Method for predicting long-term solar activity level

Abstract

The invention discloses a method for predicting a long-term solar activity level. The method comprises the following steps: S1, establishing a solar activity level prediction model; s2, establishing a sunspot relative several-month mean value database; s3, obtaining a target prediction month; s4, using an RBF neural network to predict a sunspot relative several-year monthly average value of n months after the target month, and filling the sunspot relative several-month average value database with the sunspot relative several-month average value; s5, traversing the sunspot relative several-month mean value database, and obtaining the sunspot relative several-month mean value data of the target month and the previous several months; s6, substituting the mean value data of the relative multiple months of the sunspots of the target month and the previous multiple months into the solar activity level prediction model; and S7, outputting a monthly average value of the sunspots relative to several years and a prediction result thereof. According to the method, the average value of the relative several years of the sunspots in the current month and several months after the current month can be accurately predicted, and the average value can be used as a reference basis of the solar activity level and can also be used as an input parameter for forecasting other related parameters.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of solar activity level of astrophysics, and particularly to a method for predicting long-term solar activity level. BACKGROUND

[0002] Currently, the parameters used internationally to describe long-term solar activity level are mainly the smoothed monthly mean relative sunspot number, which is commonly used by those skilled in the art as an input of a prediction model related to long-term solar activity level.

[0003] The smoothed monthly mean relative sunspot number is calculated by using the relative sunspot number of 13 consecutive months, and the smoothed monthly mean relative sunspot number (hereinafter referred to as SSN) is calculated by the following formula:

[0004]

[0005] The index actually calculates the value in the median month in each 13 months, for example, the calculation result of SSN is the value of the 7th month by substituting the data of a certain 13 months. The disadvantage of this calculation method is that the latest value is obtained with a time delay of several months. No matter how many consecutive months of data are used, there is more or less a delay. The second disadvantage of SSN calculation is that if the SSN of the current month is needed for real-time acquisition and used for subsequent research, the value of the current month actually uses the measured value of the next 6 months. If the prediction is used as an input factor, it is equivalent to bringing in the future measured information. That is, the prediction of the next 6 months using this value contains some information of the prediction month itself or even after the prediction month. It contains the factor of predicting itself, and even can be said to predict the past with the future, which is obviously unreasonable in logic. It cannot reflect real-time information, and it is impossible to use SSN for further research in academia.

[0006] The information disclosed in this BACKGROUND section is only for the purpose of increasing the understanding of the background of the present application and should not be taken as an acknowledgement or any form of suggestion that this information forms prior art that is publicly known. SUMMARY

[0007] The present application aims to provide a method for predicting long-term solar activity level, which can solve the problems in the background art.

[0008] To achieve the above object, the application provides a method for predicting long-term solar activity level, characterized in that the method comprises the following steps: S1, establishing a solar activity level prediction model; S2, establishing a database of relative sunspot number monthly average; S3, obtaining a target prediction month; S4, using an RBF neural network to predict the relative sunspot number yearly moving monthly average of the target month and the n months after the target month, and filling the relative sunspot number monthly average database; S5, traversing the relative sunspot number monthly average database to obtain the relative sunspot number monthly average data of the target month and the previous months; S6, substituting the relative sunspot number monthly average data of the target month and the previous months into the solar activity level prediction model; and S7, outputting the relative sunspot number yearly moving monthly average and the prediction result thereof.

[0009] In one or more embodiments, the step S1 comprises the following steps: S11, inputting a calculation formula of the relative sunspot number yearly moving monthly average; S12, determining input and output; and S13, generating input and output ports according to the number of input and output.

[0010] In one or more embodiments, in the step S11, the calculation formula of the relative sunspot number yearly moving monthly average is: SSN yi =∑(SN i-11 +SN i-10 +......+SN i-1 +SN i ) / 12, wherein i month refers to the target prediction month, SSN yi refers to the relative sunspot number yearly moving monthly average, SN i refers to the relative sunspot number monthly average of the i month, and SN i-n refers to the relative sunspot number monthly average of the n months before the i month.

