A Regression Analysis Method and System for Thunderstorm Winds Based on Multiple Convection Parameters

Abstract

This invention discloses a method and system for thunderstorm gale regression analysis based on multiple convection parameters, belonging to the field of data processing technology. The method includes: real-time acquisition of multiple convection parameters to construct a real-time multiple convection parameter sequence; mapping the real-time multiple convection parameter sequence to a pre-constructed multiple convection parameter coupled sensitivity field model to calculate and generate a real-time convection sensitivity field; inputting the real-time convection sensitivity field into a physical consistency constraint model to generate real-time structured convection characteristics; performing a master regression analysis under multiple convection parameter regression to establish the master regression result; after calculating the real-time local disturbance index, triggering disturbance residual compensation of the master regression result, and dynamically adjusting the weights of the residual-compensated master regression result using seasonal phase characteristics, outputting the thunderstorm gale regression analysis result. This solves the technical problem of insufficient accuracy in thunderstorm gale prediction in existing technologies, achieving the technical effect of improving the accuracy of thunderstorm gale regression analysis results.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and system for regression analysis of thunderstorm winds based on multiple convection parameters. Background Technology

[0002] Thunderstorms and strong winds are a type of sudden and disastrous weather process triggered by severe convective weather systems. They are characterized by rapid occurrence, small spatial scale, and high destructiveness, widely affecting urban operations, traffic safety, and the stability of power facilities. Existing methods for predicting and analyzing thunderstorms and strong winds are mostly based on numerical weather prediction products, radiosonde data, or empirical threshold indicators. These methods perform single-indicator or low-dimensional combination analyses of factors such as convective instability, wind shear, and water vapor conditions to determine the probability of thunderstorms and strong winds. However, these methods typically focus on static parameters or instantaneous state discrimination, making it difficult to reflect the coordinated changes of multiple convective parameters over time. On the other hand, with the deepening research into convective weather mechanisms, existing technologies have gradually introduced statistical regression or machine learning models to predict thunderstorms and strong winds. However, existing models often fail to adequately characterize the coupling sensitivity between multiple convective parameters, lack effective modeling of vertical structure consistency, convective stability constraints, and abrupt triggering mechanisms. This makes regression analysis results susceptible to local disturbances, compromising prediction accuracy and robustness. Furthermore, the dominant mechanisms of convective systems differ significantly across different seasons, and existing techniques typically do not dynamically adjust for seasonal phase characteristics, further limiting the applicability and reliability of thunderstorm wind regression analysis results. Summary of the Invention

[0003] This application provides a method and system for regression analysis of thunderstorm winds based on multiple convection parameters, which solves the technical problem of insufficient accuracy in predicting thunderstorm winds in the prior art.

[0004] The first aspect of this application provides a regression analysis method for thunderstorm winds based on multiple convection parameters, the method comprising:

[0005] Multiple convective parameters, representing convective instability, wind shear, and temporal variations of water vapor conditions, are acquired in real time from numerical models, radiosondes, or analytical data to construct a real-time multi-convective parameter sequence. This real-time multi-convective parameter sequence is mapped to a pre-constructed multi-convective parameter coupled sensitivity field model, and a real-time convective sensitivity field is generated through the coupling calculation of parameter covariance structure and gradient structure. The real-time convective sensitivity field is input into a physical consistency constraint model, and real-time structured convective features are generated based on stability constraints, vertical structure constraints, and abrupt triggering constraints. Based on the real-time structured convective features, a master regression analysis under multi-convective parameter regression is performed to establish the master regression results. After calculating the real-time local disturbance index, disturbance residual compensation of the master regression results is triggered, and the seasonal phase characteristics are used to dynamically adjust the weights of the residual-compensated master regression results, outputting the thunderstorm and gale regression analysis results.

[0006] A second aspect of this application provides a system for regression analysis of thunderstorm winds based on multiple convection parameters, the system comprising:

[0007] The system comprises the following modules: Data Acquisition Module: Real-time acquisition of multiple convective parameters characterizing convective instability, wind shear, and temporal variations of water vapor conditions from numerical models, radiosondes, or analytical data, constructing a real-time multi-convective parameter sequence; Sensitivity Field Generation Module: Mapping the real-time multi-convective parameter sequence to a pre-constructed coupled sensitivity field model, generating a real-time convective sensitivity field through the coupling calculation of parameter covariance structure and gradient structure; Feature Generation Module: Inputting the real-time convective sensitivity field into a physical consistency constraint model, generating real-time structured convective features based on stability constraints, vertical structure constraints, and abrupt triggering constraints; Principal Regression Analysis Module: Performing principal regression analysis under multi-convective parameter regression based on the real-time structured convective features, establishing principal regression results; Result Output Module: After calculating the real-time local disturbance index, triggering disturbance residual compensation of the principal regression results, and dynamically adjusting the weights of the residual-compensated principal regression results using seasonal phase characteristics, outputting the thunderstorm and gale regression analysis results.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] First, multiple convective parameters, representing convective instability, wind shear, and temporal variations of water vapor conditions, are acquired in real-time from numerical models, radiosondes, or analytical data to construct a real-time multi-convective parameter sequence. Next, this real-time multi-convective parameter sequence is mapped to a pre-constructed coupled sensitivity field model, and a real-time convective sensitivity field is generated through the coupling calculation of parameter covariance structure and gradient structure. Further, the real-time convective sensitivity field is input into a physical consistency constraint model, and real-time structured convective features are generated based on stability constraints, vertical structure constraints, and abrupt change triggering constraints. Then, based on the real-time structured convective features, a master regression analysis under multi-convective parameter regression is performed to establish the master regression results. Finally, after calculating the real-time local disturbance index, disturbance residual compensation of the master regression results is triggered, and the seasonal phase characteristics are used to dynamically adjust the weights of the residual-compensated master regression results, outputting the thunderstorm gale regression analysis results. This solves the technical problem of insufficient accuracy in predicting thunderstorm gales in existing technologies and achieves the technical effect of improving the accuracy of thunderstorm gale regression analysis results. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 A schematic diagram of the process for a regression analysis method for thunderstorm winds based on multiple convection parameters provided in this application embodiment;

