A method and device for numerical simulation of space weather

By monitoring and building numerical simulation models in real-time in space weather detection instruments, outputting data sets of different accuracy and training super-score models, the problems of insufficient resolution of traditional simulation data and high computational complexity are solved, and a higher precision space weather simulation is achieved Data meets business needs.

CN118429189BActive Publication Date: 2024-11-12NAT SATELLITE METEOROLOGICAL CENT
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410594045.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-14
Publication Date
2024-11-12
Estimated Expiration
2044-05-14

AI Technical Summary

Technical Problem

Due to the limitations of computing resources and algorithms, the traditional numerical simulation model of space weather is insufficient in the simulation data resolution and high computational complexity, which cannot meet the existing space weather business needs.

Method used

Data is obtained through real-time monitoring of space weather detection instruments, numerical simulation models are constructed, and data sets of different accuracy are output under the same conditions. These data sets are used to train super-score models to improve the accuracy of simulation data.

Benefits of technology

It effectively improves the accuracy of space weather simulation data, provides more accurate data support, and meets the needs of space weather forecasting and analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118429189B_ABST
    Figure CN118429189B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and device for numerical simulation of space weather, which belongs to the field of data processing technology, wherein the method comprises: performing real-time monitoring through a space weather detection instrument to obtain numerical simulation data of space weather; constructing a numerical simulation model based on the numerical simulation data of space weather; using the numerical simulation model to output a first precision data set and a second precision data set according to the same space weather conditions; performing data training based on the first precision data set and the second precision data set to construct a super-resolution model; using the numerical simulation data of space weather to obtain target high-precision simulation data through the super-resolution model. The present application solves the technical problem in the prior art that the resolution of simulation data is insufficient and the computational complexity is high, resulting in the inability of the numerical simulation model of space weather to meet the existing space weather business needs.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and device for numerical simulation of space weather. Background Art

[0002] Space weather numerical simulation is an important weather forecasting technology. It uses numerical models to model and simulate space environments such as the thermosphere, atmosphere and ionosphere to predict changes in space weather phenomena. However, traditional space weather numerical simulations often have problems such as insufficient resolution and poor prediction accuracy under the premise of ensuring space weather business needs, making it difficult to meet actual needs. Therefore, using deep learning technology to improve and optimize space weather numerical simulations has become one of the research hotspots.

[0003] Traditional space weather numerical simulation models are often limited by computing resources and algorithm models, resulting in low resolution of generated simulation data that cannot meet existing space weather business needs; some space weather numerical simulation methods require a lot of computing resources and time, have high computational complexity, and are difficult to meet real-time and efficiency requirements.

[0004] In summary, the present application is used to solve the technical problems in the prior art that the simulation data resolution is insufficient and the calculation complexity is high, resulting in the inability of the space weather numerical simulation model to meet the existing space weather business needs. Summary of the invention

[0005] The present application provides a method and device for numerical simulation of space weather, aiming to solve the technical problems in the prior art of insufficient simulation data resolution and high computational complexity, which result in the inability of space weather numerical simulation models to meet existing space weather business needs.

[0006] In view of the above problems, the present application provides a method and device for numerical simulation of space weather.

[0007] The first aspect disclosed in the present application provides a method for numerical simulation of space weather, the method comprising: performing real-time monitoring through a space weather detection instrument to obtain space weather numerical simulation data; constructing a numerical simulation model based on the space weather numerical simulation data; using the numerical simulation model to output a first precision data set and a second precision data set under the same space weather conditions; performing data training based on the first precision data set and the second precision data set to construct a super-resolution model; using the space weather numerical simulation data, and obtaining target high-precision simulation data through the super-resolution model.

[0008] Another aspect disclosed in the present application provides a space weather numerical simulation device, which includes: a simulation data acquisition module: used to perform real-time monitoring through a space weather detection instrument to acquire space weather numerical simulation data; a simulation model construction module: used to construct a numerical simulation model based on the space weather numerical simulation data; a data set output module: used to output a first precision data set and a second precision data set using the numerical simulation model under the same space weather conditions; a super-resolution model construction module: used to perform data training based on the first precision data set and the second precision data set to construct a super-resolution model; a high-precision simulation data acquisition module: used to use the space weather numerical simulation data to obtain target high-precision simulation data through the super-resolution model.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] This application uses real-time monitoring of space weather detection instruments to obtain space weather numerical simulation data, and then construct a numerical simulation model. Under the same conditions, the model can output data sets of different precisions, and use these data sets to train super-resolution models. The super-resolution model can upgrade the lower-precision simulation data to the target high-precision level. It solves the technical problems in the prior art that the simulation data resolution is insufficient and the calculation complexity is high, resulting in the inability of the space weather numerical simulation model to meet the existing space weather business needs. It effectively improves the accuracy of space weather simulation data and provides more accurate data support for space weather prediction and analysis.

