Method and system for extracting aperiodic disturbance absolute value

Abstract

The invention discloses an aperiodic disturbance absolute value extraction method and system. The method comprises the steps of obtaining first probability density distribution of an original data set and performing Gaussian fitting to determine a first change range corresponding to a disturbance signal in the original data set; processing the original data set through a white spectrum method to obtain a disturbance relative quantity; obtaining a second distribution characteristic of the relative disturbance quantity and performing Gaussian fitting to obtain a second change range of the relative disturbance quantity; obtaining a disturbance absolute value through a product of the disturbance relative quantity and a proportionality coefficient; the proportionality coefficient is equal to the first variation range divided by the second variation range. Conversion from a relative value to a specific value is achieved, and the scene requirement for quantizing disturbance intensity is met. Meanwhile, disturbance is separated from background signals, so that background information is discussable, and the application range is further expanded.

Description

Technical Field

[0001] The invention relates to the field of satellite meteorological technology, and in particular to a method and system for extracting the absolute value of a non-periodic disturbance. Background Art

[0002] Separating disturbances from their background has always been the core and difficulty of disturbance research in various regions of space weather. Some traditional disturbance extraction methods, such as the monthly median method and the 27-day sliding monthly median method, often define the monthly median (sliding monthly median) of the observation data at a given local time and the difference between it and the original value as the background and disturbance respectively. However, these methods are inert to periodic changes larger than the smoothing window, and therefore cannot accurately reflect the background information. Wang et al. (2014) proposed a new method for identifying non-periodic disturbances, the Spectral Whitening Method (SWM). This method is derived from the spectral whitening technique in statistical estimation theory, which makes the processed data present a uniform distribution in the spectrum (i.e., the form of a white noise spectrum). The whitening method uses the information of the upper envelope of the power spectrum of the original data set to "horizontally" the original power spectrum, which can very effectively remove the periodic component (background) in the data, so that the non-periodic component (disturbance) of interest can be clearly distinguished in the final result, and the final result approximately satisfies the normal distribution.

[0003] At present, the construction of disturbance indexes in the solar, interplanetary, magnetosphere, ionosphere and thermosphere regions has been realized by using the white spectrum method. The basic principles and implementation steps of the existing white spectrum method are mainly as follows:

[0004] 1. Data extension and Fourier transform: By periodically extending the original data set before Fourier transform, the boundary effect generated during the frequency domain conversion process can be avoided;

[0005] 2. Whitening of the power spectrum: by dividing it by the background spectrum, the power spectrum is made to appear in the form of white noise;

[0006] 3. Power spectrum intensity restoration: multiply the whitened spectrum by the mode of the extended data to make the intensity of the spectrum consistent with that before whitening;

[0007] 4. Inverse Fourier transform and truncation of spectrum: Perform inverse Fourier transform on the white spectrum result to the time domain, and truncate the middle part of the extended data to obtain the disturbance.

[0008] For a time series, the SWM algorithm for calculating disturbances is as follows:

[0009]

[0010] The above formula is to obtain the final disturbance result from the original observation data g(t) process.env (ξ) is the upper envelope function of the power spectrum of g(t), and P0 is P env the value that appears most frequently in (ξ), that is, P env the mode of (ξ).

[0011] However, the output of the above white spectrum method is limited to dimensionless relative values and cannot directly provide the absolute magnitude of the perturbation. Specifically, when initially constructing the white spectrum method, in the third step (power spectrum intensity recovery), multiplying the whitened spectrum by the mode of the extended data (i.e., P0 in the formula) is to make the intensity of the spectrum consistent with that before whitening. The white spectrum method believes that P0, as the value that appears most frequently in P env (ξ), that is, P env the mode of (ξ), approximately represents the spectral intensity of the perturbation, so that most of the spectral components in the spectrum remain basically the same before and after whitening. However, it is found that the finally extracted perturbation is not the specific value of the perturbation. This may be because P0 cannot well reflect the spectral intensity of the perturbation, or in the process of reducing the spectral intensity of the periodic component to the noise level during the white spectrum process, a part of new noise is introduced. However, since the final application of the white spectrum method is to construct an index, that is, there is still a process of standardization (dividing the extracted perturbation by its standard deviation), so whether multiplying by P0 in the formula or not will not affect the finally constructed index, because the purpose is to finally obtain only the dimensionless relative quantity after standardization.

