Method and system for predicting ocean surface temperature change and medium
By preprocessing the ocean surface temperature data and screening and fitting the Fourier series, the optimal order is determined to predict the interdecadal changes of ocean surface temperature, and the Fourier series order selection in the prior art is solved and the issue of predictive credibility improves the accuracy and reliability of predictions and provides scientific basis for climate research.
Patent Information
- Application Number
- CN202411065255.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-05
- Publication Date
- 2024-11-26
AI Technical Summary
The prior art is difficult to scientifically and objectively determine the order of Fourier series and evaluate the credibility of prediction results, resulting in insufficient accuracy of prediction of ocean surface temperature changes.
By preprocessing historical ocean surface temperature data, extracting interdecadal components, traversing the preset order sets to filter candidate orders of Fourier series, calculating cross-validation coefficients, determining the optimal order, and using the optimal order Fitting was performed to obtain periodic changes in ocean surface temperature to predict interdate changes.
It effectively improves the prediction accuracy of ocean surface temperature changes, provides scientific basis for climate change research and related policy formulation, and ensures the reliability of order selection of Fourier series and the credibility of prediction.
Smart Images

Figure CN119025992A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of ocean surface temperature change prediction, and in particular to a method, system and medium for predicting ocean surface temperature change. Background Art
[0002] In recent years, understanding and predicting changes in sea surface temperature have been very important for climate research, fishery management, navigation safety and other fields. Taking the North Atlantic as an example, its sea surface temperature change is a complex process, which is affected by many factors, including the Arctic amplification effect, the Atlantic meridional overturning circulation and global warming. These factors interact with each other, making the changes in the North Atlantic SST show significant multi-period characteristics.
[0003] At present, Fourier series analysis, as a powerful mathematical tool, is widely used in the analysis and prediction of time series. It can effectively describe the periodic changes in time series and predict future trends through the fitting process. However, how to scientifically and objectively determine the order of the Fourier series, evaluate the credibility of the prediction results, and improve the prediction accuracy are problems that need to be solved urgently. Summary of the invention
[0004] In order to improve prediction accuracy, the present application provides a method, system and medium for predicting changes in ocean surface temperature.
[0005] In a first aspect, the present application provides a method for predicting changes in ocean surface temperature, using the following technical solution: A method for predicting changes in ocean surface temperature, the method comprising: Preprocess the historical sea surface temperature data within the target time to obtain the decadal component corresponding to the sea surface temperature time series; Traversing the preset order set according to the decadal component, screening and obtaining a candidate order set of Fourier series; wherein the preset order set includes a plurality of preset orders of Fourier series; Calculating, according to the decadal component, a cross-validation coefficient of Fourier series fitting at each candidate order in the candidate order set; Based on a preset correlation coefficient test table, determining the credibility corresponding to each candidate order according to the cross-validation coefficient; Determine the candidate order with the highest credibility in the candidate order set as the optimal order; Fitting the interdecadal component by the Fourier series under the optimal order to obtain the periodic variation characteristics of the interdecadal component of the ocean surface temperature; The decadal changes in the ocean surface temperature are predicted based on the periodic variation characteristics.
[0006] By adopting the above technical solution, the historical sea surface temperature data is preprocessed to remove the interference of noise in the time series and extract the interdecadal component; then the preset order set is traversed to screen out the candidate order set of the Fourier series; then, the credibility is determined by calculating the cross-validation coefficient of the Fourier series fitting under each candidate order, and the order with the highest credibility is selected as the optimal order; finally, the interdecadal component is fitted using the Fourier series under the optimal order to obtain the periodic variation characteristics, and the interdecadal variation of the sea surface temperature is predicted accordingly. This series of steps scientifically and objectively determines the order of the Fourier series, and at the same time proposes a test method for the prediction reliability of the order, which can effectively predict the interdecadal variation trend of the sea surface temperature, thereby providing a scientific basis for climate change research and related policy formulation, and has important scientific research value and application prospects.
[0007] Optionally, the steps of preprocessing the historical sea surface temperature data within the target time to obtain the decadal component corresponding to the sea surface temperature time series include: Collect sea surface temperature time series based on historical sea surface temperature data within the target time; The sea surface temperature time series is low-pass filtered by sliding average to obtain an interdecadal component; wherein the time window of the sliding average is a preset value.
[0008] By adopting the above technical solutions, the impact of noise can be effectively reduced, long-term trends and periodic changes can be highlighted, and a solid foundation can be laid for subsequent Fourier series fitting and prediction.