[0012] Figure 2 A schematic diagram of the structure of a thunderstorm and strong wind regression analysis system based on multiple convection parameters provided in this application embodiment.

[0013] Figure labeling: Data acquisition module 11, sensitivity field generation module 12, feature generation module 13, main regression analysis module 14, result output module 15. Detailed Implementation

[0014] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structure, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0015] Example 1, as Figure 1 As shown, this application provides a regression analysis method for thunderstorm winds based on multiple convection parameters, wherein the method includes:

[0016] Real-time acquisition of multiple convective parameters, which characterize convective instability, wind shear, and temporal variations of water vapor conditions, from numerical models, radiosondes, or analytical data, is used to construct real-time multi-convective parameter sequences.

[0017] In this embodiment, basic meteorological element data related to convective activity are acquired in real time from numerical weather prediction model outputs, ground and upper-air radiosonde observation systems, or operational analysis data via a meteorological data access interface. This basic meteorological element data includes at least temperature, humidity, air pressure, wind speed, and wind direction at different altitudes. Based on this basic meteorological element data, according to preset convective parameter calculation rules, parameters characterizing convective instability, wind shear parameters characterizing the differences between low- and mid-to-high-level wind fields, and water vapor condition parameters characterizing water vapor transport and convergence are calculated. Each convective parameter corresponds to the current moment and its adjacent historical moments in the time dimension. Time synchronization and spatial registration are performed on convective parameters acquired from different sources, uniformly mapping the convective parameters to a consistent time resolution and spatial grid scale. After time synchronization and spatial registration are completed, the multiple convective parameters corresponding to each moment are combined in chronological order to form a real-time multi-convective parameter sequence containing convective instability parameters, wind shear parameters, and water vapor condition parameters, used to characterize the evolution of the convective environment over time.

[0018] Based on basic meteorological data, and following preset convective parameter calculation rules, parameters characterizing convective instability, wind shear parameters characterizing the differences between lower and middle-upper-level wind fields, and water vapor condition parameters characterizing water vapor transport and convergence are calculated, specifically including:

[0019] Vertical profiles were constructed from collected temperature, humidity, and air pressure data at different altitudes. Based on the thermodynamic changes during air parcel uplift, parameters characterizing convective instability were calculated. These parameters include, but are not limited to, convective effective potential energy, suppression energy, uplift index, K-index, and Showalter index. Wind vectors at different altitudes were calculated based on low- and mid-to-high-altitude wind speed and direction data. Wind shear parameters characterizing differences in vertical wind field structure were obtained through inter-layer wind vector difference or wind speed change rate analysis. These wind shear parameters include, but are not limited to, parameters from 0 to 1 km. Vertical wind shear intensity, wind speed difference, and wind vector shear parameters between 0~3km or 0~6km altitude layers; based on humidity, temperature, and wind field data, calculate water vapor content and its spatial and temporal variation characteristics, and combine the effect of horizontal wind field on water vapor transport to quantify water vapor transport and local convergence characteristics, forming water vapor condition parameters. These water vapor condition parameters include, but are not limited to, precipitable water, overall water vapor flux, water vapor flux divergence, lower-level specific humidity or relative humidity, and water vapor convergence intensity index; among them, convective instability parameters, wind shear parameters, and water vapor condition parameters are all output with a unified time step.

[0020] The real-time multi-convective parameter sequence is mapped to a pre-constructed multi-convective parameter coupled sensitivity field model, and the real-time convection sensitivity field is generated through the coupling calculation of parameter covariance structure and gradient structure.

[0021] In this embodiment, a real-time multi-convective parameter sequence is used as the model input. According to a preset spatial grid and isobaric height layer structure, the multi-convective parameter sequence is spatially reconstructed and vertically layered, so that each convective parameter forms a corresponding parameter field at different height layers and spatial locations. Based on the layered multi-convective parameter fields, the local covariance relationship between convective parameters is calculated within each isobaric height layer to obtain a local covariance structure characterizing the statistical correlation of multiple convective parameters within the same height layer. Furthermore, considering the physical correlation between adjacent height layers, a linkage analysis is performed on the local covariance structures of each height layer. This process involves forming a cross-layer covariance structure that reflects the vertical parameter coupling characteristics, and performing feature constraint processing on the cross-layer covariance structure to reduce the impact of spurious correlations caused by random disturbances or noise. Simultaneously, local variation gradients are calculated along the horizontal and vertical directions for the layered multi-convective parameter field to obtain gradient information reflecting the intensity and direction characteristics of parameter spatial variation. The gradient information is then scaled to construct a gradient structure with unified dimensions. The parameter covariance structure and gradient structure are coupled and calculated to generate a real-time convection sensitivity field that comprehensively reflects the statistical coupling relationship and spatial variation sensitivity of multi-convective parameters.

