst diffusion characteristics; A4, extracting catalyst-water vapor interaction characteristics; A5, calculating the rocket artificial rain enhancement operation basic influence area and the basic grid range image; A6, the rocket artificial rain enhancement operation accurate influence area, and then performing grid statistics and scale conversion to obtain the rocket artificial rain enhancement operation accurate influence area. The application can solve the problem of insufficient accuracy of rocket artificial rain enhancement influence area estimation caused by the fact that the traditional method simplifies linear source diffusion into a trapezoidal model and does not fully consider multiple factors.

Description

Technical Field

[0001] This invention relates to the field of artificial rain enhancement technology, and in particular to a method and system for estimating the area affected by rocket-based artificial rain enhancement. Background Technology

[0002] The technology for estimating the impact area of ​​rocket-based artificial rain enhancement is a supporting technology for quantifying the effectiveness of rocket-based rain enhancement operations, guiding the optimization of operational plans, and supporting the evaluation of operational effects. Its core process involves collecting basic information such as rocket operation parameters, regional wind field data, and atmospheric water vapor distribution data. Based on a diffusion model, the diffusion range of the catalyst is simulated to determine the impact area of ​​the rain enhancement operation, providing data support for subsequent rain enhancement operation layout, resource allocation, and effectiveness review. Rocket-based artificial rain enhancement operations have significant unique characteristics and complexities, with the following two key features: First, the influencing factors are closely intertwined. The rain enhancement effect is simultaneously constrained by three core elements: rocket operation parameters, regional wind field, and atmospheric water vapor. The rocket operation parameters must conform to the inherent characteristics of the rocket type and match the wind direction at launch; otherwise, it will directly affect the initial diffusion direction of the catalyst. The regional wind field is constantly evolving; subtle changes in wind speed and turbulence intensity will alter the advection diffusion trajectory of the catalyst, which is a key factor dominating the diffusion process. Atmospheric water vapor exhibits multi-scale spatial distribution characteristics, with significant differences in water vapor concentration between the cloud center and edges, and between different regions, directly affecting whether the catalyst can be effectively activated to form rainfall. Secondly, the diffusion process is complex and unique. Rocket catalyst seeding is a linear source diffusion. In actual diffusion, there is not only the coupling effect of advection and turbulence, but also dynamic changes such as water vapor adsorption and condensation. It is difficult to accurately replicate the real diffusion law through simple mathematical models.

[0003] Currently, existing research in this field mainly focuses on quantifying the impact area of ​​aircraft-based rain enhancement scenarios. Systematic research on the impact range of rocket operations is still insufficient. Most existing estimation models simplify the linear source diffusion process of rocket operations into a trapezoidal model for linear extrapolation. However, the actual catalysts dispersed by rockets diffuse in the atmospheric environment through advection-turbulence coupling, accompanied by dynamic changes in water vapor adsorption. The simplified linear model cannot accurately depict the real diffusion law and ignores the interaction between multiple factors such as meteorological elements, operational parameters, and water vapor conditions. This results in a large deviation between the estimated results and the actual impact range, making it difficult to meet the needs of precision rain enhancement operations.

[0004] While some technologies have attempted to use deep learning models to modify traditional artificial rain enhancement estimation models to improve accuracy, existing deep learning applications are mostly designed for the area or point source diffusion characteristics of aircraft-based rain enhancement, and are not specifically adapted to the diffusion characteristics of rocket line source operations. This makes it difficult to compensate for the linear simplification defects of traditional trapezoidal models and achieve high-precision quantification of the area affected by rocket-based artificial rain enhancement. Summary of the Invention

[0005] In view of this, the present invention aims to provide a method and system for estimating the impact area of ​​rocket-based artificial rain enhancement, in order to solve the problem that the traditional method simplifies the linear source diffusion into a trapezoidal model and fails to adequately consider multiple factors, resulting in insufficient accuracy in estimating the impact area of ​​rocket-based artificial rain enhancement.

[0006] A method for estimating the area affected by rocket-based artificial rainmaking includes: A1: Collect rocket rain enhancement operation parameter data, regional wind field data, and atmospheric water vapor distribution data, and preprocess them respectively to obtain preprocessed rocket rain enhancement operation parameter data, preprocessed regional wind field data, and preprocessed atmospheric water vapor distribution data. A2: Based on the preprocessed rocket rain enhancement operation parameter data, preprocessed regional wind field data, and preprocessed atmospheric water vapor distribution data, extract structured diffusion features, regional wind field features, and water vapor distribution features respectively; A3: Based on the structured diffusion characteristics, regional wind field characteristics, and water vapor distribution characteristics, cross-modal fusion is performed to obtain multimodal fusion characteristics and catalyst diffusion characteristics; A4: Extract catalyst-water vapor interaction characteristics based on catalyst diffusion characteristics, water vapor distribution characteristics, structured diffusion characteristics, and regional wind field characteristics; A5: Based on the preprocessed rocket rain enhancement operation parameter data and the preprocessed regional wind field data, calculate the basic impact area and basic grid range image of rocket artificial rain enhancement operation; A6: The basic impact area, catalyst diffusion characteristics, and catalyst-water vapor interaction characteristics of rocket-based artificial rain enhancement operations are fused to obtain a precise grid range image of rocket-based artificial rain enhancement operations. Then, grid statistics and scale conversion are performed to obtain the precise impact area of ​​rocket-based artificial rain enhancement operations.

[0007] Furthermore, step A1 also includes: A11: Collect rocket rain enhancement operation parameter data. The data type is structured numerical data, including launch azimuth, rocket seeding dosage, seeding type identification, and catalyst seeding distance. The rocket rain enhancement operation parameter data is preprocessed by linear interpolation to complete missing values, removal of constant values ​​with three times standard deviation, and linear normalization to obtain the preprocessed rocket rain enhancement operation parameter data. A12: Collect regional wind field data, which is a spatiotemporal series data, including near-surface wind speed, wind direction, and turbulence intensity; preprocess the regional wind field data by sequentially applying Gaussian filtering for noise reduction, linear normalization, and time axis alignment to obtain preprocessed regional wind field data; A13: Collect atmospheric water vapor distribution data, which is spatial raster data, including atmospheric water vapor content and relative humidity; use a preprocessing method of raster resolution linear interpolation, linear normalization, and time axis alignment to obtain preprocessed atmospheric water vapor distribution data.

[0008] Furthermore, step A2 also includes: A21: Extract structured diffusion features based on the preprocessed rocket rain enhancement operation parameter data; A22: Extract regional wind field features based on the preprocessed regional wind field data; A23: Extract water vapor distribution characteristics based on the preprocessed atmospheric water vapor distribution data.