[0011] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A flow chart of a method for numerical simulation of space weather is provided for an embodiment of the present application.

[0013] Figure 2 A flow chart of constructing a numerical simulation model in a method for numerical simulation of space weather is provided for an embodiment of the present application.

[0014] Figure 3 A structural schematic diagram of a space weather numerical simulation device is provided for an embodiment of the present application.

[0015] Explanation of the reference numerals: module for obtaining simulation data 11, module for building simulation model 12, module for outputting data set 13, module for building super-resolution model 14, module for obtaining high-precision simulation data. DETAILED DESCRIPTION

[0016] The overall idea of ​​the technical solution provided by this application is as follows:

[0017] The embodiment of the present application provides a method and device for numerical simulation of space weather. First, real-time monitoring is performed through a space weather detection instrument to collect real-time space weather data, and numerical simulation data of space weather is obtained accordingly. Then, a numerical simulation model is constructed based on the collected space weather numerical simulation data. This model can simulate and predict various parameters and states of space weather. Subsequently, according to the set equal space weather conditions, the constructed numerical simulation model is used for simulation calculation. Through multi-level configuration and single-level configuration, a first precision data set and a second precision data set are output respectively. The first precision data set has a higher spatial and temporal resolution, while the second precision data set is relatively low. Afterwards, a super-resolution model is constructed and trained based on the first precision data set and the second precision data set. The second precision data set is used as the input data of the super-resolution model, and the first precision data set is used as the target output data. By defining a loss function and an optimizer, the super-resolution model is iteratively trained until the loss difference is less than a preset threshold, and the construction of the super-resolution model is completed. Finally, new space weather numerical simulation data is used to obtain target high-precision simulation data through the trained super-resolution model.

[0018] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically described below in conjunction with the drawings in the specification.

[0019] Embodiment 1

[0020] like Figure 1 As shown, an embodiment of the present application provides a method for numerical simulation of space weather, the method comprising:

[0021] Step S100: Perform real-time monitoring through space weather detection instruments to obtain space weather numerical simulation data.

[0022] Specifically: start the space weather detection instrument to conduct real-time monitoring of the target space area, wherein the space weather detection instrument includes: a solar observer, an earth magnetosphere detector, and a cosmic ray detector; then collect various parameters and data of space weather, thereby obtaining multiple real-time weather elements as the basis for numerical simulation of space weather. Wherein, the space weather numerical simulation data includes but is not limited to data information such as solar radiation flux, plasma density, and temperature. Furthermore, it also includes converting the data format of multiple real-time weather elements, screening according to the conversion results, and generating target real-time weather elements; transmitting the target real-time weather elements to the space weather data center for simulation calculation to generate space weather numerical simulation data.

[0023] This step conducts real-time monitoring through space weather detection instruments to obtain space weather numerical simulation data, providing real-time and accurate data support for the entire monitoring and forecasting process.

[0024] Step S200: constructing a numerical simulation model based on the space weather numerical simulation data.

[0025] Specifically: A numerical simulation model is constructed based on the space weather numerical simulation data to simulate and predict the dynamic changes of space weather. The construction process includes: determining the first hyperparameter based on the model framework, updating the first hyperparameter through the back propagation algorithm, generating the second hyperparameter, data preprocessing, model training, monitoring performance indicator data, performance evaluation, iterative optimization and other steps.

[0026] This step is based on mathematical equations and algorithms, and uses computer programs to solve approximate solutions of mathematical models to reveal the inherent laws and characteristics of the system. It is applied to meteorology, geology, fluid dynamics, electromagnetism, structural mechanics and other fields, and is mainly used to simulate and predict physical phenomena, engineering problems or the dynamic processes of scientific research.

[0027] Step S300: outputting a first precision data set and a second precision data set using the numerical simulation model according to the same spatial weather conditions.

[0028] Specifically, the same space weather conditions refer to space weather conditions with the same or similar physical parameters and characteristics in the same or similar space environment. Under this condition, a first precision data set and a second precision data set are output, and the first precision data set and the second precision data set represent simulation results of different precisions. Among them, the first precision data set represents a high-precision simulation result; the second precision data set is used to represent a low-precision simulation result.

[0029] Step S400: Perform data training based on the first precision data set and the second precision data set to build a super-resolution model.