[0012] However, this method has the following deficiencies: (1) Limitations of dimensionless results: The results obtained by using the white spectrum method are dimensionless relative values and cannot directly reflect the specific physical value of the perturbation. The existing white spectrum method lacks the ability to describe the specific magnitude of the perturbation, which limits the physical analysis or verification work involving the specific value of the perturbation and also makes it impossible for us to easily separate the perturbation value from the background value in the original dataset. (2) Limited application scope: Due to the inability to provide the specific perturbation value, the results of the white spectrum method can only be used for the analysis of the degree of anomaly and cannot meet the scenario requirements of some situations that need to quantify the specific value of the perturbation intensity. For example, when analyzing the abnormal change of temperature (abnormal high temperature), etc., it is impossible to know how many degrees the temperature is abnormally higher than the normal value. Summary of the Invention

[0013] The present invention provides a method and system for extracting the absolute value of non-periodic perturbation, which is used to solve the problem that the output of the existing white spectrum method is limited to dimensionless relative values and cannot directly provide the absolute magnitude of the perturbation, resulting in its inability to meet the actual scenario requirements of specifically measuring the perturbation intensity. Thus, it limits the application scope and practicality of the white spectrum method in the fields of thermospheric perturbation analysis, space weather monitoring, and even meteorological extreme weather monitoring.

[0014] The present invention includes a method for extracting the absolute value of non-periodic perturbation, characterized in that the method includes:

[0015] Obtain the first probability density distribution of the original data set and perform Gaussian fitting to determine the first variation range corresponding to the perturbation signal in the original data set;

[0016] Process the original data set by the white spectrum method to obtain the relative perturbation amount;

[0017] Obtain the second distribution feature of the relative perturbation amount and perform Gaussian fitting to obtain the second variation range of the relative perturbation amount;

[0018] Obtain the absolute perturbation value by multiplying the relative perturbation amount by the proportionality coefficient; the proportionality coefficient is equal to the first variation range divided by the second variation range.

[0019] Optionally, before obtaining the first probability density distribution of the original data set, the method further includes detrending the original data set; after processing the original data set by the white spectrum method to obtain the relative perturbation amount, the method includes: detrending the relative perturbation amount.

[0020] Optionally, detrending the original data set includes:

[0021] Fit the linear trend T(t)=at + b of the original data set X = {x1, x2,..., x N} by the least squares method, where x represents the i-th original data, N is the total number of data, and t = {t1, t2...., t N} is time, and the slope a and intercept b are respectively:

[0022]

[0023] Calculate the detrended original data set: x i ′ = x i - T(t) = x i -(at + b), and obtain the detrended original data set X′ = {x1′, x2′,..., x N ′}.

[0024] Optionally, obtaining the first probability density distribution of the original data set includes:

[0025] Calculate the normalized histogram of the detrended original data set X' to obtain the empirical probability density distribution; wherein, the histogram is divided into M intervals, and the width of each interval is The probability density is approximately:

[0026]

[0027] Draw a normalized histogram to obtain the first probability density distribution.

[0028] Optionally, the method further includes:

[0029] Perform initial grouping on the detrended original dataset X′ through the K-means clustering algorithm to obtain K clustering centers C k (k = 1, 2,..., K) and their sample assignments;

[0030] Obtain the mixing coefficients The mean μ k = C k , the variance where S k is the sample set of the k-th cluster, and N k is the number of its samples;

[0031] Use the expectation-maximization algorithm to iteratively optimize the parameters θ = {π k , μ k , σ i 2} and generate the mixture probability density function.