[0011] In one or more embodiments, the step S2 comprises the following steps: S21, obtaining relative sunspot number monthly average data; and S22, storing the relative sunspot number monthly average data in corresponding storage spaces according to month order.

[0012] In one or more embodiments, the step S4 comprises the following steps: S41, setting a prediction algorithm of the RBF network and formulating a loss optimization strategy; S42, predicting the relative sunspot number yearly moving monthly average and filling the prediction value into the corresponding month of the relative sunspot number yearly moving monthly average database; and S43, after the prediction process, using the loss optimization strategy to reduce the error of the prediction value and adjusting the prediction algorithm.

[0013] In one or more embodiments, the step S42 comprises: S421, loading the relative sunspot number annual moving average of the required month, the format containing a time stamp and a numerical field; S422, setting the lag step to build a supervised learning format; S423, dividing the training set / test set; S424, initializing the input layer, hidden layer and output layer network parameters; S425, executing the core code of the prediction algorithm; S426, obtaining the prediction value and filling in the relative sunspot number annual moving average database.

[0014] In one or more embodiments, the step S43 comprises: S431, resetting the target prediction month, wherein the reset target prediction month is the current month and the months before the current month; S432, traversing the relative sunspot number monthly average database to read all the relative sunspot number monthly averages and corresponding months used to calculate the relative sunspot number annual moving average; S433, hiding the relative sunspot number annual moving average of the second half of the months to simulate the prediction of the relative sunspot number annual moving average of the second half of the months to obtain the prediction value; S434, comparing the prediction value with the relative sunspot number annual moving average of the second half of the months to obtain the loss gap, and optimizing the prediction algorithm through back propagation.

[0015] The second aspect of the present application provides a system for predicting long-term solar activity level, based on the same concept of the method for predicting long-term solar activity level, comprising: a first establishment module for establishing a solar activity level prediction model; a second establishment module for establishing a relative sunspot number monthly average database; a first acquisition module for acquiring a target prediction month; a first prediction module using a RBF neural network for predicting the relative sunspot number annual moving average of the n months after the target month and filling in the relative sunspot number monthly average database; a first acquisition module for acquiring the relative sunspot number monthly average data of the target month and the previous months; a first input module for inputting the relative sunspot number monthly average data of the target month and the previous months into the solar activity level prediction model; and a first output module for outputting the relative sunspot number annual moving average and its prediction result.

[0016] The third aspect of the present application provides an electronic device, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements a method for predicting long-term solar activity level when executing the computer program.

[0017] The fourth aspect of the present application provides a non-transitory computer readable storage medium having a computer program stored thereon, characterized in that the computer program is executed by a processor to implement a method for predicting long-term solar activity level.

[0018] Compared with the prior art, the multiple technical solutions and embodiments provided by the present application at least have the following technical effects or advantages:

[0019] The prediction algorithm based on the RBF network is set to accurately predict the relative monthly mean value of sunspot number in the current month and the future; a loss optimization strategy is formulated to train the prediction algorithm so that the error precision can be comparable to the acceptable error range of the actual meteorological observation of the relative monthly mean value of sunspot number, so that the actually predicted relative monthly mean value of sunspot number in the current month can reach the super-high precision recognized by the industry and be used as a reference basis for the level of solar activity. BRIEF DESCRIPTION OF DRAWINGS

[0020] The accompanying drawings, which are included to provide a further understanding of the present application and constitute a part of this application, illustrate the illustrative embodiments of the present application and their description serves to explain the present application, which can be considered as a combination of various combinations of preferred embodiments, and do not constitute an improper limitation on the present application. In the drawings:

[0021] Figure 1 The overall flowchart of the method for predicting long-term solar activity level provided by the present application is shown in the figure;

[0022] Figure 2 The method for predicting long-term solar activity level provided by the present application is used to predict the SSN index of the first month in the future from January 1950 to February 2025 y Prediction, actual measurement and error of the SSN index of the first month in the future;

[0023] Figure 3 The method for predicting long-term solar activity level provided by the present application is used to predict the SSN index of the second month in the future from January 1950 to February 2025 y Prediction, actual measurement and error of the SSN index of the second month in the future;

[0024] Among them, the purple curve represents the actual measurement value, the red curve represents the predicted value, the dotted line represents the numerical range of the actual measurement value floating by 10%, and the red curve being hidden by the purple curve represents that the predicted value is highly consistent with the actual value; the blue curve represents the absolute error and the relative error from top to bottom, respectively. DETAILED DESCRIPTION

[0025] Unless otherwise explicitly indicated, in the entire specification and claims, the term "comprise" or its variants such as "contain" or "include" or "made of" and the like will be understood to include the stated element or component, without excluding the presence of other elements or components.