[0022] Furthermore, mapping the real-time multi-pair flow parameter sequence to a pre-constructed multi-pair flow parameter coupling sensitivity field model includes:

[0023] The real-time multi-convective parameter sequence is vertically stratified according to isobaric height layers, and the local covariance matrix of the convective parameters within each height layer is calculated. The physical correlation between each height layer is obtained, and based on this physical correlation, a vertical linkage analysis is performed on the local covariance matrix to construct a cross-layer covariance matrix. Eigenvalue constraint processing is performed on the local and cross-layer covariance matrices to suppress spurious correlations caused by noise or weak disturbances, thus constructing a parameter covariance structure matrix. Local gradients are calculated along the horizontal and vertical directions for the real-time multi-convective parameter sequence, forming a gradient tensor for multiple physical quantities, including temperature gradient, wind speed gradient, and humidity gradient. Scale normalization processing is performed on the gradient tensor to construct a gradient structure matrix. The parameter covariance structure matrix and the gradient structure matrix are coupled for calculation to generate a real-time convection sensitivity field.

[0024] First, based on a pre-defined isobaric height layer division rule, the real-time multi-convective parameter sequence is vertically layered, mapping the multi-convective parameters at each moment to their corresponding isobaric height layers. Within each isobaric height layer, using the temporal variation of the multi-convective parameters as samples, a local covariance matrix is ​​calculated among the convective parameters within that height layer to characterize the statistical correlation between different convective parameters within the same height layer. Second, physical correlation information between different isobaric height layers is obtained, including at least the vertical development mechanism of the convective system and energy transfer relationships. Based on this physical correlation, a vertical linkage analysis is performed on the local covariance matrices corresponding to each height layer to construct a cross-layer covariance matrix, characterizing the coordinated change features of multi-convective parameters in the vertical structure. After obtaining the local and cross-layer covariance matrices, eigenvalue constraint processing is performed on both. By retaining the principal eigenvalues ​​and suppressing small eigenvalues, the influence of spurious correlations caused by noise or weak disturbances is weakened, thereby constructing a stable parameter covariance structure matrix. Simultaneously, local gradients are calculated along the horizontal and vertical directions in the spatial dimension for real-time multi-convective parameter sequences, forming a multi-physical quantity gradient tensor containing temperature, wind speed, and humidity gradients. This tensor characterizes the spatial variation intensity and directional features of physical quantities in the convective environment. Scale normalization is performed on the gradient tensor to map the gradients of different physical quantities to a space with the same dimensions, constructing a gradient structure matrix to eliminate the influence of numerical scale differences between different physical quantities on subsequent calculations. Finally, the parameter covariance structure matrix and the gradient structure matrix are coupled and calculated using a collaborative weighting and feature fusion approach to generate a real-time convection sensitivity field.

[0025] Coupled computation is performed through collaborative weighting and feature fusion, specifically including:

[0026] Based on the physical sensitivity of convective parameters during thunderstorm and gale formation, corresponding weight coefficients are assigned to the parameter covariance structure matrix and gradient structure matrix, respectively. These weight coefficients characterize the relative contributions of statistical coupling features and spatial variation features in convective sensitivity assessment. After weight assignment, the parameter covariance structure matrix and gradient structure matrix are weighted element-wise or partition-wise to enhance the response of highly sensitive regions in subsequent calculations, while suppressing regions with low correlation or weak variation. Based on the weighting results, feature fusion is performed on the parameter covariance structure features and gradient structure features, mapping the covariance information reflecting the statistical correlation of parameters and the gradient information reflecting the intensity of spatial variation to the same feature space to form a coupled feature representation. The coupled feature representation is spatially reconstructed to generate a coupled sensitivity feature field continuously distributed in the horizontal and vertical directions, which serves as the output of the real-time convective sensitivity field.

[0027] Furthermore, the multi-convection parameter coupling sensitivity field model is a multi-level evolutionary network model, including:

[0028] The system comprises the following layers: an input layer for receiving real-time multi-convective parameter sequences; a local covariance calculation layer for calculating the local covariance matrix of convective parameters within each isobaric height layer; a cross-layer covariance evolution layer for evolving the local covariance matrix into a cross-layer covariance matrix based on the physical correlation between height layers; a gradient tensor calculation layer for calculating temperature gradient, wind speed gradient, and humidity gradient along the horizontal and vertical directions, and normalizing them into a gradient structure matrix; a feature coupling fusion layer for coupling the parameter covariance structure matrix and the gradient structure matrix to output a coupled sensitivity feature tensor; and a sensitivity field generation layer for generating a real-time convective sensitivity field based on the coupled sensitivity feature tensor.