[0009] Furthermore, step A2 also includes: The extraction process of the structured diffusion features includes: first, processing the preprocessed rocket rain enhancement operation parameter data using a multilayer perceptron; then, processing the preprocessed rocket rain enhancement operation parameter data sequentially using a multilayer perceptron and a sigmoid function; performing a Hadamard product on the results of the two processing steps to obtain the structured diffusion intermediate features; calculating the absolute value of the difference between the launch azimuth angle and the wind direction angle at launch in the preprocessed rocket rain enhancement operation parameter data; performing a cosine calculation on this absolute value; then calculating the ratio of the rocket seeding dose to the standard seeding dose corresponding to the seeding type identifier and the ratio of the catalyst seeding distance to the standard seeding distance corresponding to the seeding type identifier in the preprocessed rocket rain enhancement operation parameter data; concatenating the cosine calculation results and the two ratios dimensionally to obtain the operation parameter domain constraint vector; and performing a Hadamard product on the structured diffusion intermediate features and the operation parameter domain constraint vector to obtain the structured association features. The process of extracting the regional wind field features includes: first, processing the preprocessed regional wind field data through a multilayer perceptron, dividing it by the square root of the attention dimension scaling factor, and then processing it through a Softmax function to obtain the wind speed temporal node mask; simultaneously, processing the preprocessed regional wind field data through a convolutional layer, a temporal convolutional network, and a bidirectional long short-term memory network in sequence, and then performing a Hadamard product between the processing result and the wind speed temporal node mask to obtain the regional wind field features. The extraction process of the water vapor distribution features includes: first, processing the preprocessed atmospheric water vapor distribution data through a multi-scale convolutional layer to obtain multi-scale raster features; performing average pooling and max pooling on the multi-scale raster features respectively, adding the two pooling results and processing them through a multilayer perceptron, and then performing a Hadamard product with the multi-scale raster features to obtain spatial attention-weighted features; performing average pooling and max pooling on the spatial attention-weighted features respectively, and summing the two pooling results element by element to obtain the water vapor distribution features.

[0010] It should be further explained that the effectiveness of rocket-based artificial rain enhancement operations is affected by three core elements: operational parameters, regional wind field, and atmospheric water vapor. Operational parameters need to be consistent with the characteristics of the rocket and the requirements of wind direction matching. The wind field has dynamic temporal evolution characteristics and directly dominates the advection diffusion trajectory of the catalyst. Water vapor exhibits multi-scale spatial distribution, and different concentrations in different regions have significantly different effects on catalyst activation. The operational patterns of these three elements need to be accurately captured to support subsequent estimation of the affected area. Traditional methods have not designed feature extraction schemes for the operational characteristics of each type of element, resulting in extracted features that are out of touch with the actual operational scenario, thus causing a large deviation in the estimation of the affected area. This invention achieves high-quality extraction of three types of features, as detailed below: When extracting structured diffusion features from preprocessed rocket rain enhancement operation parameter data, since rocket rain enhancement operation parameters are not isolated and there are inherent correlations between parameters such as launch azimuth, seeding dose, and seeding distance, this invention first inputs the preprocessed operation parameter data into a multilayer perceptron to extract basic correlation features between operation parameters, obtaining intermediate structured diffusion features. Simultaneously, the same operation parameter data is input into the multilayer perceptron again and activated by the Sigmoid function to obtain nonlinear mapping weights for parameter correlation. The two are then fused using Hadamard product operations to strengthen the correlation characteristics within the operation parameters. Subsequently, a domain constraint vector for the operation parameters is constructed. This vector includes the cosine value of the angle between the launch azimuth and wind direction, the ratio of the actual seeding dose to the standard seeding dose of the projectile type, and the actual seeding distance. The ratio of dispersal distance to standard dispersal distance of the projectile is used as the three dimensions. Then, the constraint vector is combined with the aforementioned fused structured diffusion intermediate features to perform a Hadamard product operation to obtain the final structured diffusion features. In the domain constraint vector of operational parameters, the matching degree between the launch azimuth and the wind direction directly affects the initial diffusion direction of the catalyst. The smaller the angle, the easier it is for the catalyst to diffuse along the wind direction. The ratio of the actual dispersal dose and distance to the standard value reflects the actual action capability of the catalyst. This is consistent with the actual business practice in rocket rain enhancement operations where the characteristics of the projectile determine the basic range of action and the wind direction determines the diffusion direction. This solves the problem that traditional operational parameter feature extraction only focuses on the parameter values ​​themselves and is detached from the physical boundaries of the operation and business constraints. This makes the extracted structured diffusion features include both the internal correlation of parameters and the core business rules of actual operations. When extracting regional wind field features from preprocessed regional wind field data, the preprocessed regional wind field data is first input into a convolutional layer to extract local spatial correlation features of the wind field, adapting to the spatial heterogeneity of the wind field in local areas. Then, the output features of the convolutional layer are input into a temporal convolutional network to capture the temporal evolution features of the wind field, matching the characteristics of dynamic changes in wind turbulence intensity and wind speed over time, and characterizing key changes such as sudden wind speed changes and turbulence enhancement in a short period. Subsequently, the output features of the temporal convolutional network are input into a bidirectional long short-term memory network to further capture the bidirectional temporal dependencies of the wind field. Simultaneously... The wind field data is input into a multilayer perceptron and processed by the Softmax function to obtain a wind speed temporal node mask. This mask is used to weight the wind field features at different temporal nodes, highlighting the nodes that have a key impact on catalyst diffusion. Finally, the output features of the bidirectional long short-term memory network are combined with the wind speed temporal node mask to obtain the final regional wind field features. This accurately captures the impact of dynamic wind field changes on diffusion, effectively solving the pain point of traditional methods that use static feature extraction and cannot support dynamic diffusion process modeling. This makes the extracted regional wind field features dynamic and targeted. When extracting water vapor distribution features from preprocessed atmospheric water vapor distribution data, the preprocessed atmospheric water vapor distribution data is first input into a multi-scale convolutional layer to extract water vapor raster features at different scales. This covers both the overall large-scale water vapor distribution of clouds and local small-scale water vapor accumulation areas, avoiding the limitation of traditional single-scale convolution which can only capture water vapor information at a single scale. Then, the multi-scale raster features output from the multi-scale convolutional layer are processed by average pooling and max pooling respectively. The sum of the two pooling results is then input into a multilayer perceptron to obtain the water vapor spatial attention weights. Subsequently, the attention weights are performed with the multi-scale raster features to obtain spatial attention-weighted features, which highlight high water vapor concentration areas and weaken low water vapor concentration areas, reflecting the actual situation that catalysts are easily activated in high water vapor areas and difficult to be effective in low water vapor areas. Finally, the spatial attention-weighted features are processed by average pooling and element-wise summation respectively to obtain the final water vapor distribution features, providing high-resolution support for subsequent catalyst-water vapor interaction modeling and solving the problem that traditional single-scale features tend to smooth out regional differences in water vapor. Based on the extraction of three types of features, this invention performs linear layer and batch normalization processing on the structured diffusion features, regional wind field features, and water vapor distribution features, and then concatenates them dimensionally to achieve preliminary fusion of cross-modal features, resulting in multimodal fused features. Then, a rocket rain enhancement physical constraint vector containing key physical quantities such as near-surface wind speed, turbulence intensity, atmospheric water vapor content, and relative humidity is introduced. After the multimodal fused features are input into a multilayer perceptron, a Hadamard product operation is performed with the physical constraint vector to obtain the catalyst diffusion features. The cross-modal fusion method of this invention is not a simple feature concatenation, but rather strengthens the operational relevance of the three types of single-source features through the physical constraint vector. This makes the fused features deeply aligned with the physical mechanisms of operational parameter constraints, wind field-driven diffusion, and water vapor-provided activation conditions in rocket rain enhancement operations, achieving deep coupling of the three types of elements.