[0026] The application aims to provide a method for predicting long-term solar activity level, wherein the existing smoothed monthly mean of relative sunspot number SSN is preferably calculated by selecting 13-month smoothed monthly mean of relative sunspot number as the reference value of solar activity level of the median month, for example, the 13-month smoothed monthly mean of relative sunspot number from January 2024 to January 2025 is selected to calculate the smoothed monthly mean of relative sunspot number SSN in July 2024 (the median month), and the reason for using 13-month data is that it is the most suitable number of months that meets the sample size requirement verified by the field and recognized. Since the future monthly mean of relative sunspot number cannot be predicted, the existing method for calculating the smoothed monthly mean of relative sunspot number SSN has a delay of 6 months. The application can eliminate the delay of the monthly mean of relative sunspot number prediction by calculating the annual smoothed monthly mean of relative sunspot number, and provide a method for accurately predicting the solar activity level and logically rigorous.

[0027] Embodiment one:

[0028] The embodiment provides a method for predicting long-term solar activity level, comprising:

[0029] S1: establishing a solar activity level prediction model; specifically, the solar activity level prediction model comprises a prediction formula and a prediction step, and accurate solar activity level prediction data can be obtained after inputting the original monthly mean of relative sunspot number into the solar activity level prediction model.

[0030] As a preferred embodiment of the embodiment, the step S1 comprises:

[0031] S11: inputting a calculation formula of the annual smoothed monthly mean of relative sunspot number;

[0032] S12: determining the input and output;

[0033] S13: generating an input port and an output port according to the number of the input and output.

[0034] Specifically, in order to provide a solar activity level prediction model, the prediction formula and the prediction step are set, the calculation formula of the annual smoothed monthly mean of relative sunspot number is obtained, and the input and output interfaces are determined. Preferably, the 12-month monthly mean of relative sunspot number is used as the input, and the annual smoothed monthly mean of relative sunspot number is used as the output.

[0035] As a preferred embodiment of the embodiment, in the step S11, the calculation formula of the annual smoothed monthly mean of relative sunspot number is:

[0036] SSN yi =∑(SN i-11 +SNi-10 +......+ SN i-1 + SN i ) / 12, wherein, i month refers to the target prediction month, SSN yi refers to the relative sunspot number monthly average, SN i is the relative sunspot number monthly average of the ith month, SN i-n refers to the relative sunspot number monthly average of the n months before the ith month.

[0037] As a preferred embodiment, the calculation formula of the relative sunspot number annual moving monthly average weights and averages the relative sunspot number monthly averages of different months, finds the correlation with the predicted parameters of the model, and thus improves the prediction accuracy of the model.

[0038] S2: Establish a relative sunspot number monthly average database;

[0039] As a preferred embodiment of the present embodiment, the step S2 comprises:

[0040] S21: Obtain relative sunspot number monthly average data;

[0041] S22: Store the relative sunspot number monthly average data in the corresponding storage space in the order of months.

[0042] Specifically, in order to be able to store and obtain solar activity level data at any time, the present embodiment establishes a relative sunspot number monthly average database, and according to the order of months, the storage space is divided, and when the data needs to be retrieved, the corresponding information can be retrieved and obtained according to the month.