[0009] Optionally, low-pass filtering the sea surface temperature time series by sliding average includes: Calculate the moving average Y based on the sea surface temperature time series n , the calculation formula is: In the above formula, X (n-k) is the sea surface temperature time series, k and M are the preset sliding average parameters.
[0010] Optionally, the sliding average time window is 9 years, and the 9-year sliding average Y n The expression is: In the above formula, X (n-k) is the sea surface temperature time series, and k is the preset sliding average parameter.
[0011] Optionally, the step of traversing a preset order set according to the decadal component to screen and obtain a candidate order set of Fourier series includes: According to the preset orders in the preset order set, respectively performing Fourier series fitting of different orders on the decadal components to obtain a Fourier series fitting curve of each preset order; Screening and eliminating the preset orders with abnormalities in the Fourier series fitting curve; Calculate the correction fit corresponding to each preset order in the preset order set after elimination; The preset orders whose correction fitness is greater than a preset threshold value are combined into the candidate order set.
[0012] By adopting the above technical solution, traversing the preset order set, fitting the decadal component with Fourier series of different orders, and through steps such as screening and correction of fit calculation, the candidate order set with good performance is effectively screened out to ensure that the selected model has high predictive and generalization capabilities, providing strong support for subsequent prediction work, thereby better serving climate change research and related decision-making.
[0013] Optionally, the calculation formula for the corrected goodness of fit is: In the above formula, p is the number of variables, n is the number of samples, and R 2 is the degree of fit.
[0014] Optionally, the step of calculating the cross-validation coefficient of the Fourier series fitting at each candidate order in the candidate order set according to the decadal component includes: Substituting each candidate order in the candidate order set into the Fourier series formula, fitting the observed values of the 1st to nth items in the sea surface temperature time series, substituting the coefficients obtained by fitting into the fitting equation, predicting the predicted values of the n+1th to n+10th items in the sea surface temperature time series, and calculating the correlation coefficient between the observed values of the n+1th to n+10th items in the sea surface temperature time series and the predicted values; Gradually increase the value of n and repeat the above steps to obtain multiple sets of correlation coefficients between the observed values and the predicted values from the n+1th to the n+10th items, forming a set of correlation coefficients corresponding to each candidate order; The arithmetic mean calculation is performed on the correlation coefficients in the correlation coefficient set to obtain the cross-validation coefficient of the Fourier series fitting under each candidate order.
[0015] By adopting the above technical solution, Fourier series fitting is performed on the sea surface temperature time series data, and the rolling window method and correlation coefficient calculation are used to cross-validate the model, which can effectively evaluate the prediction performance of models of different orders; by calculating the cross-validation coefficients of models of different orders, the model that can most accurately predict the trend of changes in sea surface temperature can be determined. This technical solution can not only improve the accuracy of the prediction, but also ensure that the selected model has strong generalization ability. It has important application value in fields such as climate change research and long-term weather forecasting.
[0016] Optionally, the calculation formula of the correlation coefficient is: In the above formula, X i is the independent variable of the ith observation, y i is the dependent variable of the ith observation, n represents the number of observations, represents the mean value of the independent variable, represents the mean value of the dependent variable.
[0017] In a second aspect, the present application provides a system for predicting changes in ocean surface temperature, which adopts the following technical solution: A prediction system for ocean surface temperature changes, the prediction system comprising: A data processing module is used to pre-process the historical sea surface temperature data within the target time to obtain the decadal component corresponding to the sea surface temperature time series; A candidate order set generation module, used for traversing a preset order set according to the decadal component to screen and obtain a candidate order set of Fourier series; the preset order set includes a plurality of preset orders of Fourier series; A calculation module, used for calculating the cross-validation coefficient of the Fourier series fitting under each candidate order in the candidate order set according to the decadal component; An optimal order determination module is used to determine the candidate order with the largest cross-validation coefficient in the candidate order set as the optimal order; a fitting module is used to fit the decadal component through the Fourier series under the optimal order to obtain the periodic variation characteristics of the decadal component of the sea surface temperature; The prediction module is used to predict the decadal change of the ocean surface temperature according to the periodic change characteristics.
[0018] In a third aspect, the present application provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium stores a computer program that can be loaded by a processor and execute any one of the methods in the first aspect.