[0011] Furthermore, step A4 also includes: A41: Based on the catalyst diffusion characteristics and water vapor distribution characteristics, a bidirectional modulation operation is performed through an adaptive temporal control network to extract diffusion-water vapor modulation characteristics; A42: Based on catalyst diffusion characteristics, structured diffusion characteristics, and regional wind field characteristics, diffusion-wind field modulation characteristics are extracted through a dual-gated cross-attention mechanism; A43: Residual enhancement and dynamic calibration calculations were performed on the diffusion-wind field modulation characteristics and diffusion-water vapor modulation characteristics to obtain the catalyst-water vapor interaction characteristics.

[0012] Furthermore, step A4 also includes: The adaptive temporal modulation network takes catalyst diffusion features and water vapor distribution features as inputs, obtains diffusion attention features through an attention mechanism, performs Hadamard product operation with diffusion attention features after processing by a multilayer perceptron and sigmoid activation, performs layer normalization, and then sums element-wise with the convolutional catalyst diffusion features to obtain diffusion-water vapor modulation features. The dual-gated cross-attention mechanism first processes the regional wind field features and catalyst diffusion features through a multilayer perceptron and activates them with Sigmoid to obtain wind field gating weights and diffusion gating weights. Then, it performs Hadamard product operations with the attention associations of catalyst diffusion features and structured diffusion features, and catalyst diffusion features and regional wind field features, respectively. The two operation results are then summed element-wise, processed by a Transformer network, and then subjected to Hadamard product operations with the layer-normalized catalyst diffusion features to obtain diffusion-wind field modulation features.

[0013] It needs further explanation that catalyst particles act as condensation nuclei, attracting surrounding water vapor and forming localized high-moisture areas, thereby expanding the rainfall enhancement range. Traditional methods mostly only consider the adsorption effect of water vapor on the catalyst, ignoring the reverse effect of the catalyst diffusion path simultaneously guiding water vapor accumulation. This results in modulation features failing to reflect the phenomenon of the catalyst actively influencing water vapor distribution. In addition, traditional methods use indiscriminate weighting for the interaction of catalyst, operating parameters, and wind field, without dynamically adjusting the weights according to the business scenario. For example, in a strong wind field environment, the dominant effect of the wind field on diffusion is far greater than the constraining effect of operating parameters; in a weak wind field environment, the influence of operating parameters will be enhanced. Traditional interaction modeling methods cannot achieve this dynamic weight adjustment, resulting in modulation features that do not conform to actual laws. In the diffusion-moisture modulation stage, this invention first uses catalyst diffusion characteristics as query features and water vapor distribution characteristics as both bond and value features. An attention mechanism is used to capture the deep correlation between these two types of features, resulting in diffusion attention features, which initially explore the interaction between the catalyst and water vapor. Subsequently, the diffusion attention features are combined with those processed by a multilayer perceptron and activated by a Sigmoid function using a Hadamard product. Layer normalization is then used to stabilize the feature distribution, while convolution processing is applied to the catalyst diffusion features to strengthen their spatial correlation. Finally, the two processed features are summed element-wise to obtain the diffusion-moisture modulation features. This bidirectional modulation design is not a single feature mapping but reflects both the adsorption and activation effect of water vapor on the catalyst and the reverse effect of the catalyst acting as a condensation nucleus to guide the accumulation of surrounding water vapor. This recreates the bidirectional influence between the catalyst and water vapor in rocket-based rain enhancement operations, addressing the shortcomings of traditional methods that only consider unidirectional effects and whose modulation features cannot reflect the active influence of the catalyst on water vapor distribution. In the diffusion-wind field modulation stage, this invention relies on a dual-gated cross-attention mechanism and a Transformer network to achieve dynamic weighting. First, regional wind field features and catalyst diffusion features are input into a multilayer perceptron and activated by a Sigmoid function to generate corresponding wind field gating weights and diffusion gating weights. These gating weights can be automatically adjusted according to the actual meteorological scenario. Then, through the dual-gated cross-attention mechanism, attention associations are constructed between catalyst diffusion features and structured diffusion features, and between catalyst diffusion features and regional wind field features, respectively. Finally, the wind field gating weight and diffusion gating weight are used to weight the two sets of attention results. The weighted results are summed element-wise to obtain the dual-gated cross-attention output features. Finally, this output feature is input into the Transformer network to further explore the deep correlations between features, and then a Hadamard product is performed with the catalyst diffusion features that have undergone layer normalization to obtain the diffusion-wind field modulation features. Among them, the dynamic adjustment characteristic of the gating weight is used to dynamically adapt to different scenarios. In the strong wind field environment, the wind field gating weight is automatically increased to strengthen the dominant role of the wind field on diffusion. In the weak wind field environment, the proportion of diffusion gating weight is increased to highlight the constraint effect of the operation parameters on diffusion, which solves the problem that the modulation features do not conform to the actual law of the industry due to the indiscriminate weighting of traditional methods. In the residual enhancement and dynamic calibration stage, the diffusion-wind field modulation features and the convolution-processed diffusion-water vapor modulation features are first summed element-wise to obtain residual features. Then, the residual features are summed element-wise with the residual features processed by the multilayer perceptron. At the same time, the product of the wind field gating weight and the diffusion gating weight is combined to construct an inverse weighting factor. The features are then weighted by the Hadamard product operation. Finally, global average pooling is used to obtain the final catalyst-water vapor interaction features. Among them, the product of the wind field gating weight and the diffusion gating weight is used to characterize the synergistic influence strength of wind field and diffusion. Based on this, the weighting process can filter out redundant features in non-synergistic regions and focus on the core interaction relationship among catalyst, wind field, and water vapor, making the final extracted interaction features more business-specific. Existing technologies for characterizing the interaction between catalysts, wind fields, and water vapor generally employ unidirectional feature mapping or simple weighted fusion, failing to consider the bidirectional interaction between catalysts and water vapor, and lacking a dynamic weight adjustment gating mechanism. Consequently, the extracted interaction features are disconnected from real-world business scenarios. In contrast, the advantages of this invention lie in: diffusion-water vapor modulation characterizes the bidirectional interaction; diffusion-wind field modulation enables intelligent dynamic weight adaptation; and residual calibration effectively filters redundant information. The entire process recreates the interaction mechanism of the three components, ensuring feature validity. The extracted catalyst-water vapor interaction features can directly provide data support for subsequent accurate estimation of the impact area, significantly improving the overall solution's business applicability and estimation accuracy.