[0043] S3: Obtain the target prediction month;

[0044] S4: Use the RBF neural network to predict the relative sunspot number annual moving monthly average of the n months after the target month, and fill in the relative sunspot number annual moving monthly average database;

[0045] Specifically, the existing solar activity level indicators cannot be obtained in real time, and the real-time acquisition cannot meet the timeliness required by the business prediction of the related parameters of the solar activity level, which leads to the delay of the progress of the related parameter business prediction, and the present embodiment can obtain the solar activity level through formula calculation after the end of a natural month, and improve the efficiency of the progress of the related business prediction.

[0046] As a preferred embodiment of the present embodiment, the step S4 comprises:

[0047] S41: Set the prediction algorithm of the RBF network, and develop a loss optimization strategy, such as adjusting the training step length and other parameters;

[0048] S42: predicting the relative sunspot number monthly average of the year, and filling the prediction value into the corresponding month of the relative sunspot number monthly average of the year database;

[0049] S43: after the prediction process, using a loss optimization strategy to reduce the error of the prediction value, and adjusting the prediction algorithm.

[0050] Specifically, in order to respond to the solar activity level in real time, the biggest technical problem of obtaining the solar activity level is that there is no future sunspot relative number monthly average data support. The embodiment proposes a method for responding to the solar activity level by predicting the subsequent sunspot relative number monthly average of the year. The prediction method includes a prediction algorithm and a loss optimization strategy. The prediction algorithm is used to obtain the known sunspot relative number monthly average in the sunspot relative number monthly average database, to obtain the sunspot relative number monthly average of the year, and to predict the future sunspot relative number monthly average of the year by month. The prediction algorithm has the disadvantage that the prediction accuracy is uncertain, so the loss optimization strategy is set to train the prediction algorithm and make the prediction result more accurate.

[0051] As a preferred embodiment of the present embodiment, the step S42 comprises:

[0052] S421: loading the sunspot relative number monthly average of the year of the required month, the format including a time stamp and a numerical field;

[0053] S422: setting a delay step (lag) to build a supervised learning format;

[0054] S423: dividing the training set / test set;

[0055] S424: initializing the input layer, hidden layer and output layer network parameters;

[0056] S425: executing the core code of the prediction algorithm;

[0057] S426: obtaining the prediction value and filling it into the sunspot relative number monthly average of the year database.

[0058] Specifically, the radial basis function RBF network is a kind of forward network constructed on the basis of function approximation theory. The learning of this kind of network is equivalent to finding the best fitting plane of the training data in a multi-dimensional space. This network avoids the tedious and lengthy calculation of BP network, has high operation speed and extrapolation ability, and has strong non-linear mapping function.

[0059] The RBF network is a nonlinear conversion from the input space RN to the output space RM by a linear combination of nonlinear basis functions. The sunspot number is a strong nonlinear time series, and the prediction of the sunspot number, i.e., the prediction of the future M data from the previous N data, is essentially to find the nonlinear mapping relationship from RN to RM. Therefore, it can be said that the radial basis function network is particularly suitable for the prediction of nonlinear time series such as the sunspot number.

[0060] As a preferred embodiment of the present embodiment, the step S43 comprises:

[0061] S431: resetting the target prediction month, wherein the reset target prediction month is a number of months before the current month and earlier months;

[0062] S432: traversing the sunspot relative number monthly mean database to read all sunspot relative number monthly means and corresponding months for calculating the sunspot relative number annual sliding monthly mean;

[0063] S433: hiding the sunspot relative number annual sliding monthly mean of the second half of the months, simulating the sunspot relative number annual sliding monthly mean of the second half of the months to obtain a prediction value;

[0064] S434: comparing the prediction value with the sunspot relative number annual sliding monthly mean of the second half of the months to obtain a loss gap, and optimizing the prediction algorithm by back propagation.

[0065] Specifically, the radial basis function (RBF) network method is used to predict SSNy, the loss optimization strategy is used to correct the prediction algorithm, and the problem encountered by the loss optimization strategy is how to find the loss optimization corpus. Since the prediction value still needs to be delayed for n months to obtain the actual data for comparison, the optimization efficiency is poor. Therefore, to solve this problem, the present embodiment uses the data of historical months for training. First, the target month is reset, and the target is set from the current month to the month for which the sunspot relative number annual sliding monthly mean data of a number of months before and after can be determined.