[0030] Specifically: Use the first precision data set and the second precision data set for training to build a super-resolution model, which is to restore the lost detail information to a certain extent by learning the mapping relationship of the image to improve the visual effect of the image and convert the low-resolution image into a high-resolution image. In this application, it is used to improve the simulation accuracy. The construction process includes: data preparation, that is, the second precision data set is used as the input data of the super-resolution model, and the first precision data set is used as the target output data, wherein the second precision data set (low-resolution data) is used for the training of the super-resolution model, and the first precision data set (high-resolution data) is used to verify whether the deep learning results based on the LIIF algorithm super-resolution are correct.

[0031] Step S500: using the space weather numerical simulation data, and obtaining target high-precision simulation data through the super-resolution model.

[0032] Specifically: the super-resolution model is used to improve the accuracy of space weather numerical simulation data, that is, the super-resolution model will receive low-precision input data and generate high-resolution or high-precision output data based on the mapping relationship learned during its training process, thereby obtaining the target high-precision simulation data, which is superior to the original input data in resolution and detail. The generated high-precision simulation data will be used in actual scenarios such as space weather forecasting, climate simulation, and scientific research to provide more accurate and detailed information.

[0033] Furthermore, in step S100 of the present application, real-time monitoring is performed by a space weather detection instrument to obtain space weather numerical simulation data, which also includes: starting the space weather detection instrument to perform real-time monitoring of the target space area to obtain multiple real-time weather elements; performing data format conversion on the multiple real-time weather elements, screening them according to the conversion results, and generating target real-time weather elements; transmitting the target real-time weather elements to a space weather data center for simulation calculation to generate the space weather numerical simulation data.

[0034] Specifically: according to specific monitoring needs and target space areas (such as the sun, the earth's magnetosphere, the ionosphere, etc.), start the corresponding space weather detection instruments, including solar observers, earth's magnetosphere detectors, and cosmic ray detectors. Space weather detection instruments capture and measure physical parameters in space through specific sensors and receivers, such as solar radiation, magnetic field intensity, plasma density, etc. Real-time monitoring ensures the real-time and accuracy of data.

[0035] Furthermore, since different space weather detection instruments may use different data formats and transmission standards, it is necessary to convert the formats of multiple real-time weather elements received, including: decoding the original data, converting it into a common data format (such as CSV, NetCDF, etc.), unit conversion (such as conversion from the International System of Units to other units) and coordinate conversion (such as conversion from instrument coordinates to geographic coordinates or physical space coordinates).

[0036] After the data format conversion, the data needs to be screened to generate the target real-time weather elements. The specific screening process includes: Data quality check: Check the integrity, accuracy and consistency of the data. Data points with obvious errors or anomalies may be removed or marked. Time range selection: Select data within a specific time range according to the needs of research or application. For example, if a specific solar storm event is being studied, data from a period of time before and after the event may be selected. Physical parameter selection: Select specific physical parameters according to the needs of research or application. For example, if the changes in the earth's magnetic field are being studied, parameters such as magnetic field strength and magnetic field direction may be selected. Space region selection: Select data from a specific space region according to the needs of research or application. For example, if the changes in the ionosphere in a specific area are being studied, data from that area may be selected. After screening, the target real-time weather elements are obtained, which include data of space weather elements such as solar activity, the state of the earth's magnetosphere, and the intensity of cosmic rays.

[0037] Finally, the target real-time weather elements are transmitted to the space weather data center for simulation calculation. The space weather data center is equipped with a high-performance computer system and professional space weather numerical simulation software. That is, the target real-time weather elements are simulated and calculated to generate space weather numerical simulation data. The space weather numerical simulation data includes the distribution and changes of space weather elements such as temperature, density, and speed, as well as predicted space weather events and phenomena.

[0038] Based on this step, the accuracy and real-time nature of the space weather numerical simulation data are ensured through real-time monitoring, data format conversion, screening and simulation calculation, etc. It is used to understand the changing laws of space weather and predict potential space weather events.

[0039] Further, such as Figure 2 As shown, in step S200 of the present application, a numerical simulation model is constructed based on the space weather numerical simulation data, and it also includes: determining a first hyperparameter of the numerical simulation model based on a model framework, updating the first hyperparameter through a back propagation algorithm, and generating a second hyperparameter; performing data preprocessing on the space weather numerical simulation data, performing data division according to the preprocessing result in combination with the second hyperparameter, and generating a training data set, a supervision data set, and a test data set; training the numerical simulation model using the training data set, and monitoring the performance indicator data of the training data set and the supervision data set in the numerical simulation model; performing performance evaluation on the numerical simulation model using the test data set, and generating a performance evaluation result; iteratively optimizing the numerical simulation model based on the performance indicator data in combination with the performance evaluation result until the numerical simulation model reaches the performance indicator data for output.