[0019] In summary, the present application includes at least one of the following beneficial technical effects: by scientifically and objectively determining the order of the Fourier series and proposing a test method for the predictive reliability of the order, it can effectively predict the interdecadal change trend of the ocean surface temperature, thereby providing a scientific basis for climate change research and related policy formulation, and has important scientific research value and application prospects. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] Figure 1 This is a first flow chart of a method for predicting changes in ocean surface temperature according to one embodiment of the present application.
[0021] Figure 2 This is a second flow chart of a method for predicting changes in ocean surface temperature according to one embodiment of the present application.
[0022] Figure 3 It is a third flow chart of the method for predicting ocean surface temperature changes according to one embodiment of the present application.
[0023] Figure 4 It is a schematic diagram of the North Atlantic Ocean Surface Temperature Index and Fourier series fitting curves of different orders in the actual research process of this application.
[0024] Figure 5 It is a schematic diagram of the correction fit of the Fourier series at different orders in the actual research process of this application.
[0025] Figure 6 It is a schematic diagram of the cross-validation coefficients of Fourier series fitting at different orders in the actual research process of this application. DETAILED DESCRIPTION
[0026] In order to make the purpose, technical solutions and advantages of this application more clear, the following Figure 1-6 It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0027] The embodiment of the present application discloses a method for predicting changes in ocean surface temperature.
[0028] Reference Figure 1 , a method for predicting changes in sea surface temperature, the prediction method comprising: Step S101, preprocessing the historical sea surface temperature data within the target time to obtain the decadal component corresponding to the sea surface temperature time series; Among them, the preprocessing steps can improve the quality of the data, reduce the deviation in subsequent analysis, and ensure the integrity and accuracy of the data; for example, low-pass filtering techniques such as sliding average can be used to remove short-term fluctuations and extract decadal components, so that the model can focus more on long-term trends and cyclical changes.
[0029] In one embodiment of the present application, the target selected time series may be the Atlantic Multidecadal Oscillation Index, which is obtained from the National Oceanic and Atmospheric Administration (NOAA) of the United States.
[0030] Step S102, traversing the preset order set according to the interdecadal component, screening and obtaining a candidate order set of the Fourier series; wherein the preset order set includes several preset orders of the Fourier series, and these orders are used for different complexity levels of Fourier series fitting; specifically, the orders in the order set are natural numbers, that is, the order m can be 1, 2, 3, 4...n in sequence.
[0031] Specifically, the screening process includes, for example: for each preset order, the decadal component is fitted using the Fourier series, and the fitting effect is evaluated. According to the fitting effect (such as the corrected fit), the orders whose fitting curves will obviously deviate from the reasonable range in the future are eliminated, and the orders with good performance are retained to form a candidate order set.
[0032] Step S103, calculating the cross-validation coefficient of the Fourier series fitting under each candidate order in the candidate order set according to the decadal component; Specifically, for each candidate order, the cross-validation coefficient can be used to evaluate the prediction ability of the model corresponding to each candidate order, and the cross-validation coefficient is calculated based on the correlation between the prediction results and the actual data; Among them, the cross-validation coefficient reflects the predictive ability of the prediction method. It is between -1 and 1. The larger the cross-validation coefficient, the stronger the predictive ability. For example, when the cross-validation coefficient is less than 0, it means that the prediction method is inappropriate; when the cross-validation coefficient is between 0 and 0.2, it means that the predictive ability of the prediction method is insufficient; when the cross-validation coefficient is greater than 0.4, it means that the prediction ability of the prediction method is good.
[0033] Step S104, based on a preset correlation coefficient test table, determine the credibility corresponding to each candidate order according to the cross-validation coefficient; Among them, the preset correlation coefficient test table includes the correlation coefficient critical values corresponding to different significance levels (such as 0.05, 0.01, etc.) under different degrees of freedom. By comparing the cross-validation coefficient at each order with the critical value in the preset correlation coefficient test table, it is possible to evaluate whether the prediction results of the model are statistically significant, and then determine the credibility; it can be understood that if the cross-validation coefficient is greater than the corresponding critical value, it is considered that the prediction results of the model are more credible.
[0034] Step S105, determining the candidate order with the highest credibility in the candidate order set as the optimal order; Among them, by selecting the order with the highest credibility, it can be ensured that the model finally used has strong predictive ability and its prediction results are statistically significant.
[0035] Step S106, fitting the decadal component by Fourier series under the optimal order to obtain the periodic variation characteristics of the decadal component of the sea surface temperature; Among them, accurate identification of periodic variation characteristics helps to improve prediction accuracy.