[0032] Optionally, using the expectation-maximization algorithm to iteratively optimize the parameters and generate the mixture probability density function includes:

[0033] Calculate the posterior probability γ i that each detrended original data x ik ′ belongs to the t-th Gaussian component;

[0034] Update π ik , μ k , σ k k 2 according to γ;

[0035] Iterate until the change in the log-likelihood value is less than the threshold;

[0036] Obtain the parameters {π k , μ k , σ k 2} of the K Gaussian components and generate the mixture probability density function:

[0037]

[0038] where

[0039] Optionally, the method further includes: identifying the Gaussian component corresponding to the perturbation signal from the multi-Gaussian fitting results and determining the first change range corresponding to the perturbation signal.

[0040] Optionally, the method further includes:

[0041] Analyzing the mean value μ of the K Gaussian components obtained by fitting k , variance σ k 2 and mixing coefficient π k ;

[0042] Selecting the Gaussian component that best matches the characteristics of the perturbation signal, denoted as the m-th Gaussian component (1 ≤ m ≤ K);

[0043] Defining R1 as the width of the m-th Gaussian component, and calculating [μ m - nσ m , μ m + nσ m based on the standard deviation σ m ; where x is the coverage factor;

[0044] Obtaining a first variation range R1 according to the standard deviation.

[0045] In an extraction system for the absolute value of a non-periodic perturbation according to the present invention, the system includes:

[0046] A first acquisition unit, configured to acquire a first probability density distribution of the original data set to determine a first variation range corresponding to the perturbation signal in the original data set;

[0047] A processing unit, configured to process the original data set by the white spectrum method to obtain a relative perturbation amount;

[0048] A second acquisition unit, configured to acquire a second distribution feature of the relative perturbation amount to obtain a second variation range of the relative perturbation amount;

[0049] A calculation unit, configured to obtain the absolute value of the perturbation by multiplying the relative perturbation amount by a proportionality coefficient; the proportionality coefficient is equal to the first variation range divided by the second variation range.

[0050] In a computer-readable storage medium according to the present invention, the computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the method described in any one of the above.

[0051] The method according to the present invention realizes the conversion from a relative value to a specific value, meeting the scenario requirements for quantifying the perturbation intensity. At the same time, the perturbation is separated from the background signal, making the background information also discussable, solving the problem of lack of physical meaning. At the same time, the application range is further expanded. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1It is a flowchart of a method for extracting the absolute value of non-periodic perturbations in an embodiment of the present invention;

[0053] Figure 2 It is a structural diagram of a system for extracting the absolute value of non-periodic perturbations in an embodiment of the present invention;

[0054] Figure 3 It is a multi-Gaussian fitting distribution diagram of the original target temperature data after detrending in an embodiment of the present invention;

[0055] Figure 4 It is a Gaussian distribution diagram corresponding to the white spectrum result after detrending in an embodiment of the present invention;

[0056] Figure 5 It is a schematic diagram of the relative amount of perturbations obtained by the white spectrum method in an embodiment of the present invention;

[0057] Figure 6 It is a schematic diagram of converting the relative amount of perturbations obtained by the white spectrum method into an absolute value in an embodiment of the present invention. Detailed implementation manners

[0058] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present invention, rather than limiting the present invention. In addition, it should be noted that, for the sake of description, only parts related to the present invention are shown in the drawings, rather than all structures.

[0059] It should be understood that in various embodiments herein, the magnitudes of the serial numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments herein.

[0060] Starting from the white spectrum method in the prior art itself, its purpose is neither to obtain nor to obtain the absolute value of the perturbation. Therefore, in the embodiments of the present invention, the "magnitude" of the perturbation, that is, the range or scale of the perturbation change, is explored from the original data, and the proportional coefficient is obtained by comparing with the change range of the relative amount of the perturbation to help the conversion of the relative amount of the perturbation to the absolute value.