[0040] Specifically: The model framework is built based on Swin Transformer, which is a Transformer-based architecture designed for visual tasks (such as image classification, object detection, semantic segmentation, etc.). It adopts a hierarchical construction method, similar to the pyramid structure in convolutional neural networks (CNN). First, the input image is divided into a series of small blocks (or patches), and then these small blocks are processed using multiple layers of Transformer layers. In each layer, Swin Transformer performs self-attention calculations on patches in non-overlapping window areas to capture local features in the image. In addition, Swin Transformer also introduces a cross-window connection mechanism to fuse features at different scales.

[0041] Based on the model framework, the initial first hyperparameter is determined, and the first hyperparameter includes the initial set weights, biases and other values. On this basis, the model is trained, and the first hyperparameter is updated by the back propagation algorithm during the training process to generate the second hyperparameter to minimize the loss function. Among them, the back propagation algorithm (Backpropagation Algorithm), referred to as the BP algorithm, is a commonly used algorithm for training neural networks. Its principle is to reversely calculate the gradient of the weight and bias of each neuron layer by layer according to the error between the input and the expected output of the training sample, and then update the network parameters to reduce the error.

[0042] The space weather numerical simulation data is further preprocessed to meet the model input requirements. Then, data division is performed based on the preprocessing results combined with the second hyperparameter to generate a training data set, a supervision data set, and a test data set. The division ratio can be set according to the second hyperparameter (such as 70% for training, 15% for verification, and 15% for testing). The training data set is a data set containing a large number of known inputs and outputs. In supervised learning, the model uses the training data set to learn the mapping relationship from input to output. The supervised data set is a data set with outputs, which is used to train the model to predict the label of new data. Supervised learning is a machine learning paradigm in which the model learns the mapping relationship from input data to output data. Therefore, the supervised data set is a data set containing this mapping relationship. The test data set is used to evaluate the performance of the model.

[0043] Use the training data set to train the numerical simulation model. During the training process, use the back propagation algorithm to update the weights and biases of the model. Monitor the performance indicator data of the training data set and the supervision data set in the numerical simulation model, such as loss value, accuracy, etc.

[0044] The performance of the numerical simulation model is evaluated using a test data set, and performance indicators (also called evaluation indicators or loss functions) are used to compare the prediction results generated by the model with the actual results in the test data set, where the performance indicators include accuracy, precision, recall, F1 score, mean square error, etc., to generate performance evaluation results.

[0045] Finally, the numerical simulation model is iteratively optimized based on the performance indicator data combined with the performance evaluation results. The iterative optimization method includes adjusting the model structure (such as increasing or decreasing the number of layers, changing the activation function, etc.), adjusting the model parameters (such as learning rate, batch size, etc.), using different optimization algorithms (such as SGD, Adam, etc.), performing regularization (such as L1, L2 regularization) to prevent overfitting, introducing data enhancement technology to increase the generalization ability of the model, etc. Use the same test data set to evaluate the optimized model again and record the new performance indicator data. Repeat the above steps to continuously iterate and optimize the model until the model reaches or approaches the set performance indicator target.

[0046] Furthermore, in the method described in the present application, data preprocessing of space weather numerical simulation data also includes: performing wavelet decomposition on the numerical simulation signal of the space weather numerical simulation data through a wavelet transform formula to obtain wavelet coefficients of the numerical simulation signal; performing threshold quantization based on the wavelet coefficients of the numerical simulation signal to determine the wavelet selection threshold of the numerical simulation signal; truncating the wavelet coefficients of the numerical simulation signal according to the wavelet selection threshold of the numerical simulation signal, setting the noise signal less than the wavelet selection threshold of the numerical simulation signal to zero, and obtaining valid signal information greater than the wavelet selection threshold of the numerical simulation signal; filtering and reconstructing the valid signal information to obtain the preprocessing result.

[0047] Specifically: the numerical simulation signal of the space weather numerical simulation data is decomposed by wavelet transform formula. The wavelet transform can decompose the signal into components of different frequencies and time scales, and select appropriate wavelet basis functions and decomposition layers to capture the key features of the signal. Thus, the wavelet coefficients of the numerical simulation signal are obtained; specifically, the wavelet coefficients include real and imaginary parts, where the real part represents the amplitude (amplitude) of the signal at different scales and positions, and the imaginary part represents the phase information. The amplitude of the wavelet coefficient reflects the strength of the signal at different scales and positions, which can be understood as the local energy of the signal.