[0029] The multi-convective parameter coupling sensitivity field model is constructed as a multi-level evolutionary network model. It jointly models the statistical coupling relationships and spatial variation characteristics among multiple convection parameters through hierarchical calculation and step-by-step evolution. Specifically, it includes the following functional layers: The input layer receives real-time multi-convective parameter sequences, including convective instability parameters, wind shear parameters, and water vapor condition parameters at different time steps, spatial grids, and isobaric height levels, serving as the basic input data for subsequent evolutionary calculations. The local covariance calculation layer, connected to the input layer, performs vertical hierarchical processing on the input multi-convective parameter sequences according to the isobaric height level division rules, and calculates the local covariance matrix between convection parameters within each isobaric height level to characterize the statistical correlation characteristics of multiple convection parameters within the same height level. The cross-layer covariance evolution layer is connected to the local covariance calculation layer to obtain physical correlation information between different height layers. Based on the physical correlation, the local covariance matrix of each height layer is evolved to generate a cross-layer covariance matrix, thus reflecting the coordinated change relationship of multiple convection parameters in the vertical structure. The gradient tensor calculation layer is used to calculate the spatial change gradient of the real-time multi-convective parameter sequence along the horizontal and vertical directions, obtaining the temperature gradient, wind speed gradient, and humidity gradient, respectively. The gradients are then scaled to construct a gradient structure matrix under a unified dimension, which is used to characterize the spatial change intensity and directional characteristics of multiple physical quantities. The feature coupling fusion layer is connected to the cross-layer covariance evolution layer and the gradient tensor calculation layer. It is used to couple the parameter covariance structure matrix composed of the local covariance matrix and the cross-layer covariance matrix with the gradient structure matrix. Through coordinated weighting and feature fusion, a coupling sensitivity feature tensor is generated to comprehensively reflect the statistical coupling characteristics and spatial sensitivity of multiple convection parameters. The sensitivity field generation layer is connected to the feature coupling and fusion layer. It is used to map and generate a real-time convection sensitivity field based on the coupled sensitivity feature tensor in the spatial grid and height layer dimensions, providing an input basis for subsequent physical consistency constraint modeling and regression analysis.

[0030] The real-time convection sensitivity field is input into the physical consistency constraint model, and real-time structured convection features are generated based on stability constraints, vertical structure constraints, and mutation triggering constraints.

[0031] In this embodiment, the real-time convection sensitivity field is fed into a physical consistency constraint model as input data to analyze the real-time convection sensitivity field in terms of spatial grid and isobaric height layer dimensions. A stability constraint sub-channel is invoked to assess the stability of the intensity distribution of the real-time convection sensitivity field based on preset thresholds for convective instability parameters and historical evolution trends. Weighted suppression processing is applied to sensitivity features that do not meet stability conditions to avoid physically unreasonable high-sensitivity responses. Finally, a vertical structure constraint sub-channel is invoked to correct the continuity and inter-layer consistency of the real-time convection sensitivity field in the vertical direction, taking into account the physical correlation between different isobaric height layers. This process weakens abnormal sensitivity characteristics caused by vertical structural discontinuities or violations of convection development mechanisms. It invokes a mutation-triggered constraint subchannel to monitor changes in sensitivity gradients in local regions of the real-time convection sensitivity field, identifies regions that meet mutation conditions, and performs local enhancement or attenuation processing on corresponding regions based on the degree of mutation to highlight mutation characteristics with potential thunderstorm and strong wind indication significance. After completing the processing of each constraint subchannel, the sensitivity characteristics processed by stability constraints, vertical structure constraints, and mutation-triggered constraints are hierarchically fused to generate real-time structured convection characteristics that are physically self-consistent and structurally continuous, for subsequent multi-convection parameter regression analysis.

[0032] Furthermore, inputting the real-time convection sensitivity field into the physical consistency constraint model includes:

[0033] The system invokes the stability constraint subchannel to assess the stability of the real-time convection sensitivity field based on the convective instability threshold and historical evolution characteristics, and performs weighted suppression processing. It also invokes the vertical structure constraint subchannel to correct the real-time convection sensitivity field in the vertical direction based on the physical correlation and gradient distribution between different height layers. The mutation trigger constraint subchannel is used to monitor mutation regions and gradient fluctuation change points in the implemented convection sensitivity field, and generates triggerable mutation features through local enhancement or attenuation processing. Based on the hierarchical fusion of all subchannels, it outputs real-time structured convection features.

[0034] First, the stability constraint subchannel is invoked to read the sensitivity distribution corresponding to the convective instability parameters in the real-time convective sensitivity field. Combined with a preset convective instability threshold and the historical evolution characteristics of the convective instability parameters, the overall stability of the current sensitivity field is evaluated. When a local region's sensitivity response exceeds the stability constraint range, weighted suppression processing is applied to the sensitivity value of the corresponding region to weaken anomalous sensitivity features generated under unstable convection backgrounds. Second, the vertical structure constraint subchannel is invoked to perform vertical consistency correction processing on the real-time convective sensitivity field based on the physical correlation between different isobaric height layers and the gradient distribution characteristics of the real-time convective sensitivity field in the vertical direction, ensuring continuity and physical rationality of the sensitivity field across height layers. Subsequently, the mutation trigger constraint subchannel is invoked to scan local regions in the real-time convective sensitivity field, monitoring mutation regions and gradient fluctuation points of sensitivity values ​​and their gradients. Based on the mutation amplitude and duration, local enhancement or attenuation processing is applied to the corresponding regions to generate mutation features with triggering significance. After processing the stability constraint subchannel, vertical structure constraint subchannel, and mutation trigger constraint subchannel, the results output by each subchannel are subjected to hierarchical fusion processing to form real-time structured convection characteristics under physical consistency constraints, which serve as input data for subsequent regression analysis and disturbance compensation.

[0035] Based on the real-time structured convection characteristics, a principal regression analysis under multiple convection parameter regression is performed to establish the principal regression results.