[0036] Specifically, the decadal component is fitted using the Fourier series model at the optimal order, and the periodic variation characteristics of the ocean surface temperature can be accurately identified through fitting.
[0037] Step S107, predicting the decadal variation of the ocean surface temperature based on the periodic variation characteristics.
[0038] Among them, a prediction model is constructed based on the Fourier series model and periodic change characteristics to predict the interdecadal change trend of ocean surface temperature in the future, providing strong data support for climate research and policy making.
[0039] In the above implementation, the historical sea surface temperature data is preprocessed to remove the interference of noise in the time series and extract the interdecadal component; then the preset order set is traversed to screen out the candidate order set of the Fourier series; then, the cross-validation coefficient of the Fourier series fitting under each candidate order is calculated to determine the credibility, and the order with the highest credibility is selected as the optimal order; finally, the interdecadal component is fitted using the Fourier series under the optimal order to obtain the periodic variation characteristics, and the interdecadal variation of the sea surface temperature is predicted accordingly. This series of steps scientifically and objectively determines the order of the Fourier series, and at the same time proposes a test method for the prediction reliability of the order, which can effectively predict the interdecadal variation trend of the sea surface temperature, thereby providing a scientific basis for climate change research and related policy formulation, and has important scientific research value and application prospects.
[0040] This application makes full use of the multi-time scale periodic characteristics of the Fourier series, closely grasps the multi-time scale periodic variation characteristics of the ocean surface temperature time series, and avoids the human factors in the selection of the Fourier series order. By quantifying the prediction ability of each order, an effective evaluation of the prediction reliability is achieved.
[0041] Reference Figure 2 As an implementation of step S101, the step of preprocessing the historical sea surface temperature data within the target time to obtain the decadal component corresponding to the sea surface temperature time series includes: Step S201, collecting sea surface temperature time series according to historical sea surface temperature data within a target time; Among them, the sea surface temperature records within a specified time period are obtained from reliable data sources. These records are usually temperature values measured regularly (such as daily or monthly). The collected temperature data are arranged in chronological order to form a time series. Each data point represents the sea surface temperature at a certain moment. By constructing a time series, the trend of temperature changes over time can be clearly observed.
[0042] Step S202, low-pass filtering the ocean surface temperature time series by sliding average to obtain an interdecadal component.
[0043] The time window of the sliding average is a preset value. The sliding average is a simple and effective low-pass filtering method that smoothes the time series by calculating the average value of data points within a fixed length (i.e., the time window). The time window moves along the time series and calculates the average value of the data in each window. Specifically, by using sliding average filtering, the regularly changing components of the time series can be effectively extracted and the randomly changing components in the time series can be filtered out, thereby providing a prerequisite for reliable prediction of the time series. In some embodiments, the sliding average method can be replaced by a limiting filter method, a median filter method, and a weighted recursive average filter method.
[0044] In the above implementation, the influence of noise can be effectively reduced, long-term trends and periodic changes can be highlighted, and a solid foundation is laid for subsequent Fourier series fitting and prediction.
[0045] As an implementation of step S202, the time window of the sliding average may be an odd number, and low-pass filtering the sea surface temperature time series by sliding average includes: Calculate the moving average Y from the sea surface temperature time series n , the calculation formula is: In the above formula, X (n-k) is the sea surface temperature time series, k and M are the preset sliding average parameters.
[0046] For example, in one embodiment of the present application, the time window of the sliding average is 9 years, and the 9-year sliding average is used to extract the decadal component of the sea surface temperature time series. n The expression is: In the above formula, X (n-k) is the sea surface temperature time series, and k is the preset sliding average parameter.
[0047] In addition, in other embodiments, the time window of the sliding average can also be an even number, and the corresponding sliding average Y n The calculation formula is: In the above formula, X (n-k) is the sea surface temperature time series, and k is the preset sliding average parameter.
[0048] Reference Figure 3 As an implementation of step S102, the step of traversing the preset order set according to the decadal component and screening the candidate order set of the Fourier series includes: Step S301, performing Fourier series fitting of different orders on the decadal components according to preset orders in the preset order set, to obtain a Fourier series fitting curve of each preset order; Among them, Fourier series is a method of representing periodic functions, which can decompose any periodic function into the sum of a series of sine and cosine functions. In this step, the filtered sea surface temperature time series is input into the software for Fourier series fitting, and Fourier series fitting curves of different orders are drawn. These curves can be used to describe the periodic changes of the interdecadal component.