[0014] Furthermore, step A5 also includes: A51: Based on the preprocessed rocket rain enhancement operation parameter data and the preprocessed regional wind field data, the basic impact area of ​​rocket artificial rain enhancement operation under the traditional trapezoidal model is calculated using the line source diffusion method; in the traditional trapezoidal model, the upper base of the trapezoid is the catalyst dispensing distance, the lower base is the turbulent diffusion width, the height is the advection diffusion length, and the extension direction is consistent with the wind direction; A52: Set the grid resolution with the rocket launch point as the grid origin; A53: Along the direction of trapezoidal extension, divide the advection diffusion length by the grid resolution and then round down to obtain the number of grid cells in the advection direction; A54: Along the width of the trapezoid, divide the turbulence diffusion width by 2, divide by the grid resolution, and then round down to obtain the number of grids in each direction on both sides; A55: Based on the number of grid cells in the advection direction and the number of grid cells in each direction on both sides, and referring to the trapezoidal boundary of the traditional trapezoidal model, the grid cells are arranged to obtain the basic grid range image.

[0015] Furthermore, step A6 also includes: A61: Based on the basic impact area of ​​rocket-based artificial rain enhancement operations, catalyst diffusion characteristics, and catalyst-water vapor interaction characteristics, feature fusion is performed to obtain a precise raster range image of rocket-based artificial rain enhancement operations; A62: Based on the precise grid range image of rocket-based artificial rain enhancement operations, grid statistics and scale conversion are performed to obtain the precise area of ​​impact of rocket-based artificial rain enhancement operations.

[0016] It should be further explained that in estimating the impact area of ​​rocket-based artificial rain enhancement, the traditional line source diffusion method only roughly calculates the basic area based on the trapezoidal model without correcting deviations by incorporating actual operational factors. This results in the estimation results not matching the actual rain enhancement coverage area, failing to meet the precise operational needs for operational effect evaluation and benefit accounting. It is necessary to combine core elements such as catalyst diffusion and water vapor interaction to correct the theoretical model deviations, while simultaneously achieving accurate mapping of raster data to the actual geographical area. This invention employs a multilayer perceptron to process the basic influence area derived from the trapezoidal model. This area is then fused with catalyst diffusion features and catalyst-water vapor interaction features to form the input features for the corrected model. These features are further processed by the multilayer perceptron and residual connection blocks to obtain intermediate features for deviation prediction. Next, spatial features of the basic raster range image are extracted through a convolutional layer. The dimensions of the intermediate features for deviation prediction are reshaped and fused with the Hadamard product of the basic raster spatial features. A precise raster range image is then generated using a Sigmoid function, thus correcting the theoretical trapezoidal raster for actual business elements and ensuring the raster range matches the effective catalyst coverage area. Finally, through raster statistics and scale conversion, the raster data is transformed into actual geographical area. Existing technologies rely solely on trapezoidal models or estimate precise areas using fixed empirical coefficients, lacking business feature fusion correction and grid spatial feature adaptation design, resulting in insufficient accuracy. The advantage of this invention is that it abandons empirical coefficients, uses multi-core business feature fusion to correct the inherent bias of the trapezoidal model, and combines grid spatial feature processing to achieve accurate correction. The final quantification result is more in line with the actual rainfall enhancement range, significantly improving accuracy.

[0017] This invention also discloses a system for estimating the area affected by rocket-based artificial rainmaking, comprising: Rocket operation data acquisition module: Collects rocket rain enhancement operation parameter data, regional wind field data, and atmospheric water vapor distribution data, and preprocesses them respectively to obtain preprocessed rocket rain enhancement operation parameter data, preprocessed regional wind field data, and preprocessed atmospheric water vapor distribution data; Basic feature extraction module: Based on the preprocessed rocket rain enhancement operation parameter data, preprocessed regional wind field data, and preprocessed atmospheric water vapor distribution data, structured diffusion features, regional wind field features, and water vapor distribution features are extracted respectively. Cross-modal feature fusion module: Based on structured diffusion features, regional wind field features, and water vapor distribution features, cross-modal fusion is performed to obtain multimodal fused features and catalyst diffusion features; Catalyst-water vapor interaction module: Extract catalyst-water vapor interaction characteristics based on catalyst diffusion characteristics, water vapor distribution characteristics, structured diffusion characteristics, and regional wind field characteristics; Basic Impact Area Calculation Module: Based on the preprocessed rocket rain enhancement operation parameter data and preprocessed regional wind field data, calculate the basic impact area and basic grid range image of rocket artificial rain enhancement operation; Precise Impact Area Calculation Module: This module integrates the basic impact area of ​​rocket-based artificial rain enhancement operations, catalyst diffusion characteristics, and catalyst-water vapor interaction characteristics to obtain the precise impact area of ​​rocket-based artificial rain enhancement operations. Then, it performs grid statistics and scale conversion to obtain the precise impact area of ​​rocket-based artificial rain enhancement operations.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) To address the problem that traditional methods simplify rocket line source diffusion into a trapezoidal model and ignore the interaction of multiple factors, resulting in large deviations in the estimation of the impact area of ​​rocket-based artificial rain enhancement operations, this invention improves the accuracy of the estimation of the impact area of ​​rocket-based artificial rain enhancement by integrating operational and meteorological data, feature extraction, and multi-dimensional fusion. This invention first systematically processes operational parameters, regional wind fields, and atmospheric water vapor data, fully explores various key features, and achieves cross-dimensional fusion. At the same time, it focuses on capturing the interaction between catalyst and water vapor, and then performs in-depth fusion calibration in combination with the basic impact area to finally achieve accurate quantification. This invention can truly restore the diffusion law of catalyst and effectively avoids the neglect of the interaction of dynamic meteorological elements, operational parameters, and water vapor conditions in traditional methods, thus reducing the deviation between the estimation results and the actual impact range.

[0019] (2) In response to the traditional methods in the process of extracting operational parameters, wind field, and water vapor features, this invention designs feature extraction ideas respectively: structured diffusion features are extracted by combining multilayer perceptron with domain constraint vectors, dynamic wind field features are captured by using convolutional layers and temporal networks with masks, high-resolution water vapor features are obtained by using multi-scale convolution and attention mechanisms, and multimodal deep coupling is achieved by initial splicing and fusion with physical constraint vectors; among them, structured diffusion features match the elastic characteristics and wind direction requirements to ensure that the initial diffusion of the catalyst meets expectations and avoids dissemination deviation; dynamic wind field features capture the dynamics of the wind field in real time to provide a basis for judging the diffusion range; high-resolution water vapor features distinguish the differences in water vapor concentration, focus on high water vapor areas to improve catalytic efficiency, and after feature fusion, provide a feature basis for the area affected by subsequent rainfall.