[0048] According to the wavelet coefficients of the numerical simulation signal, a suitable threshold is determined. This threshold is usually set based on the noise level of the signal and the required denoising degree. The threshold quantization can be based on a soft threshold or a hard threshold method. The soft threshold method shrinks the wavelet coefficients to a value near zero, while the hard threshold method directly sets the wavelet coefficients less than the threshold to zero.

[0049] According to the numerical simulation signal wavelet selection threshold, the numerical simulation signal wavelet coefficients are truncated, and the wavelet coefficients less than the threshold (usually considered as noise signals) are set to zero, while the wavelet coefficients greater than the threshold (effective signal information) are retained. Based on this step, the noise in the signal can be effectively removed while retaining the main features of the signal.

[0050] Finally, the effective signal information is filtered and reconstructed, that is, the wavelet coefficients processed by threshold value are subjected to inverse wavelet transform to reconstruct the signal. The effective signal information after noise removal is recombined into a continuous signal to obtain the preprocessing result.

[0051] Furthermore, in the method described in this application, the wavelet transform formula is:

[0052]

[0053] Among them, W t (a, b) are the wavelet coefficients of the numerical simulation signal, which are used to characterize the characteristics of the numerical simulation signal f(t) at scale a and time position b. f(t) is the numerical simulation signal, a is the scale parameter, which is used to control the expansion and contraction of the wavelet function, b is the translation parameter, which is used to control the translation of the wavelet function on the time axis, t represents the position of the numerical simulation signal on the time axis, ψ * (t) is the complex conjugate of ψ(t), and ψ(t) is the mother wavelet function.

[0054] Specifically: the expansion and contraction of the wavelet function means that the larger the scale, the wider the wavelet function, corresponding to the low-frequency part of the signal; the smaller the scale, the narrower the wavelet function, corresponding to the high-frequency part of the signal; the mother wavelet function has an oscillating waveform with zero mean and finite energy.

[0055] Furthermore, in step S300 of the present application, the numerical simulation model is used to output a first precision data set and a second precision data set according to the same spatial weather conditions, and it also includes: setting the same spatial weather conditions based on multiple real-time weather elements; according to the same spatial weather conditions, multi-parameter configuration is performed through the numerical simulation model, and the multi-parameters include network resolution parameters, time step parameters, and physical simulation parameters; the first precision data set is output by performing multi-level configuration of the network resolution parameters, the time step parameters, and the physical simulation parameters; the second precision data set is output by performing single-level configuration of the network resolution parameters, the time step parameters, and the physical simulation parameters.

[0056] Specifically, the multiple real-time weather elements refer to temperature, humidity, wind speed, wind direction, air pressure, solar radiation, etc. Based on this, one or more equivalent spatial weather conditions are determined. The equivalent spatial weather conditions reflect the real atmospheric conditions as much as possible to ensure the accuracy of the simulation results.

[0057] According to the same spatial weather conditions, multi-parameter configuration is performed through the numerical simulation model, and the multi-parameters include network resolution parameters, time step parameters, and physical simulation parameters; wherein the network resolution parameters are used to control the size and number of grids in the simulation, affecting the spatial resolution of the simulation, the time step parameters are used to control the time step of the simulation, that is, the time interval of each iteration, affecting the time resolution of the simulation, and the physical simulation parameters include parameterization schemes of various physical processes.

[0058] By performing multi-level configuration on the network resolution parameters, time step parameters, and physical simulation parameters, the multi-level configuration refers to finer grid division, higher calculation accuracy, or more complex physical processes. Multi-level configuration means that the above parameters can be set at different levels according to different accuracy requirements or computing resource limitations. For example, when high-precision results are required, a smaller time step and a higher grid resolution can be used; when resources are limited, a larger time step and a lower grid resolution can be used. Output the first precision data set. Through multi-level configuration, a first precision data set can be output, which has a higher spatial resolution and temporal resolution and can more accurately reflect the weather conditions in the simulation area.

[0059] Furthermore, if you need to obtain simulation results with lower precision, you need to configure the network resolution parameters, time step parameters and physical simulation parameters at a single level. Including: Network resolution parameters: select a lower network resolution to reduce the amount of calculation. Time step parameters: select a larger time step to speed up the simulation. Physical simulation parameters: select simpler physical simulation parameters to reduce the complexity of calculation. Through single-level configuration, you can output a second precision data set, which has high calculation efficiency and is suitable for occasions where precision is not required.