[0066] In order to avoid predicting the current situation with the data of the current month, the system makes the algorithm substitute the data of the previous n months to predict the data of the next n months, that is, assuming that the data of the next n months is not obtained, the predicted value is compared with the actual value by predicting the data of the next n months, and then the algorithm is optimized to make the predicted value close to the actual value. Since the relative sunspot number data has a long history, the relative sunspot number monthly average database contains rich training corpus, and through a large number of operations, the prediction algorithm can accurately predict the relative sunspot number annual sliding monthly average. The loss optimization strategy operation of the embodiment aims to reduce the difference between the predicted value and the actual value to the allowable error range of the international measured relative sunspot number annual sliding monthly average, that is, the measured relative sunspot number annual sliding monthly average also has an error, and the predicted value of the embodiment can be optimized to the error range allowed interval, so the predicted value can actually be recognized by the industry as being equivalent to the measured value. The relative sunspot number annual sliding monthly average calculated according to this can also be recognized as an accurate value in the industry.

[0067] For example, the data from January 1750 to December 1949 is selected to train the radial basis function (RBF) network, the observation values D(n-2), D(n-1) and D(n) of the previous three months are used as the input vectors of the network, and the observation values D(n+1) and D(n+2) of the first and second months after the previous three months are used as the output vectors to train the network. The data from January 1950 to February 2025 is selected for prediction test, the observation values D(n-2), D(n-1) and D(n) of the previous three months are used as the input vectors of the network, and the output vectors are the predicted values D(n+1) and D(n+2) of the first and second months in the future. Compare with the actual value, optimize the prediction algorithm to make the error of the predicted value within the allowable error range of the actual relative sunspot number annual sliding monthly average.

[0068] S5: traversing the relative sunspot number annual sliding monthly average database to obtain the relative sunspot number monthly average data of the target month and the previous months;

[0069] S6: substituting the relative sunspot number annual sliding monthly average data of the target month and the previous months into the solar activity level prediction model;

[0070] S7: outputting the relative sunspot number annual sliding monthly average and the prediction result.

[0071] Embodiment two:

[0072] The embodiment provides a system for predicting long-term solar activity level, based on the same concept of the method for predicting long-term solar activity level, comprising:

[0073] The first establishing module is configured to establish a solar activity level prediction model;

[0074] The second establishing module is configured to establish a database of relative sunspot number monthly average values;

[0075] The first obtaining module is configured to obtain a target prediction month;

[0076] The first prediction module is configured to use an RBF neural network to predict relative sunspot number yearly sliding monthly average values of n months after the target month and fill in the database of relative sunspot number monthly average values;

[0077] The first obtaining module is configured to obtain relative sunspot number monthly average value data of the target month and previous months;

[0078] The first inputting module is configured to input the relative sunspot number monthly average value data of the target month and previous months into the solar activity level prediction model;

[0079] The first output module is configured to output the relative sunspot number yearly sliding monthly average values and prediction results thereof.

[0080] Embodiment Three

[0081] The embodiment provides an electronic device based on the same concept, and is used for implementing the method for predicting long-term solar activity level in the embodiment one. The electronic device comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor. The processor implements the method for predicting long-term solar activity level when executing the computer program.

[0082] Embodiment Four

[0083] The embodiment provides a non-transitory computer readable storage medium, which stores a computer program. The computer program is executed by a processor to implement the method for predicting long-term solar activity level.

[0084] The foregoing description of specific example embodiments of the application is provided for the purpose of explanation and illustration. These descriptions are not intended to limit the application to the precise forms disclosed, and many modifications and variations are possible in light of the above teachings. The example embodiments were chosen and described in order to explain the principles of the application and its practical application, and to enable others skilled in the art to implement and utilize the application in various different example embodiments and various different combinations and modifications. The scope of the application is intended to be defined by the claims and their equivalents.