[0036] In this embodiment, real-time structured convection features are used as the input feature set for regression analysis. These real-time structured convection features include at least the convection sensitivity distribution features, vertical structure features, and abrupt triggering features constrained by physical consistency. Based on the contribution of different convection parameters to the formation of thunderstorms and strong winds, and combined with historical thunderstorm and strong wind sample data, the weight coefficients of each convection parameter and its corresponding structured feature in the regression analysis are dynamically determined, and the real-time structured convection features are weighted. Based on this weighting, the interaction relationships between multiple convection parameters are introduced into the regression analysis process to model higher-order coupling features, thereby constructing regression terms that reflect the physical interaction effects between convective instability, wind shear, and water vapor conditions. Based on the weighted structured features and the interaction regression terms, the main regression results characterizing the intensity or probability of thunderstorms and strong winds are generated through regression coefficient calculation and linear or nonlinear combinations, used to characterize the regression analysis output of thunderstorms and strong winds under the current convective environment.

[0037] Furthermore, based on the aforementioned real-time structured convection characteristics, a principal regression analysis under multiple convection parameter regression is performed to establish the principal regression results, including:

[0038] The weights of each convective parameter are dynamically assigned based on their contribution to thunderstorm winds, and real-time structured convective features are weighted. Regression analysis is performed on the higher-order coupling relationships between multiple convective parameters to construct physical interaction effects. The main regression results are established by calculating and combining the weighted features and interaction terms through regression coefficients.

[0039] First, based on the historical contribution of each convective parameter to the formation of thunderstorm winds and the sensitivity distribution of the current real-time structured convective features, weight coefficients are dynamically assigned to each convective parameter. Weighting is then applied to the real-time structured convective features based on these weight coefficients to highlight the convective parameter features that play a dominant role in thunderstorm wind formation. Second, after weighting, higher-order coupling relationships between multiple convective parameters are introduced into the regression analysis process. Interaction terms between different convective parameters are constructed and modeled, resulting in regression features reflecting the physical interaction effects between convective instability, wind shear, and water vapor conditions. Finally, based on the weighted structured features and interaction regression features, regression coefficients are calculated, and linear or nonlinear combinations of the regression terms are performed to generate the main regression results, which characterize the regression analysis output of thunderstorm winds under the current convective environment.

[0040] After calculating the real-time local disturbance index, the disturbance residual compensation of the main regression results is triggered, and the seasonal phase characteristics are used to dynamically adjust the weights of the main regression results after residual compensation, and the regression analysis results of thunderstorm and strong wind are output.

[0041] In this embodiment, based on real-time structured convection characteristics, the intensity changes and spatial distribution characteristics of locally highly sensitive areas are quantitatively analyzed to calculate a real-time local disturbance index reflecting the degree of local convective instability. The real-time local disturbance index is compared with a preset trigger threshold. When the real-time local disturbance index meets the disturbance triggering condition, it is determined that the current master regression result is affected by local disturbance, and a disturbance residual compensation process for the master regression result is triggered. During the disturbance residual compensation process, based on the real-time local disturbance index and the spatial distribution characteristics of the corresponding sensitive areas, the residuals in the master regression result are... The parameters are modified to weaken the bias effect of local anomalous disturbances on the main regression results. After completing the disturbance residual compensation, the seasonal phase features corresponding to the current time are extracted based on the seasonal distribution pattern of historical thunderstorm and gale events. The main regression results after residual compensation are dynamically weighted according to the seasonal phase features to match the regression analysis results with the convective dominance mechanism under different seasonal backgrounds. Finally, based on the main regression results after disturbance residual compensation and seasonal phase dynamic weight adjustment, the thunderstorm and gale regression analysis results are output and used as the basis for thunderstorm and gale warning or risk assessment.

[0042] Furthermore, calculating the real-time local disturbance index includes:

[0043] The intensity and gradient changes of sensitive areas in real-time structured convection characteristics are normalized to quantify the amplitude of local convective disturbances. The spatial concentration of disturbances is calculated by combining the distribution data of local horizontal and vertical convection characteristics. The real-time local disturbance index is generated by weighting the amplitude and concentration of local convective disturbances.

[0044] First, regions with high convection sensitivity are identified from real-time structured convection features as sensitive areas for local disturbance analysis. For these sensitive areas, the intensity values ​​of their convection features and the corresponding spatial gradient changes are normalized to eliminate the influence of differences in the dimensions of different physical quantities. The normalized results are used as the amplitude of local convective disturbances to quantify the strength of local convective activity. Second, combining the horizontal and vertical distribution data of convection features in the sensitive areas, the spatial aggregation of local disturbances is analyzed. By calculating the density of disturbance features at the spatial grid and height layers, a concentration index characterizing the spatial concentration of disturbances is obtained. Finally, based on the amplitude of local convective disturbances and the concentration index, a weighted calculation is performed according to a preset weighting rule to generate a real-time local disturbance index reflecting the intensity and spatial aggregation characteristics of local convective disturbances. This index is used to determine whether to trigger subsequent disturbance residual compensation processing.

[0045] Furthermore, it is determined whether the real-time local disturbance index meets the preset triggering conditions. If the preset triggering conditions are met, the disturbance residual compensation of the main regression result is triggered.