[0049] Specifically, the expression of the Fourier series is as follows: In the above formula, m is the order of the Fourier series, a n is a constant term, b n and w are coefficients; Step S302, screening and eliminating the preset orders of Fourier series fitting curves with abnormalities; Among them, during the fitting process, some high-order Fourier series may be overfitted, that is, the model is too complex and captures noise instead of the real trend. These curves may perform poorly in predicting future data; In some embodiments, abnormal fitting curves can be identified through visual inspection or quantitative indicators (such as residual analysis), and the orders corresponding to these curves are removed from the candidate set to avoid selecting an overly complex model and reduce the risk of overfitting.
[0050] Step S303, calculating the correction fitting degree corresponding to each preset order in the preset order set after elimination; Among them, the adjusted R-squared is an indicator to measure the quality of model fitting. It takes into account the influence of model complexity. The higher the adjusted R-squared, the better the model fitting effect.
[0051] Specifically, the fit and the corrected fit are dimensionless numbers ranging from 0 to 1. The closer the value is to 1, the better the fit effect. 2 It is equal to the square of the correlation coefficient between the fitting value and the fitting object, and the corrected goodness of fit is the change in goodness of fit brought about by the change in the number of variables. In practical applications, since the independence between different variables cannot be guaranteed, goodness of fit and corrected goodness of fit are usually used together. When the corrected goodness of fit and goodness of fit no longer increase with the increase of variables, it proves that the newly added variables are meaningless; when the corrected goodness of fit and goodness of fit both decrease with the increase of variables, it proves that the newly added variables have a negative effect, and this is overfitting. Therefore, using goodness of fit and corrected goodness of fit to determine the appropriate number of variables can reasonably reduce the amount of calculation.
[0052] Specifically, the calculation formula for the corrected fit is: In the above formula, p is the number of variables, n is the number of samples, and R 2 is the degree of fit.
[0053] Step S304: The preset orders whose correction fit is greater than a preset threshold are grouped into a candidate order set.
[0054] Among them, the preset threshold can be pre-set according to the experience value or the specific application scenario, and the orders with the correction fit greater than the preset threshold are composed of a candidate order set, and the models corresponding to these orders are considered to be candidate models with good performance.
[0055] In the above implementation, the preset order set is traversed, Fourier series fitting of different orders is performed on the interdecadal component, and through steps such as screening and correction of fit calculation, a candidate order set with good performance is effectively screened out to ensure that the selected model has high predictive and generalization capabilities, providing strong support for subsequent prediction work, thereby better serving climate change research and related decision-making.
[0056] As an implementation of step S103, the step of calculating the cross-validation coefficient of the Fourier series fitting under each candidate order in the candidate order set according to the decadal component includes: Substitute each candidate order in the candidate order set into the Fourier series formula, fit the observed values of the 1st to nth items in the sea surface temperature time series, substitute the coefficients obtained by fitting into the fitting equation, predict the predicted values of the n+1th to n+10th items in the sea surface temperature time series, and calculate the correlation coefficient between the observed values and the predicted values of the n+1th to n+10th items in the sea surface temperature time series; Gradually increase the value of n and repeat the above steps to obtain multiple sets of correlation coefficients between the observed values and the predicted values from the n+1th to the n+10th items, forming a set of correlation coefficients corresponding to each candidate order; The arithmetic average is calculated based on the correlation coefficients in the correlation coefficient set to obtain the cross-validation coefficient of the Fourier series fitting under each candidate order.
[0057] In the above steps, each candidate order in the candidate order set is used to fit the first n observations in the known sea surface temperature time series according to the mathematical expression of the Fourier series, and the coefficients obtained by fitting are used to predict the sea surface temperature value of the next time period (n+1th to n+10th) according to the Fourier series formula; by comparing the predicted value with the actual observed value, the correlation coefficient between them is calculated, which is used as a measure of the accuracy of the model prediction.
[0058] It can be understood that by calculating the correlation coefficient, the prediction ability of the model in a specific time period can be evaluated, and the correlation coefficients at different candidate orders can be used to compare the prediction performance of different models and provide a basis for the subsequent selection of the best model.