[0060] Furthermore, in step S400 of the present application, data training is performed based on the first precision data set and the second precision data set to construct a super-resolution model, which also includes: using the second precision data set as input data of the super-resolution model, and using the first precision data set as target output data; defining a loss function for the super-resolution model, and obtaining a loss difference through the loss function; based on the input data, using a stochastic gradient descent optimizer to iteratively train the super-resolution model until the loss difference is less than a preset threshold, and outputting the super-resolution model.

[0061] Specifically: the second precision data set is used as the input data of the super-resolution model, and the first precision data set is used as the target output data; the first precision data set: this is a data set with higher resolution and accuracy, used as the target output data for super-resolution model training. The second precision data set: this is a data set with lower resolution, used as the input data of the super-resolution model. It is low-resolution and high-resolution data simulated by TIEGCM, which is a global circulation model of thermosphere and ionosphere electrodynamics. It is mainly used to simulate and study the thermosphere and ionosphere dynamics in the earth's atmosphere. TIEGCM can calculate the three-dimensional global, full-current system ionospheric current distribution, and then calculate the ionospheric magnetic field from the ionospheric current distribution. In addition, it can also simulate and predict the changing characteristics of the thermosphere atmospheric density, and provide security services for the safety of low-orbit spacecraft in orbit. When using magnetic satellites to establish a geomagnetic field model, TIEGCM can eliminate the interference of the ionospheric magnetic field from the observation data and provide a more accurate geomagnetic field model. The first precision data set and the second precision data set obtained by TIEGCM must ensure the uniform data format and perform preprocessing, including removing outliers and normalization.

[0062] A loss function is further defined for the super-resolution model, and the loss function is used to quantify the gap between the output predicted by the super-resolution model and the true target (i.e., the first precision data set). The loss function includes mean square error (MSE), mean absolute error (MAE), and structural similarity index (SSIM), etc. The loss difference is obtained through the loss function, and the loss difference refers to the difference between the high-precision data output by the super-resolution model and the actual high-precision data.

[0063] Finally, based on the input data, the super-resolution model is iteratively trained using a stochastic gradient descent optimizer. Before the training begins, the weights and bias parameters of each layer of the super-resolution model need to be initialized, the second precision data set (low-resolution data) is input into the super-resolution model, the predicted high-resolution output is calculated by the forward propagation of the model, the loss difference between the predicted output and the first precision data set (real high-resolution data) is calculated using a defined loss function, the gradient of the loss function with respect to the model parameters is calculated by a back-propagation algorithm, and the parameters of the model are updated according to the calculated gradient using stochastic gradient descent (SGD) or other optimizers (such as Adam, RMSprop, etc.), and the steps of forward propagation, loss calculation, back-propagation, and parameter update are repeated until the loss difference is less than a preset threshold or the specified number of training rounds is reached. The process stops when the loss difference is less than a preset threshold, and the super-resolution model is output to improve the low-precision data to a higher level of accuracy to meet specific application requirements.

[0064] In summary, the space weather numerical simulation method provided by the embodiment of the present application has the following technical effects:

[0065] 1. Through real-time monitoring of space weather detection instruments, we can obtain space weather numerical simulation data and then build a numerical simulation model. Under the same conditions, the model can output data sets of different precisions and use these data sets to train super-resolution models. Super-resolution models can improve the simulation data with lower precision to the target high-precision level. This effectively improves the precision of space weather simulation data and provides more accurate data support for space weather prediction and analysis.

[0066] 2. Through real-time monitoring, data format conversion, screening and simulation calculation, the accuracy and real-time nature of space weather numerical simulation data are ensured, which is used to understand the changing laws of space weather and predict potential space weather events.

[0067] 3. Through multi-level configuration, the first precision dataset can be output, which has high spatial resolution and temporal resolution, and can more accurately reflect the weather conditions in the simulation area; Through multi-level configuration, the first precision dataset can be output, which has high spatial resolution and temporal resolution, and can more accurately reflect the weather conditions in the simulation area.

[0068] Embodiment 2

[0069] Based on the same inventive concept as the method for numerical simulation of space weather in the aforementioned embodiment, Figure 3 As shown, an embodiment of the present application provides a space weather numerical simulation device, the device comprising:

[0070] Acquisition simulation data module 11: used for real-time monitoring through space weather detection instruments to acquire space weather numerical simulation data;

[0071] A simulation model building module 12 is used to build a numerical simulation model based on the space weather numerical simulation data;

[0072] Dataset output module 13: used to output a first precision dataset and a second precision dataset using the numerical simulation model according to the same spatial weather conditions;

[0073] A super-resolution model building module 14 is configured to perform data training based on the first precision data set and the second precision data set to build a super-resolution model;

[0074] The module 14 for obtaining high-precision simulation data is used to obtain target high-precision simulation data by using the space weather numerical simulation data through the super-resolution model.