[0046] Specifically, the real-time local disturbance index is compared with a pre-set disturbance trigger threshold, which is set based on the statistical characteristics of historical thunderstorm and strong wind samples or operational experience. When the real-time local disturbance index is greater than or equal to the disturbance trigger threshold, it is determined that there is a significant local disturbance in the current convective environment, meeting the triggering condition for disturbance residual compensation. After determining that the preset triggering condition is met, a disturbance compensation trigger command is generated to start the disturbance residual compensation process for the main regression results, so as to correct the residual bias in the main regression results caused by local anomalous disturbances.

[0047] Furthermore, the seasonal phase characteristics are used to dynamically adjust the weights of the residual-compensated main regression results, including:

[0048] Based on the historical probability of thunderstorms and strong winds and seasonal patterns, the contribution factors of each convective parameter under different seasonal phases are calculated; the contribution factors are combined with the main regression results to adjust the relative weights of each convective parameter in the main regression results.

[0049] Based on historical thunderstorm and gale event data, this study statistically analyzes the probability distribution characteristics of thunderstorm and gale occurrences across different seasonal phases. Combining this with the evolutionary patterns of convective systems under different seasonal backgrounds, it identifies the dominant convective characteristics corresponding to each seasonal phase. Furthermore, for different seasonal phases, it calculates the relative contribution of each convective parameter to the formation process of thunderstorms and gales, generating convective parameter contribution factors reflecting seasonal differences. These contribution factors are then fused with the perturbation-recovered residuals of the main regression results. Weight adjustments are made to the regression terms corresponding to each convective parameter in the main regression results based on the contribution factors, ensuring that convective parameters contributing more significantly under the current seasonal phase have higher weights in the regression analysis results. This dynamic weight adjustment process allows the residual-compensated main regression results to adapt to the differences in convective mechanisms under different seasonal backgrounds, improving the accuracy and adaptability of thunderstorm and gale regression analysis results in cross-seasonal application scenarios.

[0050] Furthermore, the output of the thunderstorm and strong wind regression analysis results includes:

[0051] Based on the regression analysis results of thunderstorms and strong winds, early warning information is matched to establish early warning signals; the early warning signals are then used to implement early warning issuance management.

[0052] Specifically, based on the regression analysis results of thunderstorms and strong winds, and combined with pre-set thunderstorm and strong wind warning judgment rules, the regression analysis results are compared with thresholds and judged to determine the corresponding thunderstorm and strong wind risk level. A thunderstorm and strong wind warning signal is then established based on this, characterizing the probability or intensity level of thunderstorm and strong wind occurrence. After establishing the warning signal, output control is performed according to the warning release sequence and regional matching rules to complete the reporting and management of thunderstorm and strong wind warning information, enabling the regression analysis results of thunderstorms and strong winds to be directly used for business warnings or risk alerts.

[0053] In summary, the embodiments of this application have at least the following technical effects:

[0054] First, multiple convective parameters, representing convective instability, wind shear, and temporal variations of water vapor conditions, are acquired in real-time from numerical models, radiosondes, or analytical data to construct a real-time multi-convective parameter sequence. Next, this real-time multi-convective parameter sequence is mapped to a pre-constructed coupled sensitivity field model, and a real-time convective sensitivity field is generated through the coupling calculation of parameter covariance structure and gradient structure. Further, the real-time convective sensitivity field is input into a physical consistency constraint model, and real-time structured convective features are generated based on stability constraints, vertical structure constraints, and abrupt change triggering constraints. Then, based on the real-time structured convective features, a master regression analysis under multi-convective parameter regression is performed to establish the master regression results. Finally, after calculating the real-time local disturbance index, disturbance residual compensation of the master regression results is triggered, and the seasonal phase characteristics are used to dynamically adjust the weights of the residual-compensated master regression results, outputting the thunderstorm gale regression analysis results. This solves the technical problem of insufficient accuracy in predicting thunderstorm gales in existing technologies and achieves the technical effect of improving the accuracy of thunderstorm gale regression analysis results.

[0055] Example 2, based on the same inventive concept as the thunderstorm wind regression analysis method based on multiple convection parameters in the previous examples, such as... Figure 2 As shown, this application provides a thunderstorm wind regression analysis system based on multiple convection parameters, wherein the system includes:

[0056] Data acquisition module 11: Real-time acquisition of multiple convection parameters characterizing convective instability, wind shear, and temporal variations of water vapor conditions from numerical models, radiosondes, or analytical data, constructing a real-time multiple convection parameter sequence; Sensitivity field generation module 12: Mapping the real-time multiple convection parameter sequence to a pre-constructed multiple convection parameter coupled sensitivity field model, generating a real-time convection sensitivity field through coupling calculation of parameter covariance structure and gradient structure; Feature generation module 13: Inputting the real-time convection sensitivity field into a physical consistency constraint model, generating real-time structured convection features based on stability constraints, vertical structure constraints, and abrupt triggering constraints; Principal regression analysis module 14: Based on the real-time structured convection features, performing principal regression analysis under multiple convection parameter regression, establishing principal regression results; Result output module 15: After calculating the real-time local disturbance index, triggering disturbance residual compensation of the principal regression results, and dynamically adjusting the weights of the residual-compensated principal regression results using seasonal phase characteristics, outputting the thunderstorm and gale regression analysis results.