[0059] In this technical solution, the rolling window method is adopted. As the value of n gradually increases, the updated time series data is used to fit the model each time, and the correlation coefficient between the new predicted value and the actual observed value is calculated, so that the correlation coefficients of multiple different time periods can be obtained to form a correlation coefficient set. Furthermore, all the correlation coefficients in the correlation coefficient set corresponding to each candidate order are averaged to obtain an average correlation coefficient representing the prediction ability of the model of the order. This average correlation coefficient is the cross-validation coefficient of the Fourier series fitting under the order.
[0060] In the above implementation, Fourier series fitting is performed on the sea surface temperature time series data, and the rolling window method and correlation coefficient calculation are used to cross-validate the model, which can effectively evaluate the prediction performance of models of different orders; by calculating the cross-validation coefficients of models of different orders, the model that can most accurately predict the trend of changes in sea surface temperature can be determined. This technical solution can not only improve the accuracy of the prediction, but also ensure that the selected model has a strong generalization ability, and has important application value in fields such as climate change research and long-term weather forecasting.
[0061] As an implementation method of the correlation coefficient, the calculation formula of the correlation coefficient is: In the above formula, X i is the independent variable of the ith observation, y i is the dependent variable of the ith observation, n represents the number of observations, represents the mean value of the independent variable, represents the mean value of the dependent variable.
[0062] It should be noted that in the embodiments of the present application, it is assumed that the time series is continuous and obeys the normal distribution, so the above-mentioned Pearson correlation coefficient is used in the embodiments of the present application to measure the linear relationship between the two variables. If the time series is not continuous (for example, when the data is expressed in the form of levels), even if these level data obey the normal distribution, the Spearman correlation coefficient should be used because it is more suitable for the evaluation of the monotonic relationship between level data and non-continuous data. On the other hand, if the time series does not obey the normal distribution, regardless of whether the data is continuous, the Kendall correlation coefficient should be used because it is a non-parametric correlation coefficient for non-normally distributed data, which can effectively evaluate the monotonic correlation between the two variables.
[0063] Therefore, for continuous and normally distributed data, the Pearson correlation coefficient is used; for graded or non-continuous but normally distributed data, the Spearman correlation coefficient is used; and for non-normally distributed data, whether continuous or not, the Kendall correlation coefficient is used. By ensuring that the selected correlation coefficient accurately reflects the relationship between the variables, the accuracy and reliability of the prediction model can be improved. In the actual research process of this application, the internationally accepted North Atlantic Sea Surface Temperature Index was selected as the time series to perform the following steps: Step a: Take a 9-year moving average of the North Atlantic Sea Surface Temperature Index to obtain the regularly changing component of the North Atlantic Sea Surface Temperature Index; Among them, the 9-year moving average Y n The calculation formula is: In the above formula, X (n-k) is the sea surface temperature time series, and k is the preset sliding average parameter.
[0064] Step b: Select a preset order set of natural numbers from 1 to 8, and plot the fitting results of Fourier series of different orders; Among them, the expression of Fourier series is as follows: In the above formula, m is the order of the Fourier series, a n is a constant term, b n and w are coefficients; Reference Figure 4 As shown, the North Atlantic sea surface temperature index (gray) and the Fourier series fitting curves (black) at different orders m are shown. Figure 4 It can be seen that at different orders, the Fourier series can fit the North Atlantic Sea Surface Temperature Index well, and there is no situation where the fitting results deviate significantly from the reasonable range, so there is no need to eliminate them.
[0065] Step c: The corrected fit of the Fourier series fitting at different orders was further calculated to objectively demonstrate the fitting effect.
[0066] The calculation formula for the correction fit is: In the above formula, p is the number of variables, n is the number of samples, and R 2 is the degree of fit.
[0067] Reference Figure 5 As shown in, it is the correction fit of Fourier series under different orders m. Figure 5 It can be seen that when the order m≤4, the correction fitting degree of the Fourier series is less than 90%; when the order m>4, the correction fitting degree of the Fourier series is greater than 94%. The above results show that the fitting effect of the Fourier series increases with the increase of the order, and the fitting effect of the order m>4 is more ideal, so the orders 5, 6, 7, and 8 are selected to form the candidate order set.
[0068] Step d: In order to further determine the optimal order m, it is necessary to further calculate the cross-validation coefficients of Fourier series fitting under different orders. Specifically, fit the 1st to 10th items of the time series, bring the coefficients obtained from the fitting into the fitting equation, calculate the fitting equation, obtain the predicted values of the 11th to 20th items of the time series, and calculate the correlation coefficient between the predicted value and the 11th to 20th items of the time series, denoted as r 1~10Repeat the above steps n times, fit the time series from item 1 to item n, bring the coefficients obtained from the fitting into the fitting equation, calculate the fitting equation, get the predicted values of the time series from item n+1 to item n+10, calculate the correlation coefficient between the predicted values and the time series from item n+1 to item n+10, denoted as r 1~10n ; The calculation formula of the correlation coefficient is: In the above formula, X i is the independent variable of the ith observation, y i is the dependent variable of the ith observation, n represents the number of observations, represents the mean value of the independent variable, represents the mean value of the dependent variable.