[0075] Furthermore, the device is also used for:

[0076] Starting the space weather detection instrument to monitor the target space area in real time to obtain multiple real-time weather elements;

[0077] Convert the data format of the multiple real-time weather elements, filter them according to the conversion results, and generate target real-time weather elements;

[0078] The target real-time weather elements are transmitted to a space weather data center for simulation calculation to generate the space weather numerical simulation data.

[0079] Furthermore, the device is also used for:

[0080] Determine a first hyperparameter of the numerical simulation model based on the model framework, update the first hyperparameter through a back propagation algorithm, and generate a second hyperparameter;

[0081] Performing data preprocessing on the space weather numerical simulation data, dividing the data according to the preprocessing result combined with the second hyperparameter, and generating a training data set, a supervision data set, and a test data set;

[0082] Using the training data set to train the numerical simulation model, and monitoring the performance indicator data of the training data set and the supervision data set in the numerical simulation model;

[0083] Using the test data set to perform performance evaluation on the numerical simulation model to generate a performance evaluation result;

[0084] The numerical simulation model is iteratively optimized based on the performance indicator data in combination with the performance evaluation result until the numerical simulation model reaches the performance indicator data for output.

[0085] Furthermore, the device is also used for:

[0086] Performing wavelet decomposition on the numerical simulation signal of the space weather numerical simulation data using a wavelet transform formula to obtain wavelet coefficients of the numerical simulation signal;

[0087] Perform threshold quantization according to the wavelet coefficients of the numerical simulation signal to determine the wavelet selection threshold of the numerical simulation signal;

[0088] According to the numerical simulation signal wavelet selection threshold, the wavelet coefficients of the numerical simulation signal are intercepted, the noise signal less than the numerical simulation signal wavelet selection threshold is set to zero, and the effective signal information greater than the numerical simulation signal wavelet selection threshold is obtained;

[0089] The effective signal information is filtered and reconstructed to obtain the preprocessing result.

[0090] Furthermore, the device is also used for:

[0091] The wavelet transform formula is:

[0092]

[0093] Among them, W t (a, b) are the wavelet coefficients of the numerical simulation signal, which are used to characterize the characteristics of the numerical simulation signal f(t) at scale a and time position b. f(t) is the numerical simulation signal, a is the scale parameter, which is used to control the expansion and contraction of the wavelet function, b is the translation parameter, which is used to control the translation of the wavelet function on the time axis, t represents the position of the numerical simulation signal on the time axis, ψ * (t) is the complex conjugate of ψ(t), and ψ(t) is the mother wavelet function.

[0094] Furthermore, the device is also used for:

[0095] Setting the equal spatial weather conditions based on a plurality of real-time weather elements;

[0096] According to the same spatial weather conditions, multi-parameter configuration is performed through the numerical simulation model, wherein the multi-parameters include network resolution parameters, time step parameters, and physical simulation parameters;

[0097] Outputting the first precision data set by performing multi-level configuration on the network resolution parameter, the time step parameter, and the physical simulation parameter;

[0098] The second precision data set is output by performing single-stage configuration on the network resolution parameters, the time step parameters, and the physical simulation parameters.

[0099] Furthermore, the device is also used for:

[0100] Using the second precision data set as input data of the super-resolution model, and using the first precision data set as target output data;

[0101] Defining a loss function for the super-resolution model, and obtaining a loss difference through the loss function;

[0102] Based on the input data, the super-resolution model is iteratively trained using a stochastic gradient descent optimizer until the loss difference is less than a preset threshold, and the super-resolution model is output. In summary, any step of the method described above can be stored as a computer instruction or program in an unlimited computer memory, and can be called and recognized by an unlimited computer processor to implement any method in the embodiments of the present application, and no unnecessary restrictions are made here.

[0103] Furthermore, the first or second mentioned above may not only represent an order relationship, but may also represent a specific concept, and / or refer to multiple elements that can be selected individually or in full. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A method for numerical simulation of space weather, characterized in that: The method comprises: Real-time monitoring is carried out through space weather detection instruments to obtain space weather numerical simulation data; Constructing a numerical simulation model based on the space weather numerical simulation data; According to the same spatial weather conditions, the numerical simulation model is used to output a first precision data set and a second precision data set, wherein the first precision data set and the second precision data set represent simulation results of different precisions, wherein the first precision data set represents a high-precision simulation result, and the second precision data set represents a low-precision simulation result; Performing data training based on the first precision data set and the second precision data set to build a super-resolution model; Using the space weather numerical simulation data, obtaining target high-precision simulation data through the super-resolution model; Wherein, outputting a first precision data set and a second precision data set using the numerical simulation model according to the same space weather conditions includes: Setting the equal spatial weather conditions based on a plurality of real-time weather elements; According to the same spatial weather conditions, multi-parameter configuration is performed through the numerical simulation model, wherein the multi-parameters include network resolution parameters, time step parameters, and physical simulation parameters; Outputting the first precision data set by performing multi-level configuration on the network resolution parameter, the time step parameter, and the physical simulation parameter, wherein the multi-level configuration refers to performing different levels of settings according to different precision requirements or computing resource limitations; The second precision data set is output by performing single-level configuration on the network resolution parameter, the time step parameter, and the physical simulation parameter, wherein the single-level configuration refers to a low-precision, simple setting.