[0057] Furthermore, the sensitivity field generation module 12 is used to perform the following method:

[0058] The real-time multi-convective parameter sequence is vertically stratified according to isobaric height layers, and the local covariance matrix of the convective parameters within each height layer is calculated. The physical correlation between each height layer is obtained, and based on this physical correlation, a vertical linkage analysis is performed on the local covariance matrix to construct a cross-layer covariance matrix. Eigenvalue constraint processing is performed on the local and cross-layer covariance matrices to suppress spurious correlations caused by noise or weak disturbances, thus constructing a parameter covariance structure matrix. Local gradients are calculated along the horizontal and vertical directions for the real-time multi-convective parameter sequence, forming a gradient tensor for multiple physical quantities, including temperature gradient, wind speed gradient, and humidity gradient. Scale normalization processing is performed on the gradient tensor to construct a gradient structure matrix. The parameter covariance structure matrix and the gradient structure matrix are coupled for calculation to generate a real-time convection sensitivity field.

[0059] Furthermore, the sensitivity field generation module 12 is used to perform the following method:

[0060] The system comprises the following layers: an input layer for receiving real-time multi-convective parameter sequences; a local covariance calculation layer for calculating the local covariance matrix of convective parameters within each isobaric height layer; a cross-layer covariance evolution layer for evolving the local covariance matrix into a cross-layer covariance matrix based on the physical correlation between height layers; a gradient tensor calculation layer for calculating temperature gradient, wind speed gradient, and humidity gradient along the horizontal and vertical directions, and normalizing them into a gradient structure matrix; a feature coupling fusion layer for coupling the parameter covariance structure matrix and the gradient structure matrix to output a coupled sensitivity feature tensor; and a sensitivity field generation layer for generating a real-time convective sensitivity field based on the coupled sensitivity feature tensor.

[0061] Furthermore, the feature generation module 13 is used to perform the following method:

[0062] The system invokes the stability constraint subchannel to assess the stability of the real-time convection sensitivity field based on the convective instability threshold and historical evolution characteristics, and performs weighted suppression processing. It also invokes the vertical structure constraint subchannel to correct the real-time convection sensitivity field in the vertical direction based on the physical correlation and gradient distribution between different height layers. The mutation trigger constraint subchannel is used to monitor mutation regions and gradient fluctuation change points in the implemented convection sensitivity field, and generates triggerable mutation features through local enhancement or attenuation processing. Based on the hierarchical fusion of all subchannels, it outputs real-time structured convection features.

[0063] Furthermore, the main regression analysis module 14 is used to perform the following methods:

[0064] The weights of each convective parameter are dynamically assigned based on their contribution to thunderstorm winds, and real-time structured convective features are weighted. Regression analysis is performed on the higher-order coupling relationships between multiple convective parameters to construct physical interaction effects. The main regression results are established by calculating and combining the weighted features and interaction terms through regression coefficients.

[0065] Furthermore, the result output module 15 is used to perform the following method:

[0066] The intensity and gradient changes of sensitive areas in real-time structured convection characteristics are normalized to quantify the amplitude of local convective disturbances. The spatial concentration of disturbances is calculated by combining the distribution data of local horizontal and vertical convection characteristics. The real-time local disturbance index is generated by weighting the amplitude and concentration of local convective disturbances.

[0067] Furthermore, the result output module 15 is used to perform the following method:

[0068] Determine whether the real-time local disturbance index meets the preset triggering conditions. If the preset triggering conditions are met, then trigger the disturbance residual compensation of the main regression result.

[0069] Furthermore, the result output module 15 is used to perform the following method:

[0070] Based on the historical probability of thunderstorms and strong winds and seasonal patterns, the contribution factors of each convective parameter under different seasonal phases are calculated; the contribution factors are combined with the main regression results to adjust the relative weights of each convective parameter in the main regression results.

[0071] Furthermore, the result output module 15 is used to perform the following method:

[0072] Based on the regression analysis results of thunderstorms and strong winds, early warning information is matched to establish early warning signals; the early warning signals are then used to implement early warning issuance management.

[0073] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A regression analysis method for thunderstorm winds based on multiple convection parameters, characterized in that, The method includes: Real-time acquisition of multiple convective parameters that characterize convective instability, wind shear, and temporal variations of water vapor conditions from numerical models, radiosondes, or analytical data, and construction of real-time multiple convective parameter sequences; The real-time multi-convective parameter sequence is mapped to a pre-constructed multi-convective parameter coupled sensitivity field model, and the real-time convection sensitivity field is generated by coupling calculation of parameter covariance structure and gradient structure. The real-time convection sensitivity field is input into the physical consistency constraint model, and real-time structured convection features are generated based on stability constraints, vertical structure constraints, and mutation triggering constraints. Based on the real-time structured convection characteristics, perform principal regression analysis under multiple convection parameter regression and establish principal regression results; After calculating the real-time local disturbance index, the disturbance residual compensation of the main regression result is triggered, and the seasonal phase characteristics are used to dynamically adjust the weight of the main regression result after residual compensation, and the regression analysis results of thunderstorm wind are output. Mapping the real-time multi-pair flow parameter sequence to a pre-constructed multi-pair flow parameter coupling sensitivity field model includes: The real-time multi-convective parameter sequence is vertically layered according to isobaric height layers, and the local covariance matrix of the convection parameters in each height layer is calculated respectively. Obtain the physical correlation between each height layer, and perform vertical linkage analysis on the local covariance matrix based on the physical correlation to construct a cross-layer covariance matrix; The local covariance matrix and the cross-layer covariance matrix are used to perform eigenvalue constraint processing to suppress spurious correlations caused by noise or weak perturbations, and to construct a parametric covariance structure matrix. Local gradients are calculated along the horizontal and vertical directions for real-time multi-convection parameter sequences to form a gradient tensor of multiple physical quantities, including temperature gradient, wind speed gradient, and humidity gradient. Perform scale normalization on the gradient tensor to construct the gradient structure matrix; The parameter covariance structure matrix and gradient structure matrix are coupled and calculated to generate a real-time convection sensitivity field.