[0061] The embodiments of the present invention provide a method for extracting the absolute value of non-periodic perturbations, as Figure 1 shown, the method includes:

[0062] Step 100: Obtain the first probability density distribution of the original data set and perform Gaussian fitting to determine the first variation range corresponding to the perturbation signal in the original data set. Here, the original data set is the target original data to be analyzed. The original data set is processed to obtain the first probability density distribution, that is, the data distribution characteristics of the original data set are obtained. Subsequently, the data distribution characteristics are decomposed by Gaussian fitting to obtain the corresponding Gaussian fitting curve and the data variation range corresponding to the Gaussian fitting curve. This data variation range is the first variation range corresponding to the perturbation signal. Specifically, identify the Gaussian component corresponding to the perturbation signal from the multi-Gaussian fitting results and determine its data variation range as the first variation range.

[0063] Step 200: Process the original data set by the white spectrum method to obtain the relative perturbation quantity. Specifically, apply the white spectrum method (SWM) to the original data set X = {x1, x2,..., x N}, and extract the relative quantity Y = {y1, y2,..., y N} of the perturbation signal. The specific steps of applying the white spectrum method will not be elaborated here.

[0064] Step 300: Obtain the second distribution characteristic of the relative perturbation quantity and perform Gaussian fitting to obtain the second variation range of the relative perturbation quantity. The specific implementation process is similar to Step 100, and finally the second variation range of the relative perturbation quantity is obtained, which will not be elaborated here.

[0065] Step 400: Obtain the absolute perturbation value by multiplying the relative perturbation quantity by the proportionality coefficient; the proportionality coefficient is equal to the first variation range divided by the second variation range. Specifically, a proportionality coefficient is obtained by dividing the variation range of the perturbation signal (i.e., noise) by the variation range of the relative perturbation quantity after the white spectrum method. This proportionality coefficient is the key factor for converting the relative perturbation quantity into an absolute value. Obtaining the product of this proportionality coefficient and the relative perturbation quantity can obtain the absolute perturbation value.

[0066] Specifically, if the first variation range is R1 and the second variation range is R2, then the proportionality coefficient

[0067] is used to multiply the relative perturbation quantity obtained by the white spectrum method to obtain the final absolute perturbation value Z = {z1, z2,..., z N}.

[0068]

[0069] The method described in the above embodiments of the present invention realizes the conversion from relative values to specific numerical values, meeting the scenario requirements that need to quantify the perturbation intensity. At the same time, the perturbation is separated from the background signal, enabling the background information to be discussable and solving the problem of lack of physical meaning. It also further expands the application scope. The new method can be applied to practical scenarios that require absolute numerical data, such as the calculation of extreme temperature thresholds in meteorology. In satellite orbit decay monitoring, it provides accurate absolute values of thermospheric perturbations to optimize trajectory planning and lifetime prediction.

[0070] There are significant limitations in directly performing probability density distribution and Gaussian fitting analysis on the original data. The original data (such as time series temperature data) usually contains multiple signal components such as long-term trends, seasonal variations, and short-term perturbations. If the overall statistical analysis is directly carried out on it, the obtained probability density distribution and Gaussian fitting results will reflect the mixed statistical characteristics of all signals, and the characteristics of the perturbation signal cannot be effectively separated. For example, the numerical range obtained by direct fitting may be significantly affected by the long-term trend, and its magnitude is much larger than the typical magnitude of the perturbation signal, so that the distribution characteristics of the relative amount of perturbation cannot be accurately corresponded. Therefore, in the method for extracting the absolute value of non-periodic perturbation described in the specific embodiments of the present invention, preferably, before obtaining the first probability density distribution of the original data set, the method further includes detrending the original data set; after processing the original data set by the white spectrum method to obtain the relative amount of perturbation, the method includes: detrending the relative amount of perturbation. Specifically, the purpose of detrending the original data set and the relative amount of perturbation is to remove the long-term linear trend in the signal to exclude the interference of the long-term linear trend.