2. The method according to claim 1, characterized in that Real-time monitoring is performed through space weather detection instruments to obtain space weather numerical simulation data, including: Starting the space weather detection instrument to monitor the target space area in real time to obtain multiple real-time weather elements; Convert the data format of the multiple real-time weather elements, filter them according to the conversion results, and generate target real-time weather elements; The target real-time weather elements are transmitted to a space weather data center for simulation calculation to generate the space weather numerical simulation data.

3. The method according to claim 1, characterized in that Constructing a numerical simulation model based on the space weather numerical simulation data, the method comprising: Determine a first hyperparameter of the numerical simulation model based on the model framework, update the first hyperparameter through a back propagation algorithm, and generate a second hyperparameter; Performing data preprocessing on the space weather numerical simulation data, dividing the data according to the preprocessing result combined with the second hyperparameter, and generating a training data set, a supervision data set, and a test data set; Using the training data set to train the numerical simulation model, and monitoring the performance indicator data of the training data set and the supervision data set in the numerical simulation model; Using the test data set to perform performance evaluation on the numerical simulation model to generate a performance evaluation result; The numerical simulation model is iteratively optimized based on the performance indicator data in combination with the performance evaluation result until the numerical simulation model reaches the performance indicator data for output.

4. The method according to claim 3, characterized in that The space weather numerical simulation data is preprocessed, the method comprising: Performing wavelet decomposition on the numerical simulation signal of the space weather numerical simulation data using a wavelet transform formula to obtain wavelet coefficients of the numerical simulation signal; Perform threshold quantization according to the wavelet coefficients of the numerical simulation signal to determine the wavelet selection threshold of the numerical simulation signal; According to the numerical simulation signal wavelet selection threshold, the wavelet coefficients of the numerical simulation signal are intercepted, the noise signal less than the numerical simulation signal wavelet selection threshold is set to zero, and the effective signal information greater than the numerical simulation signal wavelet selection threshold is obtained; The effective signal information is filtered and reconstructed to obtain the preprocessing result.

5. The method according to claim 4, characterized in that The wavelet transform formula is: Among them, W t (a, b) are the wavelet coefficients of the numerical simulation signal, which are used to characterize the characteristics of the numerical simulation signal f(t) at scale a and time position b. f(t) is the numerical simulation signal, a is the scale parameter, which is used to control the expansion and contraction of the wavelet function, b is the translation parameter, which is used to control the translation of the wavelet function on the time axis, t represents the position of the numerical simulation signal on the time axis, ψ * (t) is the complex conjugate of ψ(t), and ψ(t) is the mother wavelet function.

6. The method according to claim 1, characterized in that Performing data training based on the first precision data set and the second precision data set to build a super-resolution model, the method comprising: Using the second precision data set as input data of the super-resolution model, and using the first precision data set as target output data; Defining a loss function for the super-resolution model, and obtaining a loss difference through the loss function; Based on the input data, the super-resolution model is iteratively trained using a stochastic gradient descent optimizer until the loss difference is less than a preset threshold, and the super-resolution model is output.

7. A space weather numerical simulation device, characterized in that: For executing the method according to any one of claims 1 to 6, the device comprises: Acquisition simulation data module: used to conduct real-time monitoring through space weather detection instruments and obtain space weather numerical simulation data; A simulation model building module is used to build a numerical simulation model based on the space weather numerical simulation data; A data set output module is used to output a first precision data set and a second precision data set using the numerical simulation model according to the same spatial weather conditions, wherein the first precision data set and the second precision data set represent simulation results of different precisions, wherein the first precision data set represents a high-precision simulation result, and the second precision data set represents a low-precision simulation result; A super-resolution model building module is used to perform data training based on the first precision data set and the second precision data set to build a super-resolution model; A module for obtaining high-precision simulation data is used to adopt the space weather numerical simulation data and obtain target high-precision simulation data through the super-resolution model.