Claims

1. A method for predicting long-term solar activity levels, characterized in that: include: S1: Establish a solar activity level prediction model; S2: Establish a database of monthly mean values ​​of the relative number of sunspots; S3: Get the target forecast month; S4: using the RBF neural network to predict the monthly average of the relative number of sunspots in the n months after the target month, and filling in the monthly average database of the relative number of sunspots; S5: traversing the monthly mean value database of the relative number of sunspots to obtain the monthly mean value data of the relative number of sunspots for the target month and several previous months; S6: Substituting the monthly average data of sunspots in the target month and the previous months into the solar activity level prediction model; S7: Output the monthly mean of the relative number of sunspots and its prediction results.

2. A method for predicting long-term solar activity levels according to claim 1, characterized in that: The step S1 comprises: S11: Input the calculation formula of the sunspot relative annual and monthly mean; S12: Determine input and output quantities; S13: Generate input ports and output ports according to the number of input quantities and output quantities.

3. A method for predicting long-term solar activity levels according to claim 2, characterized in that: In step S11, the calculation formula of the sunspot relative annual monthly mean is: SSN yi =∑(SN i-11 +SN i-10 +......+SN i-1 +SN i ) / 12, where month i refers to the target forecast month, SSN yi Refers to the sunspot relative to the monthly average over several years, SN i is the monthly mean of the number of sunspots in month i, SN i-n It refers to the monthly average of sunspots in n months before month i.

4. A method for predicting long-term solar activity levels according to claim 3, characterized in that: The step S2 comprises: S21: Obtain the monthly mean data of the relative number of sunspots; S22: storing the monthly mean data of the relative number of sunspots into corresponding storage spaces in order of months.

5. A method for predicting long-term solar activity levels according to claim 4, characterized in that: The step S4 comprises: S41: Set the prediction algorithm of the RBF network and formulate the loss optimization strategy; S42: Predicting the relative annual and monthly mean of sunspots, and filling the predicted value into the corresponding month of the relative annual and monthly mean database of sunspots; S43: After the prediction process, a loss optimization strategy is adopted to reduce the error of the prediction value and adjust the prediction algorithm.

6. A method for predicting long-term solar activity levels according to claim 5, characterized in that: The step S42 includes: S421: Load the sunspot relative annual sliding monthly mean for the required month, in a format that includes a timestamp and a numeric field; S422: Setting the delay step (lag) to construct a supervised learning format; S423: Divide the training set / test set; S424: Initialize and set the network parameters of the input layer, hidden layer and output layer; S425: Execute the core code of the prediction algorithm; S426: Obtain the predicted value and fill it into the sunspot relative annual and monthly mean database.

7. A method for predicting long-term solar activity levels according to claim 6, characterized in that: The step S43 includes: S431: Resetting the target forecast month, wherein the reset target forecast month is a number of months before the current month or an earlier month; S432: traverse the monthly mean value database of the relative number of sunspots, and read all the monthly mean values ​​of the relative number of sunspots and the corresponding months for calculating the annual and monthly mean value of the relative number of sunspots; S433: Hiding the sunspot values ​​for the second half of the month relative to the monthly mean over several years, and simulating and predicting the sunspot values ​​for the second half of the month relative to the monthly mean over several years to obtain a predicted value; S434: Compare the predicted value with the sunspot average over several years in the second half of the month to obtain the loss gap, and optimize the prediction algorithm through back propagation.

8. A system for predicting long-term solar activity levels, based on the same concept as the method for predicting long-term solar activity levels as claimed in any one of claims 1 to 7, characterized in that: include: The first building module is used to build a solar activity level prediction model; The second establishment module is used to establish a database of monthly mean values ​​of the relative number of sunspots; The first acquisition module is used to obtain the target forecast month; The first prediction module uses an RBF neural network to predict the monthly mean of the relative number of sunspots n months after the target month and fills the monthly mean database of the relative number of sunspots; The first acquisition module is used to obtain the monthly average data of sunspots in the target month and the previous months; A first input module is used to substitute the monthly average data of sunspots in the target month and the previous months into the solar activity level prediction model; The first output module is used to output the sunspot relative annual monthly mean and its prediction results.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for predicting the long-term solar activity level according to any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting the long-term solar activity level according to any one of claims 1 to 7 is implemented.

Images