[0069] Step e: Calculate the arithmetic average of the correlation coefficients obtained above. The calculation formula of the average correlation coefficient is as follows:
[0070] Reference Figure 6 , showing the cross-validation coefficient of Fourier series fitting under different orders m (i.e., the average correlation coefficient calculated above), given by Figure 6 It can be seen that when the order m = 7, the cross-validation coefficient is 0.48, which is the maximum; when the order m≠7, the cross-validation coefficient is <0.3. Figure 5 and Figure 6 The results show that when the order m = 5, 6, and 8, the Fourier series can fit the North Atlantic Ocean Surface Temperature Index well, but the prediction ability for the future is poor; when the order m = 7, the Fourier series has good fitting effect and prediction ability. In addition, after obtaining the cross-validation coefficient of each order, the credibility of each candidate order can be determined by comparing the correlation coefficient test table with 9 degrees of freedom. The final results show that the credibility corresponding to the order m = 7 is the highest. Therefore, it can be determined that the optimal order finally determined in this actual research process is 7.
[0071] The embodiment of the present application also discloses a system for predicting changes in ocean surface temperature.
[0072] A prediction system for sea surface temperature changes, the prediction system comprising: A data processing module is used to pre-process the historical sea surface temperature data within the target time to obtain the decadal component corresponding to the sea surface temperature time series; A candidate order set generation module is used to traverse the preset order set according to the decadal component to screen and obtain a candidate order set of Fourier series; the preset order set includes several preset orders of Fourier series; A calculation module, for calculating the cross-validation coefficient of the Fourier series fitting at each candidate order in the candidate order set according to the decadal component; The optimal order determination module is used to determine the candidate order with the largest cross-validation coefficient in the candidate order set as the optimal order; the fitting module is used to fit the decadal component through the Fourier series under the optimal order to obtain the periodic variation characteristics of the decadal component of the sea surface temperature; The prediction module is used to predict the decadal changes of sea surface temperature based on the periodic variation characteristics.
[0073] A prediction system for sea surface temperature changes in an embodiment of the present application can implement any of the above prediction methods, and the specific working process of each module in the prediction system can refer to the corresponding process in the above method embodiment.
[0074] In the several embodiments provided in this application, it should be understood that the provided methods and systems can be implemented in other ways. For example, the system embodiments described above are only illustrative; for example, the division of a certain module is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.
[0075] The embodiment of the present application also discloses a computer-readable storage medium.
[0076] A computer-readable storage medium stores a computer program that can be loaded by a processor and execute any one of the above-mentioned methods for predicting changes in sea surface temperature.
[0077] Among them, computer-readable storage media can be any tangible medium that contains or stores a program, which can be used by or in conjunction with an instruction execution system, apparatus or device; the program code contained on the computer-readable medium can be transmitted using any appropriate medium, including but not limited to wireless, wire, optical cable, RF, etc., or any suitable combination of the above.
[0078] It should be noted that in the above embodiments, the description of each embodiment has different emphases, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0079] The above are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Any feature disclosed in this specification (including the abstract and drawings), unless otherwise stated, can be replaced by other equivalent or alternative features with similar purposes. That is, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.
Claims
1. A method for predicting changes in sea surface temperature, It is characterized in that The prediction method comprises: Preprocess the historical sea surface temperature data within the target time to obtain the decadal component corresponding to the sea surface temperature time series; Traversing the preset order set according to the decadal component, screening and obtaining a candidate order set of Fourier series; wherein the preset order set includes a plurality of preset orders of Fourier series; Calculating, according to the decadal component, a cross-validation coefficient of Fourier series fitting at each candidate order in the candidate order set; Based on a preset correlation coefficient test table, determining the credibility corresponding to each candidate order according to the cross-validation coefficient; Determine the candidate order with the highest credibility in the candidate order set as the optimal order; Fitting the interdecadal component by the Fourier series under the optimal order to obtain the periodic variation characteristics of the interdecadal component of the ocean surface temperature; The decadal changes in the ocean surface temperature are predicted based on the periodic variation characteristics.
