A method and device for predicting abnormality of meteorological and marine environment

Abstract

The present invention discloses a method and device for predicting abnormalities in a meteorological and marine environment, the method comprising: obtaining a multi-source meteorological and marine data set; the multi-source meteorological and marine data set comprises a data sequence of multiple meteorological parameters collected at the same location; preprocessing the multi-source meteorological and marine data set to obtain a preprocessed data set; performing identification processing on the preprocessed data set and a discriminant model set to obtain an applied discriminant model; performing abnormal prediction processing on the preprocessed data set using the applied discriminant model to obtain an abnormal meteorological probability value. The present invention adopts identification processing to effectively integrate various types of multi-source data information obtained, thereby achieving accurate and efficient advance prediction of abnormal ocean weather and providing a decision-making basis for various marine operations.

Description

Technical Field

[0001] The invention relates to the field of marine meteorological forecasting, and in particular to a method and device for predicting abnormalities in a meteorological and marine environment. Background Art

[0002] At present, with the increasing demand for the development and utilization of marine resources, marine activities are increasing day by day, and modern marine technology and marine engineering equipment have been more and more widely used in marine activities. With the widespread application of information technology in the sea, the impact and restriction of marine operations by meteorological and marine environmental factors are becoming increasingly prominent. The meteorological and marine environment is complex and changeable, has a significant impact on marine activities, and is an indispensable and important part of the marine environment.

[0003] With the continuous development and maturity of meteorological detection technology, more and more high-resolution real-time conventional and unconventional detection data and data subcategories are provided. At present, there are still problems such as low prediction accuracy and complex implementation in extracting abnormal features from various multi-source data information obtained.

[0004] In the field of meteorological and marine environmental forecasting, how to effectively integrate various multi-source data information obtained, accurately and efficiently predict abnormal ocean weather in advance, and provide decision-making basis for various marine operations is an issue that urgently needs to be solved. Summary of the invention

[0005] Aiming at the problem of how to effectively integrate various multi-source meteorological and oceanographic data information obtained and accurately and efficiently predict abnormal ocean weather in advance, the present invention discloses a method and device for predicting abnormal weather in the meteorological and oceanic environment.

[0006] In a first aspect of the embodiments of the present application, a method for predicting abnormalities in a meteorological and marine environment is disclosed, comprising:

[0007] S1, obtaining a multi-source meteorological and oceanographic data set; the multi-source meteorological and oceanographic data set includes a data sequence of multiple meteorological parameters at the same location;

[0008] S2, preprocessing the multi-source meteorological and oceanographic data set to obtain a preprocessed data set;

[0009] S3, performing identification processing on the preprocessed data set and the discriminant model set to obtain an applied discriminant model;

[0010] S4, using the application discrimination model to perform abnormal prediction processing on the preprocessed data set to obtain an abnormal meteorological probability value.

[0011] The preprocessing of the multi-source meteorological and oceanographic data set to obtain a preprocessed data set includes:

[0012] S21, performing outlier removal and smoothing processing on the multi-source meteorological and oceanographic dataset to obtain a first dataset;

[0013] S22, performing acquisition time alignment processing on the first data set to obtain a second data set;

[0014] S23, performing numerical normalization processing on the second data set to obtain a preprocessed data set.

[0015] The multi-source meteorological and oceanographic data set includes a temperature data sequence, a potential height data sequence, a meridional wind speed data sequence, a latitudinal wind speed data sequence, a relative humidity data sequence, a pressure height data sequence, a wind direction data sequence and a dew point temperature data sequence collected at the same location.

[0016] The discriminant model set includes a first discriminant model and a second discriminant model;

[0017] The first discriminant model comprises:

[0018] Obtain the data matrix A;

[0019] The data matrix A is subjected to singular value decomposition to obtain an orthogonal matrix U, a diagonal matrix P and an orthogonal matrix V; the singular value decomposition includes:

[0020] A=UPV;

[0021] For each row vector of the data matrix, using the data collection information corresponding to the row vector as an independent variable and the data of the row vector as a dependent variable, autoregression-sliding average modeling is performed to obtain a regression model for each row vector;

[0022] Using the regression model of each row vector, the mean of the row vectors with the same row number of the orthogonal matrix U is calculated to obtain the weight value of each row vector; the weight values ​​of all row vectors are combined to obtain a weight vector; the weight vector includes m data;

[0023] For each element of the data matrix A, entropy normalization is performed to obtain the entropy matrix B;

[0024] Perform correlation calculation on the data matrix and entropy matrix to obtain the eigenvector;

[0025] The calculation expression of the association calculation is:

[0026]

[0027] Among them, b ij is the element in the i-th row and j-th column of the entropy matrix B, a ijis the element in the i-th row and j-th column of the data matrix A, where the dimension of the data matrix A is m×n, c i is the i-th element of the eigenvector;

[0028] The weight vector is used to perform weighted summation processing on the characteristic vector to obtain the abnormal weather probability value.