[0020] (3) In view of the problems of traditional methods ignoring the reverse effect of catalyst-guided water vapor accumulation and not adjusting the influence weight of wind field and operation parameters in actual operation scenarios such as strong / weak wind fields, this invention designs a bidirectional diffusion-water vapor modulation and dynamic diffusion-wind field weighting scheme, which, combined with the previous feature extraction, has a significant effect: the diffusion-water vapor bidirectional modulation not only reflects the adsorption and activation of water vapor on the catalyst, but also captures the effect of catalyst-guided water vapor accumulation, restores the real interaction law, and improves the efficiency of catalytic rainfall; the diffusion-wind field dynamic weighting can automatically adapt to strong and weak wind field scenarios, dynamically adjust the weight of each element, and ensure that the modulation features are in line with reality under different meteorological conditions, which helps to achieve high-quality interaction of various features and to make accurate rainfall area estimation in subsequent steps.

[0021] (4) In the estimation of the impact area of ​​rocket-based artificial rain enhancement, the traditional line source diffusion method only calculates the basic area based on the trapezoidal model without combining actual business elements to correct the deviation, resulting in the estimation result not matching the actual rain enhancement coverage. This invention combines the basic impact area obtained from the trapezoidal model and corrects the deviation of the trapezoidal model with the features extracted in the early stage, so that the final quantitative result is more in line with the actual rain enhancement range and the accuracy is significantly improved. Attached Figure Description

[0022] Figure 1 A flowchart illustrating a method for estimating the impact area of ​​rocket-based artificial rain enhancement provided by this invention; Figure 2 A comparison diagram of the traditional trapezoidal model provided by this invention and the rainfall area calculated by this invention. Detailed Implementation

[0023] The present invention will be further described below with reference to the accompanying drawings, but this is not intended to limit the present invention in any way. Any modifications or substitutions made based on the teachings of the present invention shall fall within the protection scope of the present invention.

[0024] Example 1: A method for estimating the area affected by rocket-based artificial rainmaking, such as... Figure 1 As shown, it includes the following steps: A1: Collect rocket rain enhancement operation parameter data, regional wind field data, and atmospheric water vapor distribution data, and preprocess them respectively to obtain preprocessed rocket rain enhancement operation parameter data, preprocessed regional wind field data, and preprocessed atmospheric water vapor distribution data, including: A11: Collect rocket rain enhancement operation parameter data. The data type is structured numerical data, including launch azimuth, rocket seeding dosage, seeding type identification, and catalyst seeding distance. The rocket rain enhancement operation parameter data is preprocessed by linear interpolation to complete missing values, removal of constant values ​​with three times standard deviation, and linear normalization to obtain the preprocessed rocket rain enhancement operation parameter data. A12: Collect regional wind field data, which is a spatiotemporal series data, including near-surface wind speed, wind direction, and turbulence intensity; preprocess the regional wind field data by sequentially applying Gaussian filtering for noise reduction, linear normalization, and time axis alignment to obtain preprocessed regional wind field data; A13: Collect atmospheric water vapor distribution data, which is spatial raster data, including atmospheric water vapor content and relative humidity; use a preprocessing method of raster resolution linear interpolation, linear normalization, and time axis alignment to obtain preprocessed atmospheric water vapor distribution data.

[0025] A2: Based on the preprocessed rocket rain enhancement operation parameter data, preprocessed regional wind field data, and preprocessed atmospheric water vapor distribution data, structured diffusion features, regional wind field features, and water vapor distribution features are extracted, including: A21: Based on the preprocessed rocket rain enhancement operation parameter data, structured diffusion characteristics are extracted, and the calculation method is as follows: ; ; ; in, This is an intermediate feature of structured diffusion. It is a multilayer perceptron. The data represents the preprocessed parameters for rocket-based rain enhancement operations. For the Sigmoid function, For the domain constraint vector of the operation parameters, This is a dimension splicing operation. For cosine calculation, To take the absolute value, The launch azimuth angle in the preprocessed rocket rain enhancement operation parameter data. The wind direction angle at launch. The rocket seeding dosage in the preprocessed rocket rain enhancement operation parameter data. The standard application dose corresponding to the bullet type designation. For catalyst dispersal distance, The standard dispersing distance corresponding to the bullet type marking. For structured association features, For Hadamah accumulation; A22: Based on the preprocessed regional wind field data, extract the regional wind field features. The calculation method is as follows: ; ; in, This is the wind speed timing node mask. For the Softmax function, This is the preprocessed regional wind field data. The scaling factor for the attention dimension. Characteristics of the regional wind field. It is a bidirectional long short-term memory network. It is a temporal convolutional network. It is a convolutional layer; A23: Based on the preprocessed atmospheric water vapor distribution data, water vapor distribution characteristics are extracted. The calculation method is as follows: ; ; ; in, For multi-scale raster features, It is a multi-scale convolutional layer. This is the preprocessed atmospheric water vapor distribution data. Spatial attention-weighted features, For average pooling, For max pooling, Characteristics of water vapor distribution This is for summing element by element.

[0026] A3: Based on the structured diffusion characteristics, regional wind field characteristics, and water vapor distribution characteristics, cross-modal fusion is performed to obtain multimodal fusion characteristics and catalyst diffusion characteristics. The calculation method is as follows: ; ; ; in, For multimodal fusion features, For linear layers, For batch normalization, Physical constraint vector for rocket-assisted rain enhancement. This refers to the near-surface wind speed in the preprocessed regional wind field data. The turbulence intensity in the preprocessed regional wind field data, This refers to the atmospheric water vapor content in the preprocessed atmospheric water vapor distribution data. This refers to the relative humidity in the preprocessed atmospheric water vapor distribution data. This is a characteristic of catalyst diffusion.

[0027] A4: Based on catalyst diffusion characteristics, water vapor distribution characteristics, structured diffusion characteristics, and regional wind field characteristics, extract catalyst-water vapor interaction characteristics, including: A41: Based on the catalyst diffusion characteristics and water vapor distribution characteristics, a bidirectional modulation operation is performed using an adaptive temporal control network to extract the diffusion-water vapor modulation characteristics. The calculation method is as follows: ; ; in, To diffusing attention characteristics, For attention mechanisms, It is a diffusion-water vapor modulation characteristic. For layer normalization; A42: Based on catalyst diffusion characteristics, structured diffusion characteristics, and regional wind field characteristics, diffusion-wind field modulation characteristics are extracted using a dual-gated cross-attention mechanism. The calculation method is as follows: ; in, This is a diffusion-wind field modulation characteristic. For Transformer networks, This is a dual-gated cross-attention mechanism, and its calculation method is as follows: ; ; ; in, For wind farm gating weights, For diffusion gating weights; A43: Residual enhancement and dynamic calibration calculations are performed on the diffusion-wind field modulation characteristics and diffusion-water vapor modulation characteristics to obtain the catalyst-water vapor interaction characteristics. The calculation method is as follows: ; ; in, For residual characteristics, The catalyst-water vapor interaction characteristics, This is global average pooling.