[0071] In the method for extracting the absolute value of non-periodic perturbation described in the specific embodiments of the present invention, preferably, the detrending of the original data set includes:

[0072] Fitting the linear trend T(t)=at + b of the original data set X = {x1, x2,..., x N} by the least squares method, where x i represents the i-th original data, N is the total number of data, t = {t1, t2...., t N} is time, and the slope a and intercept b are respectively:

[0073]

[0074] Calculating the detrended original data set: x i ′ = x i - T(t) = x i - (at + b), and obtaining the detrended original data set X′ = {x1′, x2′,..., x NThe detrending process for the relative disturbance amount can also be performed by using the above-mentioned similar steps, which will not be described in detail.

[0075] The method for extracting the absolute value of the non-periodic disturbance described in the specific embodiment of the present invention preferably obtains the first probability density distribution of the original data set, including:

[0076] Calculate the normalized histogram of the original data set X' after detrending to obtain the empirical probability density distribution; the histogram is divided into M intervals, and the width of each interval is The probability density is approximately:

[0077]

[0078] A normalized histogram is plotted to obtain the first probability density distribution.

[0079] Specifically, the purpose of obtaining the first probability density distribution of the original data set is to perform probability density analysis on the detrended data set and decompose its distribution characteristics through Gaussian fitting.

[0080] In a preferred embodiment, the second probability density distribution of the relative amount of disturbance after detrending is calculated in the same manner, and its distribution characteristics are decomposed by Gaussian fitting.

[0081] The method for extracting the absolute value of the non-periodic disturbance described in the specific embodiment of the present invention preferably further comprises:

[0082] The K-means clustering algorithm is used to initially group the original data set X′ after detrending, and K cluster centers C are obtained. k (k=1, 2, ..., K) and its sample distribution;

[0083] Get the mixing coefficient Mean μ k =C k ,variance Where S k is the sample set of the kth cluster, N k is the number of samples; the mixing coefficient is the weight of K Gaussian distributions.

[0084] The expectation maximization algorithm (EM) is used to iteratively optimize the parameters θ = {π k ,μ k ,σ k 2} and generate a mixed probability density function.

[0085] The method for extracting the absolute value of the non-periodic disturbance described in the specific embodiment of the present invention preferably uses an expectation maximization algorithm to iteratively optimize parameters and generate a mixed probability density function, including:

[0086] Calculate the detrended original data x i The posterior probability γ that ′ belongs to the k-th Gaussian component ik ;

[0087] According to γ ik Update π k , μ k , σ k 2 ;

[0088] Iterate until the change in the log-likelihood value is less than the threshold;

[0089] Obtain the parameters {π k , μ k , σ k 2} of the K Gaussian components, and generate the mixed probability density function:

[0090]

[0091] where

[0092] Preferably, for the detrended relative perturbation amount dataset Y′, calculating its second probability density distribution and analyzing its statistical characteristics through Gaussian fitting can adopt the above similar steps.

[0093] For the method for extracting the absolute value of non-periodic perturbation described in the specific embodiment of the present invention, preferably, the method further includes: identifying the Gaussian component corresponding to the perturbation signal from the multi-Gaussian fitting results and determining the first change range corresponding to the perturbation signal.

[0094] For the method for extracting the absolute value of non-periodic perturbation described in the specific embodiment of the present invention, preferably, the method further includes:

[0095] According to the K Gaussian components obtained by fitting, analyze their means μ k , variances σ k 2 and mixing coefficients π k ;

[0096] Select the Gaussian component that best matches the characteristics of the perturbation signal, denoted as the m-th Gaussian component (1 ≤ m ≤ K);

[0097] Define R1 as the width of the m-th Gaussian component, and calculate [μ m - nσ m - nσ m , μ m + nσ m; where n is the coverage factor; for example, n = 3, corresponding to a probability coverage range of approximately 99.73%.

[0098] Obtain the first variation range R1 according to the standard deviation. The obtained first variation range R1 is 6σ m .

[0099] Preferably, the same method as above can also be used to determine the second variation range R2 corresponding to the relative perturbation amount.