2. The method for regression analysis of thunderstorm winds based on multiple convection parameters as described in claim 1, characterized in that, The multi-convection parameter coupling sensitivity field model is a multi-level evolutionary network model, including: The input layer is used to receive real-time multi-pair flow parameter sequences; The local covariance calculation layer is used to calculate the local covariance matrix of convection parameters within each isobaric height layer. The cross-layer covariance evolution layer is used to evolve the local covariance matrix into a cross-layer covariance matrix based on the high inter-layer physical correlation. The gradient tensor calculation layer is used to calculate the temperature gradient, wind speed gradient, and humidity gradient along the horizontal and vertical directions, and normalize them into a gradient structure matrix. The feature coupling fusion layer is used to couple the calculation of the parameter covariance structure matrix and the gradient structure matrix, and output the coupling sensitivity feature tensor. A sensitivity field generation layer is used to generate a real-time convection sensitivity field based on the coupled sensitivity feature tensor.

3. The method for regression analysis of thunderstorm winds based on multiple convection parameters as described in claim 1, characterized in that, The real-time convection sensitivity field is input into the physical consistency constraint model, including: The stability constraint subchannel is invoked to evaluate the stability of the real-time convection sensitivity field based on the convective instability threshold and historical evolution characteristics, and weighted suppression processing is performed. The vertical structure constrains the sub-channels and corrects the real-time convection sensitivity field in the vertical direction based on the physical correlation and gradient distribution between different height layers. The mutation trigger constraint subchannel is used to monitor mutation regions and gradient fluctuation change points in the implemented convection sensitivity field, and generates triggerable mutation features through local enhancement or attenuation processing. Based on hierarchical fusion of all sub-channels, real-time structured convection characteristics are output.

4. The method for regression analysis of thunderstorm winds based on multiple convection parameters as described in claim 1, characterized in that, Based on the aforementioned real-time structured convection characteristics, a principal regression analysis under multiple convection parameter regression is performed to establish the principal regression results, including: The weights of each convective parameter are dynamically assigned based on their contribution to thunderstorm winds, and real-time weighted processing of convective characteristics is performed. Regression analysis was performed on the higher-order coupling relationships among multiple convection parameters to construct physical interaction effects; The main regression results are established by calculating the regression coefficients and combining the weighted features and interaction terms.

5. The method for regression analysis of thunderstorm winds based on multiple convection parameters as described in claim 1, characterized in that, Calculating the real-time local disturbance index includes: The intensity and gradient changes of sensitive regions in real-time structured convection features are normalized to quantify the amplitude of local convection disturbances. By combining the distribution data of local horizontal and vertical convection characteristics, the spatial concentration of disturbances is calculated; A real-time local disturbance index is generated by weighting the magnitude and concentration of local convective disturbances.

6. The method for regression analysis of thunderstorm winds based on multiple convection parameters as described in claim 5, characterized in that, Determine whether the real-time local disturbance index meets the preset triggering conditions. If the preset triggering conditions are met, then trigger the disturbance residual compensation of the main regression result.

7. The method for regression analysis of thunderstorm winds based on multiple convection parameters as described in claim 1, characterized in that, Dynamic weight adjustment of the residual-compensated main regression results using seasonal phase characteristics includes: Based on the historical probability of thunderstorms and strong winds and seasonal patterns, the contribution factors of each convective parameter under different seasonal phases are calculated. The contribution factors are combined with the main regression results, and the relative weights of each convection parameter in the main regression results are adjusted.

8. The method for regression analysis of thunderstorm winds based on multiple convection parameters as described in claim 1, characterized in that, Output the regression analysis results of thunderstorms and strong winds, including: Based on the regression analysis results of the thunderstorm and strong wind, early warning information is matched to establish early warning signals; The aforementioned warning signal is used to manage early warning issuance.

9. A thunderstorm wind regression analysis system based on multiple convection parameters, characterized in that, The system is used to implement the thunderstorm wind regression analysis method based on multiple convection parameters as described in any one of claims 1-8, the system comprising: Data acquisition module: Real-time acquisition of multiple convective parameters that characterize convective instability, wind shear, and water vapor condition temporal variations in numerical models, radiosondes, or analytical data, and construction of real-time multiple convective parameter sequences; Sensitivity field generation module: maps the real-time multi-convective parameter sequence to a pre-constructed multi-convective parameter coupled sensitivity field model, and generates a real-time convection sensitivity field through the coupling calculation of parameter covariance structure and gradient structure; Feature generation module: Inputs the real-time convection sensitivity field into the physical consistency constraint model, and generates real-time structured convection features based on stability constraints, vertical structure constraints, and abrupt triggering constraints; Main Regression Analysis Module: Based on the real-time structured convection characteristics, performs main regression analysis under multiple convection parameter regression and establishes main regression results; Results output module: After calculating the real-time local disturbance index, it triggers the disturbance residual compensation of the main regression results, and uses seasonal phase characteristics to dynamically adjust the weights of the main regression results after residual compensation, and outputs the regression analysis results of thunderstorm and strong wind.

Images