[0028] Specifically, for complex wind field scenarios with high turbulence intensity, this invention also provides a residual enhancement and dynamic calibration calculation method to strengthen turbulence feature weights, replacing step A43. The calculation method is as follows: ; ; in, It is the hyperbolic tangent function.

[0029] A5: Based on the preprocessed rocket-based rain enhancement operation parameter data and the preprocessed regional wind field data, calculate the basic impact area and basic grid range image of the rocket-based artificial rain enhancement operation, including: A51: Based on the preprocessed rocket rain enhancement operation parameter data and the preprocessed regional wind field data, the basic impact area of ​​rocket artificial rain enhancement operation under the traditional trapezoidal model is calculated using the line source diffusion method; in the traditional trapezoidal model, the upper base of the trapezoid is the catalyst dispensing distance, the lower base is the turbulent diffusion width, the height is the advection diffusion length, and the extension direction is consistent with the wind direction; A52: Set the grid resolution with the rocket launch point as the grid origin; A53: Along the direction of trapezoidal extension, divide the advection diffusion length by the grid resolution and then round down to obtain the number of grid cells in the advection direction; A54: Along the width of the trapezoid, divide the turbulence diffusion width by 2, divide by the grid resolution, and then round down to obtain the number of grids in each direction on both sides; A55: Based on the number of grid cells in the advection direction and the number of grid cells in each direction on both sides, and referring to the trapezoidal boundary of the traditional trapezoidal model, the grid cells are arranged to obtain the basic grid range image.

[0030] A6: By fusing the basic impact area, catalyst diffusion characteristics, and catalyst-water vapor interaction characteristics of rocket-based artificial rain enhancement operations, the precise impact area of ​​rocket-based artificial rain enhancement operations is obtained. Then, grid statistics and scale conversion are performed to obtain the precise impact area of ​​rocket-based artificial rain enhancement operations, including: A61: Based on the basic impact area of ​​rocket-based artificial rain enhancement operations, catalyst diffusion characteristics, and catalyst-water vapor interaction characteristics, feature fusion is performed to obtain a precise raster range image of rocket-based artificial rain enhancement operations. The calculation method is as follows: ; ; ; ; ; in, To correct the input features of the model, The area affected by rocket-based artificial rain enhancement operations. For bias prediction of intermediate features, For residual connection blocks, Based on the raster range image, Based on the basic raster range feature, It is a convolutional neural network. For precise raster range features, To be Matched Dimensional reshaping operations, Precise raster-area images for rocket-based artificial rain enhancement operations; A62: Based on the precise grid range image of rocket-based artificial rain enhancement operations, grid statistics and scale conversion are performed to obtain the precise area of ​​impact of rocket-based artificial rain enhancement operations.

[0031] In this embodiment, the neural network module parameters for generating the precise raster range image of rocket-based artificial rain enhancement operations described in step A61 are configured as follows: 1. A multilayer perceptron for processing the basic impact area of ​​rocket-assisted artificial rainmaking operations is set up with a 3-layer fully connected structure. The number of neurons in the input layer is consistent with the dimension of the basic impact area feature. The number of neurons in the hidden layer is 128 and 64 respectively, and the number of neurons in the output layer is 32. Each layer uses the ReLU activation function. 2. A multilayer perceptron for generating intermediate features for bias prediction has 96 neurons in the input layer, 256 and 128 neurons in the hidden layers respectively, and 64 neurons in the output layer. The activation function and Dropout layer configuration are the same as those of the aforementioned multilayer perceptron. The residual connection block contains two convolutional layers, each with 64 kernels, a stride of 1, and SAME padding. A batch normalization layer is connected after the convolutional layers, and finally, the element-wise summation of the input features and the convolutional features is achieved through a short-circuit branch. 3. The convolutional layer used to extract the basic grid range features is set as a single convolutional layer structure with 64 kernels, a stride of 1, SAME padding, and ReLU activation function. 4. The dimension reshaping operation adopts the method of dimension expansion and copying to adjust the dimension of the intermediate features of the deviation prediction to match the spatial dimension of the basic raster range features; finally, the sigmoid activation function is used to process the fused features, and the output value range is mapped to the precise raster range image in the 0-1 interval.

[0032] The training process of the neural network module described in this invention adopts a supervised learning training method, and the specific process is as follows: First, training and validation datasets were constructed, with labeled rocket rain enhancement operation-related data divided in an 8:2 ratio. The input to the training dataset consisted of preprocessed rocket rain enhancement operation parameter data, regional wind field data, and atmospheric water vapor distribution data, with corresponding labels being the precise impact area raster data obtained through field measurements after the actual rocket rain enhancement operation. The validation dataset was configured with the same data format and used for model performance verification during the training process. Secondly, the parameters of each module in the neural network are initialized. The weight parameters of modules such as multilayer perceptron, convolutional layer, multi-scale convolutional layer, bidirectional long short-term memory network, temporal convolutional network, Transformer network and attention mechanism are initialized using the Xavier initialization method, and the bias parameters are uniformly initialized to 0. Next, the training-related hyperparameters are set, and the loss function chosen is the mean squared error loss function suitable for regression tasks; the optimizer is the Adam optimizer, and the initial learning rate is set... During training, a cosine annealing strategy is employed, with the learning rate dynamically adjusted every 10 training epochs to balance the model's convergence speed and accuracy; simultaneously, a coefficient is introduced. The L2 regularization mechanism suppresses model overfitting and improves model generalization ability; the training batch size is set to 32; Subsequently, iterative training of the model is performed. The training dataset is input into the neural network in fixed batches to sequentially complete the forward computation process of extracting structured diffusion features, regional wind field features, water vapor distribution features, cross-modal feature fusion, catalyst-water vapor interaction feature extraction, and precise grid range image generation. The loss value between the predicted value and the true label is calculated based on the loss function. The loss value is then propagated back to each module through the backpropagation algorithm. The Adam optimizer updates the weights and bias parameters of each module. Finally, an early stopping strategy is used to complete model training. After each round of training, the performance of the current model is verified using the validation dataset. If the loss value on the validation set does not decrease for 5 consecutive rounds, training is terminated and the current optimal model parameters are saved.