[0100] A specific embodiment of the present invention also provides a system for extracting the absolute value of aperiodic perturbations, such as Figure 2 shown, the system includes:

[0101] The first acquisition unit 201 is used to acquire the first probability density distribution of the original data set to determine the first variation range corresponding to the perturbation signal in the original data set;

[0102] The processing unit 202 is used to process the original data set by the white spectrum method to obtain the relative perturbation amount;

[0103] The second acquisition unit 203 is used to acquire the second distribution feature of the relative perturbation amount to obtain the second variation range of the relative perturbation amount;

[0104] The calculation unit 204 is used to obtain the absolute value of the perturbation by multiplying the relative perturbation amount by the proportionality coefficient; the proportionality coefficient is equal to the first variation range divided by the second variation range.

[0105] The present invention also provides a specific embodiment. This embodiment takes the summer daily average temperature data of 699 meteorological stations in China from the "China Surface Climate Data Daily Value Data Set" as an example to show the application of the method of the present invention. The target original data set X contains the daily average temperature in summer from 1961 to 2022 for a total of 62 years (defined as from June 1st to August 31st each year, a total of 92 days), and the total number of samples N = 62 × 92 = 5704.

[0106] First, perform detrending processing on the target original data set. The target original data set X = {x1, x2,..., x 5704}, where x i is the daily average temperature on the i-th day (unit: °C), and the time indices 1, 2,..., 5704 represent the number of days in 62 years of summer from 1961 to 2022. Apply the least squares method to fit the linear trend T(t) = αt + b to X and perform detrending processing to obtain the detrended data set X' = {x1', x2',..., x 5704 '}.

[0107] Then calculate the normalized histogram for X′, and set the number of intervals M = 30. The histogram shows that the probability density distribution of the data after detrending presents multi-peak characteristics.

[0108] Perform multi-Gaussian fitting: set the number of Gaussian components K = 2, and use K-means clustering initialization parameters - cluster center, initial variance and initial mixing coefficient (which is initialized to uniform distribution, so the mixing coefficient π1 = π2 = 0.5).

[0109] Then determine the Gaussian fitting curve and data variation range R1 corresponding to the disturbance signal in the target original data. In the process of disturbance signal identification, first analyze the two Gaussian components, and determine the component with a high mean and a small variance as the summer high temperature disturbance signal. Figure 3 Yellow line (Gaussian fit component 2).

[0110] Range R1: Take coverage factor n = 3 (about 99.73% coverage), calculate R1 = 6σ m .

[0111] The EM algorithm was used for iterative optimization, and the convergence condition was set to the log-likelihood value change of less than 10. -4 , and get the final parameters.

[0112] Then, the target original data is processed by the White Spectrum Method and the relative amount of disturbance is detrended; specifically, the target original data set X = {x1, x2, ..., x 5704} as g(t) and input into the White Spectrum Method formula:

[0113]

[0114] The relative disturbance amount Y = {y1, y2, ..., y 5704}.

[0115] Remove the fitted linear trend T′(t)=a′t+b′ from Y and get the relative disturbance amount after detrending Y′={y1′, y2′,...,y 5704 ′}.

[0116] Calculate the probability density distribution of the relative amount of disturbance after detrending and perform Gaussian fitting; specifically, calculate the normalized histogram of Y′, such as Figure 4 As shown in the figure, the probability density distribution of the data after detrending presents a unimodal characteristic. Then, Gaussian fitting is performed to obtain the mean and variance.

[0117] Determine the data variation range R2 corresponding to the Gaussian fitting curve of the relative amount of disturbance after detrending; also take the coverage factor n = 3 (about 99.73% coverage range) and calculate R2 = 6σ m .

[0118] Finally, multiply the relative perturbation by R1 / R2 to obtain the absolute value Z of the perturbation.