[0029] The second discriminant model comprises:

[0030] Obtain the data matrix A;

[0031] Performing time-frequency transformation processing on each row vector of the data matrix A to obtain a corresponding time-frequency sequence;

[0032] Using all the time-frequency sequences, a time-frequency matrix is ​​constructed; the row vector of the time-frequency matrix is ​​the time-frequency sequence;

[0033] Performing cross-correlation calculation on all time-frequency sequences to obtain a cross-correlation matrix; the element in the i-th row and j-th column of the cross-correlation matrix is ​​the normalized cross-correlation value between the i-th time-frequency sequence and the j-th time-frequency sequence;

[0034] The time-frequency matrix is ​​subjected to a first eigenvalue extraction process to obtain a first eigenvector; the expression for the first eigenvalue extraction process is:

[0035]

[0036] Among them, m and n represent the row dimension and column dimension of the time-frequency matrix respectively, and z + is the first eigenvector, z i + represents the i-th data of the first eigenvector, z ij Represents the element in the i-th row and j-th column of the time-frequency matrix;

[0037] The time-frequency matrix is ​​subjected to a second eigenvalue extraction process to obtain a second eigenvector; the expression for the second eigenvalue extraction process is:

[0038]

[0039] Among them, z - is the second eigenvector, z i - Represents the i-th data of the second eigenvector;

[0040] The diagonal elements of the cross-correlation matrix are used to perform a weighted summation of the differences between the first eigenvector and the second eigenvector to obtain an abnormal weather probability value.

[0041] The expression of the difference weighted summation process is:

[0042]

[0043] Among them, t is the probability value of abnormal weather, d i is the i-th diagonal element of the cross-correlation matrix.

[0044] The performing identification processing on the preprocessed data set and the discriminant model set to obtain an applied discriminant model includes:

[0045] The preprocessed data set is represented as a data matrix A; the row vector of the data matrix A is a collection data sequence of a meteorological parameter at the same location;

[0046] The data matrix A is subjected to singular value decomposition to obtain an orthogonal matrix U, a diagonal matrix P and an orthogonal matrix V; the singular value decomposition includes:

[0047] A=UPV;

[0048] Calculate the mean e1 and variance e2 of all diagonal elements of the diagonal matrix P;

[0049] Calculate the quotient of mean e1 and variance e2, e1 / e2;

[0050] Determine whether the quotient is greater than a preset discrimination threshold f to obtain a first discrimination result; if the first discrimination result is greater than, determine that the applied discrimination model is the first discrimination model; if the first discrimination result is not greater than, determine that the applied discrimination model is the second discrimination model.

[0051] In a second aspect of the embodiment of the present application, a device for predicting abnormalities in a meteorological and marine environment is disclosed, the device comprising:

[0052] A memory storing executable program code;

[0053] a processor coupled to the memory;

[0054] The processor calls the executable program code stored in the memory to execute the abnormal prediction method of the meteorological and marine environment.

[0055] In a third aspect of an embodiment of the present application, a computer storable medium is disclosed, wherein the computer storage medium stores computer instructions, and when the computer instructions are called, they are used to execute the abnormality prediction method of the meteorological and marine environment.

[0056] In a fourth aspect of the embodiments of the present application, an information data processing terminal is disclosed, and the information data processing terminal is used to implement the abnormal prediction method of the meteorological and oceanic environment.

[0057] The beneficial effects of the present invention are:

[0058] The present invention realizes the effective fusion of various multi-source data information obtained in the field of meteorological and marine environment prediction, and makes accurate and efficient advance prediction of abnormal ocean weather, providing accurate and timely decision-making basis for various marine operations.

[0059] The present invention performs identification processing on the preprocessed data set and the discriminant model set, uses variance and mean to discriminate the volatility of the data, and adopts different discriminant models for data with different waveform degrees, thereby effectively improving the accuracy of anomaly prediction and reducing the complexity of anomaly prediction.

[0060] In the first discriminant model, the present invention obtains the feature vector by performing correlation calculation on the data matrix and the entropy matrix, thereby effectively realizing the extraction of abnormal features. In the second discriminant model, the present invention realizes the effective fusion and accumulation of various difference values ​​by establishing a difference weighted summation process, thereby ensuring the accuracy of the abnormal prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a flow chart for implementing the method of the present invention. DETAILED DESCRIPTION

[0062] In order to better understand the content of the present invention, an embodiment is given here.

[0063] Figure 1 It is a flow chart for implementing the method of the present invention.

[0064] In a first aspect of the embodiments of the present application, a method for predicting abnormalities in a meteorological and marine environment is disclosed, comprising:

[0065] S1, obtaining a multi-source meteorological and oceanographic data set; the multi-source meteorological and oceanographic data set includes a data sequence of multiple meteorological parameters at the same location;

[0066] S2, preprocessing the multi-source meteorological and oceanographic data set to obtain a preprocessed data set;

[0067] S3, performing identification processing on the preprocessed data set and the discriminant model set to obtain an applied discriminant model;

[0068] S4, using the application discrimination model to perform abnormal prediction processing on the preprocessed data set to obtain an abnormal meteorological probability value.