[0033] This experiment selected the Sanbanxi Reservoir area and its surrounding area in southeastern Guizhou as the experimental setting. This area is a typical mountainous reservoir area with significant topographic relief and uneven water vapor distribution. Rocket-based artificial rain enhancement operations were conducted at 2 PM on the day of the experiment, with 10 operation points set up (4 in Bageng, Taijiang County, and 2 in Xingshan, Majiang County), and a total of 44 rockets launched. During the operation, the average near-surface wind speed was 3.2 meters per second, and the turbulence intensity reached 0.8 (a high-turbulence wind field). The relative humidity in the core area of ​​the reservoir was 78%, while the relative humidity in the northwest edge area of ​​the reservoir was only 42% (a low-moisture ineffective zone). The objective of the operation was to increase the runoff replenishment of the Sanbanxi Reservoir. The basic impact area of ​​rocket-based artificial rain enhancement operations in this scenario was calculated using a traditional trapezoidal model. The final basic impact area was approximately 6,200 square kilometers, with its coverage area distributed in a trapezoidal shape. The line connecting all operation points is the center line, and the extension direction is consistent with the prevailing wind direction. The upper base length is equal to the catalyst dissemination distance (average 12 kilometers), the lower base is the turbulent diffusion width (average 38 kilometers), and the height is the advection diffusion length (average 162 kilometers). This model did not consider the invalidity of the low-moisture area on the northwest edge of the reservoir area, including this entire area in the impact range. At the same time, it did not correct for the catalyst diffusion offset caused by the high turbulent wind field, and concluded that "the reservoir area and the surrounding 6,200 square kilometers will be affected by rain enhancement, and runoff interception and early warning should be carried out." The method of this invention calculates the precise impact area of ​​rocket-based artificial rain enhancement operations to be approximately 4,800 square kilometers. The precise grid map shows that the affected area is mainly concentrated in the core area of ​​the reservoir, the vegetation-covered area to the southeast, and the downstream runoff collection area. The grid resolution is set to 1 km × 1 km, with approximately 4,800 effective grids. The method explicitly excludes the low-moisture ineffective area (approximately 1,400 square kilometers) to the northwest of the reservoir where relative humidity is below 50%. It also corrects for the northeastward shift of the catalyst due to high-turbulence winds, reducing the impact area of ​​the northeastern region without moisture replenishment (by approximately 200 square kilometers). The final conclusion is that "the precise impact area is 4,800 square kilometers, and only the reservoir and surrounding effective moisture areas will generate rain enhancement runoff; therefore, early warning should be focused on the core area." The traditional trapezoidal model and the rainfall area calculated by this invention, such as Figure 2 As shown, the gray area represents the predicted area of ​​the traditional trapezoidal model, while the irregular shape represents the rainfall area calculated by this invention.

[0034] Example 2: This invention also discloses a system for estimating the area affected by rocket-based artificial rainmaking, comprising: Rocket operation data acquisition module: Collects rocket rain enhancement operation parameter data, regional wind field data, and atmospheric water vapor distribution data, and preprocesses them respectively to obtain preprocessed rocket rain enhancement operation parameter data, preprocessed regional wind field data, and preprocessed atmospheric water vapor distribution data; Basic feature extraction module: Based on the preprocessed rocket rain enhancement operation parameter data, preprocessed regional wind field data, and preprocessed atmospheric water vapor distribution data, structured diffusion features, regional wind field features, and water vapor distribution features are extracted respectively. Cross-modal feature fusion module: Based on structured diffusion features, regional wind field features, and water vapor distribution features, cross-modal fusion is performed to obtain multimodal fused features and catalyst diffusion features; Catalyst-water vapor interaction module: Extract catalyst-water vapor interaction characteristics based on catalyst diffusion characteristics, water vapor distribution characteristics, structured diffusion characteristics, and regional wind field characteristics; Basic Impact Area Calculation Module: Based on the preprocessed rocket rain enhancement operation parameter data and preprocessed regional wind field data, calculate the basic impact area and basic grid range image of rocket artificial rain enhancement operation; Precise Impact Area Calculation Module: This module integrates the basic impact area of ​​rocket-based artificial rain enhancement operations, catalyst diffusion characteristics, and catalyst-water vapor interaction characteristics to obtain the precise impact area of ​​rocket-based artificial rain enhancement operations. Then, it performs grid statistics and scale conversion to obtain the precise impact area of ​​rocket-based artificial rain enhancement operations.

[0035] It should be noted that the sequence numbers of the above embodiments of the present invention are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0036] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0037] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for estimating the area affected by rocket-based artificial rainmaking, characterized in that, Includes the following steps: A1: Collect rocket rain enhancement operation parameter data, regional wind field data, and atmospheric water vapor distribution data, and preprocess them respectively to obtain preprocessed rocket rain enhancement operation parameter data, preprocessed regional wind field data, and preprocessed atmospheric water vapor distribution data. Step A1 includes: A11: Collect rocket rain enhancement operation parameter data. The data type is structured numerical data, including launch azimuth, rocket seeding dosage, seeding type identification, and catalyst seeding distance. The rocket rain enhancement operation parameter data is preprocessed by linear interpolation to complete missing values, removal of constant values ​​with three times standard deviation, and linear normalization to obtain the preprocessed rocket rain enhancement operation parameter data. A12: Collect regional wind field data, which is a spatiotemporal series data, including near-surface wind speed, wind direction, and turbulence intensity; preprocess the regional wind field data by sequentially applying Gaussian filtering for noise reduction, linear normalization, and time axis alignment to obtain preprocessed regional wind field data; A13: Collect atmospheric water vapor distribution data, which is spatial raster data, including atmospheric water vapor content and relative humidity; use a preprocessing method of raster resolution linear interpolation, linear normalization, and time axis alignment to obtain preprocessed atmospheric water vapor distribution data. A2: Based on the preprocessed rocket rain enhancement operation parameter data, preprocessed regional wind field data, and preprocessed atmospheric water vapor distribution data, extract structured diffusion features, regional wind field features, and water vapor distribution features respectively; Step A2 includes: A21: Extract structured diffusion features based on the preprocessed rocket rain enhancement operation parameter data; A22: Extract regional wind field features based on the preprocessed regional wind field data; A23: Extract water vapor distribution characteristics based on the preprocessed atmospheric water vapor distribution data; The extraction process of the structured diffusion features includes: first, processing the preprocessed rocket rain enhancement operation parameter data using a multilayer perceptron; then, processing the preprocessed rocket rain enhancement operation parameter data sequentially using a multilayer perceptron and a sigmoid function; performing a Hadamard product on the results of the two processing steps to obtain the structured diffusion intermediate features; calculating the absolute value of the difference between the launch azimuth angle and the wind direction angle at launch in the preprocessed rocket rain enhancement operation parameter data; performing a cosine calculation on this absolute value; then calculating the ratio of the rocket seeding dose to the standard seeding dose corresponding to the seeding type identifier and the ratio of the catalyst seeding distance to the standard seeding distance corresponding to the seeding type identifier in the preprocessed rocket rain enhancement operation parameter data; concatenating the cosine calculation results and the two ratios dimensionally to obtain the operation parameter domain constraint vector; and performing a Hadamard product on the structured diffusion intermediate features and the operation parameter domain constraint vector to obtain the structured association features. The process of extracting the regional wind field features includes: first, processing the preprocessed regional wind field data through a multilayer perceptron, dividing it by the square root of the attention dimension scaling factor, and then processing it through a Softmax function to obtain the wind speed temporal node mask; simultaneously, processing the preprocessed regional wind field data through a convolutional layer, a temporal convolutional network, and a bidirectional long short-term memory network in sequence, and then performing a Hadamard product between the processing result and the wind speed temporal node mask to obtain the regional wind field features. The extraction process of the water vapor distribution features includes: first, processing the preprocessed atmospheric water vapor distribution data through a multi-scale convolutional layer to obtain multi-scale raster features; performing average pooling and max pooling on the multi-scale raster features respectively, adding the two pooling results and processing them through a multilayer perceptron, and then performing a Hadamard product with the multi-scale raster features to obtain spatial attention-weighted features; performing average pooling and max pooling on the spatial attention-weighted features respectively, and summing the two pooling results element by element to obtain the water vapor distribution features; A3: Based on the structured diffusion characteristics, regional wind field characteristics, and water vapor distribution characteristics, cross-modal fusion is performed to obtain multimodal fusion characteristics and catalyst diffusion characteristics; A4: Extract catalyst-water vapor interaction characteristics based on catalyst diffusion characteristics, water vapor distribution characteristics, structured diffusion characteristics, and regional wind field characteristics; A5: Based on the preprocessed rocket rain enhancement operation parameter data and the preprocessed regional wind field data, calculate the basic impact area and basic grid range image of rocket artificial rain enhancement operation; A6: The basic impact area, catalyst diffusion characteristics, and catalyst-water vapor interaction characteristics of rocket-based artificial rain enhancement operations are fused to obtain a precise grid range image of rocket-based artificial rain enhancement operations. Then, grid statistics and scale conversion are performed to obtain the precise impact area of ​​rocket-based artificial rain enhancement operations.