[0119]

[0120] Figure 5 and Figure 6 are the relative perturbation obtained by the white spectrum method and the absolute value of the perturbation calculated in this embodiment, respectively. The relative perturbation obtained from Figure 5 cannot correspond to specific physical quantities (such as air temperature), and the specific perturbation intensity cannot be obtained, lacking intuitive physical meaning and interpretability. In contrast, the specific values of the summer air temperature perturbation in China from 1961 to 2022 can be seen from Figure 6 . From 1961 to 2022, the overall average summer air temperature anomaly in China showed a significant increasing trend. Especially since 2005, the absolute value of the perturbation shows that the summer air temperature in China is higher than the same period in history, showing a thermal anomaly. The air temperature anomaly in the summer of 2022 was extremely high, reaching the highest value since 1961, with the temperature anomaly about 1.15°C higher than normal, indicating that the summer air temperature thermal anomaly in 2022 has strong extremity. Extreme climate events are related to many influencing factors, including the internal variability of the atmospheric circulation (such as the variation of the Western Pacific subtropical high and the mid-latitude westerly belt), multi-sphere interactions (such as tropical ocean-atmosphere interactions and Arctic ice-atmosphere interactions), etc. In the summer of 2022, the Western Pacific subtropical high was stronger, and the central and eastern parts of China were controlled by an extremely strong abnormal high pressure, with significantly higher convective activities, resulting in extremely severe high-temperature weather in the central and eastern parts of China. In June 2022, the 500hPa geopotential height in the northwestern and northern regions of China was positively anomalous, and these anomalies led to the emergence of a strong high-pressure ridge over the northwestern-northern regions of China, weakening the convective activities, thus increasing the number of extreme high-temperature events in the northwestern and northern regions. These factors may have all affected the extremely high summer air temperature anomaly in China in 2022.

[0121] A specific embodiment of the present invention also provides a computer-readable storage medium, which stores one or more programs, and the one or more programs can be executed by one or more processors to implement the method described in any one of the above embodiments.

[0122] The present invention is an improvement based on the existing white spectrum method, breaking through the limitation in the prior art that only the relative quantity of non-periodic perturbations can be extracted. As a result, the white spectrum method can not only extract relative perturbation information but also further quantitatively obtain the specific physical value of the perturbation. Compared with the prior art, the present invention has the following remarkable advantages: (1) Specific value quantification analysis ability: By calculating and introducing a proportionality coefficient, the result obtained by the white spectrum method in the present invention is no longer an abstract relative quantity but a specific value with clear physical meaning. This means that in many research scenarios, a more in-depth analysis of the physical meaning of its abnormal signals can be carried out. (2) More direct physical meaning: The specific value of the perturbation calculated by the new method is linked to the actual observation or basic physical quantity, and the output has clear physical meaning and unit, thus enhancing the scientificity and practicality of the result. (3) Wider applicability: The present invention expands the scope of application of the white spectrum method in practice, supports various quantitative analysis tasks that require specific values, and is of great value for constructing more accurate perturbation models, etc. For example, in meteorological research, quantifying the change in temperature perturbation can obtain specific temperature thresholds, thereby better identifying extreme weather.

[0123] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0124] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 the function specified in one block or multiple blocks.

[0125] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device that implements the functions in Figure 1 one process or multiple processes and / or blocksFigure 1 The functions specified in one or more boxes.

[0126] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide for implementing the steps of the functions specified in one Figure 1 process or more processes and / or boxes Figure 1 or the steps of the functions specified in one or more boxes.

[0127] The foregoing description of the specific exemplary embodiments of the present invention is for purposes of illustration and exemplification. These descriptions are not intended to limit the present invention to the precise forms disclosed, and it is apparent that many changes and variations are possible in light of the above teachings. The purpose of selecting and describing the exemplary embodiments is to explain the specific principles of the present invention and its practical applications, so that those skilled in the art can implement and utilize the various different exemplary embodiments of the present invention, as well as various different selections and changes. The scope of the present invention is intended to be defined by the claims and their equivalents.

Claims

1. A method for extracting the absolute value of an aperiodic perturbation, characterized in that, The method includes: Obtaining a first probability density distribution of the original data set and performing Gaussian fitting to determine a first variation range corresponding to a perturbation signal in the original data set; Processing the original data set by a white spectrum method to obtain a relative perturbation amount; Obtaining a second distribution feature of the relative perturbation amount and performing Gaussian fitting to obtain a second variation range of the relative perturbation amount; Obtaining an absolute perturbation value by multiplying the relative perturbation amount by a proportionality coefficient; the proportionality coefficient is equal to the first variation range divided by the second variation range.