2. A method for predicting changes in sea surface temperature according to claim 1, It is characterized in that The steps of preprocessing the historical sea surface temperature data within the target time to obtain the decadal component corresponding to the sea surface temperature time series include: Collect sea surface temperature time series based on historical sea surface temperature data within the target time; The sea surface temperature time series is low-pass filtered by sliding average to obtain an interdecadal component; wherein the time window of the sliding average is a preset value.
3. A method for predicting changes in sea surface temperature according to claim 2, It is characterized in that Low-pass filtering the sea surface temperature time series by sliding average includes: Calculate the moving average Y based on the sea surface temperature time series n , the calculation formula is: In the above formula, X (n-k) is the sea surface temperature time series, k and M are the preset sliding average parameters.
4. A method for predicting changes in sea surface temperature according to claim 3, Features: The time window of the sliding average is 9 years, and the 9-year sliding average Y n The expression is: In the above formula, X (n-k) is the sea surface temperature time series, and k is the preset sliding average parameter.
5. A method for predicting changes in ocean surface temperature according to claim 1, It is characterized in that The step of traversing the preset order set according to the decadal component to screen and obtain a candidate order set of Fourier series comprises: According to the preset orders in the preset order set, respectively performing Fourier series fitting of different orders on the decadal components to obtain a Fourier series fitting curve of each preset order; Screening and eliminating the preset orders with abnormalities in the Fourier series fitting curve; Calculate the correction fit corresponding to each preset order in the preset order set after elimination; The preset orders whose correction fitness is greater than a preset threshold value are combined into the candidate order set.
6. A method for predicting changes in ocean surface temperature according to claim 5, It is characterized in that The calculation formula for the corrected goodness of fit is: In the above formula, p is the number of variables, n is the number of samples, and R 2 is the degree of fit.
7. A method for predicting changes in sea surface temperature according to any one of claims 1 to 6, It is characterized in that The step of calculating the cross-validation coefficient of the Fourier series fitting under each candidate order in the candidate order set according to the decadal component comprises: Substituting each candidate order in the candidate order set into the Fourier series formula, fitting the observed values of the 1st to nth items in the sea surface temperature time series, substituting the coefficients obtained by fitting into the fitting equation, predicting the predicted values of the n+1th to n+10th items in the sea surface temperature time series, and calculating the correlation coefficient between the observed values of the n+1th to n+10th items in the sea surface temperature time series and the predicted values; Gradually increase the value of n and repeat the above steps to obtain multiple sets of correlation coefficients between the observed values and the predicted values from the n+1th to the n+10th items, forming a set of correlation coefficients corresponding to each candidate order; The arithmetic mean calculation is performed on the correlation coefficients in the correlation coefficient set to obtain the cross-validation coefficient of the Fourier series fitting under each candidate order.
8. A method for predicting changes in ocean surface temperature according to claim 7, It is characterized in that The calculation formula of the correlation coefficient is: In the above formula, X i is the independent variable of the ith observation, y i is the dependent variable of the ith observation, n represents the number of observations, represents the mean value of the independent variable, represents the mean value of the dependent variable.
9. A prediction system for changes in ocean surface temperature. It is characterized in that The prediction system comprises: A data processing module is used to pre-process the historical sea surface temperature data within the target time to obtain the decadal component corresponding to the sea surface temperature time series; A candidate order set generation module, used for traversing a preset order set according to the decadal component to screen and obtain a candidate order set of Fourier series; the preset order set includes a plurality of preset orders of Fourier series; A calculation module, used for calculating the cross-validation coefficient of the Fourier series fitting under each candidate order in the candidate order set according to the decadal component; An optimal order determination module, used to determine the candidate order with the largest cross-validation coefficient in the candidate order set as the optimal order; A fitting module, used for fitting the interdecadal component by Fourier series under the optimal order to obtain the periodic variation characteristics of the interdecadal component of the sea surface temperature; The prediction module is used to predict the decadal variation of the ocean surface temperature according to the periodic variation characteristics.
10. A computer-readable storage medium, Features: A computer program is stored which can be loaded by a processor and execute the method according to any one of claims 1 to 8.
Citation Information
Patent Citations
Ultra-short-term wind power prediction method and system based on dynamic harmonic regression
CN112070320A
Method and device for predicting rainfall inter-annual change and storage medium
CN116882572A
IDP201502535A