2. The method for estimating the area affected by rocket-based artificial rainmaking according to claim 1, characterized in that, The A4 step includes: A41: Based on the catalyst diffusion characteristics and water vapor distribution characteristics, a bidirectional modulation operation is performed through an adaptive temporal control network to extract diffusion-water vapor modulation characteristics; A42: Based on catalyst diffusion characteristics, structured diffusion characteristics, and regional wind field characteristics, diffusion-wind field modulation characteristics are extracted through a dual-gated cross-attention mechanism; A43: Residual enhancement and dynamic calibration calculations were performed on the diffusion-wind field modulation characteristics and diffusion-water vapor modulation characteristics to obtain the catalyst-water vapor interaction characteristics.

3. The method for estimating the area affected by rocket-based artificial rainmaking according to claim 2, characterized in that, The A4 step includes: The adaptive temporal modulation network takes catalyst diffusion features and water vapor distribution features as inputs, obtains diffusion attention features through an attention mechanism, performs Hadamard product operation with diffusion attention features after processing by a multilayer perceptron and sigmoid activation, performs layer normalization, and then sums element-wise with the convolutional catalyst diffusion features to obtain diffusion-water vapor modulation features. The dual-gated cross-attention mechanism first processes the regional wind field features and catalyst diffusion features through a multilayer perceptron and activates them with Sigmoid to obtain wind field gating weights and diffusion gating weights. Then, it performs Hadamard product operations with the attention associations between catalyst diffusion features and structured diffusion features, and between catalyst diffusion features and regional wind field features, respectively. The two operation results are then summed element-wise, processed by a Transformer network, and then subjected to Hadamard product operations with the layer-normalized catalyst diffusion features to obtain diffusion-wind field modulation features.

4. The method for estimating the impact area of ​​rocket-based artificial rainmaking according to claim 3, characterized in that, Step A5 includes: A51: Based on the preprocessed rocket rain enhancement operation parameter data and the preprocessed regional wind field data, the basic impact area of ​​rocket artificial rain enhancement operation under the traditional trapezoidal model is calculated using the line source diffusion method; in the traditional trapezoidal model, the upper base of the trapezoid is the catalyst dispensing distance, the lower base is the turbulent diffusion width, the height is the advection diffusion length, and the extension direction is consistent with the wind direction; A52: Set the grid resolution with the rocket launch point as the grid origin; A53: Along the direction of trapezoidal extension, divide the advection diffusion length by the grid resolution and then round down to obtain the number of grid cells in the advection direction; A54: Along the width of the trapezoid, divide the turbulence diffusion width by 2, divide by the grid resolution, and then round down to obtain the number of grids in each direction on both sides; A55: Based on the number of grid cells in the advection direction and the number of grid cells in each direction on both sides, and referring to the trapezoidal boundary of the traditional trapezoidal model, the grid cells are arranged to obtain the basic grid range image.

5. The method for estimating the area affected by rocket-based artificial rainmaking according to claim 4, characterized in that, Step A6 includes: A61: Based on the basic impact area of ​​rocket-based artificial rain enhancement operations, catalyst diffusion characteristics, and catalyst-water vapor interaction characteristics, feature fusion is performed to obtain a precise raster range image of rocket-based artificial rain enhancement operations; A62: Based on the precise grid range image of rocket-based artificial rain enhancement operations, grid statistics and scale conversion are performed to obtain the precise area of ​​impact of rocket-based artificial rain enhancement operations.

6. A system for estimating the impact area of ​​rocket-assisted artificial rainmaking, characterized in that, include: Rocket operation data acquisition module: Collects rocket rain enhancement operation parameter data, regional wind field data, and atmospheric water vapor distribution data, and preprocesses them respectively to obtain preprocessed rocket rain enhancement operation parameter data, preprocessed regional wind field data, and preprocessed atmospheric water vapor distribution data; Basic feature extraction module: Based on the preprocessed rocket rain enhancement operation parameter data, preprocessed regional wind field data, and preprocessed atmospheric water vapor distribution data, structured diffusion features, regional wind field features, and water vapor distribution features are extracted respectively. Cross-modal feature fusion module: Based on structured diffusion features, regional wind field features, and water vapor distribution features, cross-modal fusion is performed to obtain multimodal fused features and catalyst diffusion features; Catalyst-water vapor interaction module: Extract catalyst-water vapor interaction characteristics based on catalyst diffusion characteristics, water vapor distribution characteristics, structured diffusion characteristics, and regional wind field characteristics; Basic Impact Area Calculation Module: Based on the preprocessed rocket rain enhancement operation parameter data and preprocessed regional wind field data, calculate the basic impact area and basic grid range image of rocket artificial rain enhancement operation; The precise impact area calculation module integrates the basic impact area of ​​rocket-based artificial rain enhancement operations, catalyst diffusion characteristics, and catalyst-water vapor interaction characteristics to obtain the precise impact area of ​​rocket-based artificial rain enhancement operations. Then, it performs grid statistics and scale conversion to obtain the precise impact area of ​​rocket-based artificial rain enhancement operations; thereby realizing the rocket-based artificial rain enhancement impact area estimation method as described in any one of claims 1-5.

Images