2. The method for extracting the absolute value of the non-periodic disturbance according to claim 1, wherein Before obtaining the first probability density distribution of the original data set, the method further includes detrending the original data set; After processing the original data set by a white spectrum method to obtain a relative perturbation amount, the method includes: detrending the relative perturbation amount.

3. The method for extracting the absolute value of the non-periodic disturbance according to claim 2, characterized in that, Detrending the original data set includes: Fitting the linear trend T(t)=at + b of the original data set X={x1, x2,..., x N} by the least squares method, where x i represents the i-th original data, N is the total number of data, t={t1, t2...., t N} is the time, and the slope a and the intercept b are respectively: Calculate the detrended original data set: x i ′ = x i - T(t) = x i - (at + b), to obtain the detrended original data set X′ = {x1′, x2′,..., x N ′}.

4. The method for extracting the absolute value of the non-periodic disturbance according to claim 3, wherein Obtaining the first probability density distribution of the original data set includes: Calculate the normalized histogram of the detrended original data set X' to obtain the empirical probability density distribution; where the histogram is divided into M intervals, and the width of each interval is The probability density is approximated as: Drawing a normalized histogram to obtain the first probability density distribution.

5. The method for extracting the absolute value of the non-periodic disturbance according to claim 4, characterized in that The method further includes: The detrended original dataset X′ is initially grouped by the K-means clustering algorithm to obtain K cluster centers C k (k = 1, 2,..., K) and their sample assignments; Obtain the mixing coefficient Mean μ k = C k , Variance where S k is the sample set of the k-th cluster, and N k is its number of samples; Use the Expectation-Maximization algorithm to iteratively optimize the parameter θ = {π k , μ k , σ k 2}, and generate the mixture probability density function.

6. The method for extracting the absolute value of the non-periodic perturbation according to claim 5, wherein Iteratively optimizing parameters and generating a mixture probability density function by using an expectation maximization algorithm includes: Calculate each detrended original data x i The posterior probability γ that ′ belongs to the k-th Gaussian component ik ; According to γ ik Update π k , μ k , σ k 2 ; Iterate until the change in the log-likelihood value is less than the threshold; Obtain the parameters {π k , μ k , σ k 2} of K Gaussian components, and generate the mixed probability density function: Among them 7. The method for extracting the absolute value of the non-periodic perturbation according to claim 6, wherein The method further includes: identifying a Gaussian component corresponding to a perturbation signal from multi-Gaussian fitting results and determining a first variation range corresponding to the perturbation signal.

8. The method for extracting the absolute value of the non-periodic disturbance according to claim 7, wherein, The method further includes: Analyze the mean value μ of the K Gaussian components obtained by fitting k , variance σ k 2 and mixing coefficient π k ; Selecting a Gaussian component that best matches the characteristics of the perturbation signal, denoted as the m-th Gaussian component (1 ≤ m ≤ K); Define R1 as the width of the m-th Gaussian component, based on the standard deviation σ m Calculate [μ m - nσ m , μ m + nσ m ; where n is the coverage factor; Obtaining a first variation range R1 according to the standard deviation.

9. An extraction system for the absolute value of an aperiodic disturbance, characterized in that, The system includes: A first acquisition unit, configured to obtain a first probability density distribution of the original data set to determine a first variation range corresponding to a perturbation signal in the original data set; A processing unit, configured to process the original data set by a white spectrum method to obtain a relative perturbation amount; A second acquisition unit, configured to obtain a second distribution feature of the relative perturbation amount to obtain a second variation range of the relative perturbation amount; A calculation unit, configured to obtain an absolute perturbation value by multiplying the relative perturbation amount by a proportionality coefficient; the proportionality coefficient is equal to the first variation range divided by the second variation range.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the method according to any one of claims 1-8.

Images