[0069] The abnormal weather probability value is used to represent the probability value of abnormal weather occurring at the location where the multi-source meteorological and oceanographic data set is collected.

[0070] The preprocessing of the multi-source meteorological and oceanographic data set to obtain a preprocessed data set includes:

[0071] S21, performing outlier removal and smoothing processing on the multi-source meteorological and oceanographic dataset to obtain a first dataset;

[0072] S22, performing acquisition time alignment processing on the first data set to obtain a second data set;

[0073] S23, performing numerical normalization processing on the second data set to obtain a preprocessed data set.

[0074] The multi-source meteorological and oceanographic data set includes a temperature data sequence, a potential height data sequence, a meridional wind speed data sequence, a zonal wind speed data sequence, a relative humidity data sequence, a pressure height data sequence, a wind direction data sequence, and a dew point temperature data sequence collected at the same location;

[0075] The discriminant model set includes a first discriminant model and a second discriminant model;

[0076] The first discriminant model comprises:

[0077] Obtain the data matrix A;

[0078] The data matrix A is subjected to singular value decomposition to obtain an orthogonal matrix U, a diagonal matrix P and an orthogonal matrix V; the singular value decomposition includes:

[0079] A=UPV;

[0080] For each row vector of the data matrix, using the data collection information corresponding to the row vector as an independent variable and the data of the row vector as a dependent variable, autoregression-sliding average modeling is performed to obtain a regression model for each row vector;

[0081] Using the regression model of each row vector, the mean of the row vectors with the same row number of the orthogonal matrix U is calculated to obtain the weight value of each row vector; the weight values ​​of all row vectors are combined to obtain a weight vector; the weight vector includes m data;

[0082] For each element of the data matrix A, entropy normalization is performed to obtain the entropy matrix B;

[0083] Perform correlation calculation on the data matrix and entropy matrix to obtain the eigenvector;

[0084] The calculation expression of the association calculation is:

[0085]

[0086] Among them, b ij is the element in the i-th row and j-th column of the entropy matrix B, a ijis the element in the i-th row and j-th column of the data matrix A, where the dimension of the data matrix A is m×n, c i is the i-th element of the eigenvector;

[0087] The weight vector is used to perform weighted summation processing on the characteristic vector to obtain the abnormal weather probability value.

[0088] The second discriminant model comprises:

[0089] Obtain the data matrix A;

[0090] Performing time-frequency transformation processing on each row vector of the data matrix A to obtain a corresponding time-frequency sequence;

[0091] Using all the time-frequency sequences, a time-frequency matrix is ​​constructed; the row vector of the time-frequency matrix is ​​the time-frequency sequence;

[0092] Perform cross-correlation calculation on all time-frequency sequences to obtain a cross-correlation matrix; the element in the i-th row and j-th column of the cross-correlation matrix is ​​the normalized cross-correlation value between the i-th time-frequency sequence and the j-th time-frequency sequence;

[0093] The first eigenvalue extraction process is performed on the time-frequency matrix to obtain a first eigenvector; the expression of the first eigenvalue extraction process is:

[0094]

[0095] Among them, m and n represent the row dimension and column dimension of the time-frequency matrix respectively, and z + is the first eigenvector, z i + represents the i-th data of the first eigenvector, z ij Represents the element in the i-th row and j-th column of the time-frequency matrix;

[0096] The second eigenvalue extraction process is performed on the time-frequency matrix to obtain a second eigenvector; the expression of the second eigenvalue extraction process is:

[0097]

[0098] Among them, z - is the second eigenvector, z i - Represents the i-th data of the second eigenvector;

[0099] Using the diagonal elements of the cross-correlation matrix, performing a weighted summation process on the first eigenvector and the second eigenvector to obtain an abnormal weather probability value;

[0100] The expression of the weighted summation process is:

[0101]

[0102] Among them, t is the probability value of abnormal weather, d i is the i-th diagonal element of the cross-correlation matrix.

[0103] The performing identification processing on the preprocessed data set and the discriminant model set to obtain an applied discriminant model includes:

[0104] The preprocessed data set is represented as a data matrix A; the row vector of the data matrix A is a collection data sequence of a meteorological parameter at the same location;

[0105] The data matrix A is subjected to singular value decomposition to obtain an orthogonal matrix U, a diagonal matrix P and an orthogonal matrix V; the singular value decomposition includes:

[0106] A=UPV;

[0107] Calculate the mean e1 and variance e2 of all diagonal elements of the diagonal matrix P;

[0108] Calculate the quotient of mean e1 and variance e2, e1 / e2;

[0109] Determine whether the quotient is greater than a preset discrimination threshold f to obtain a first discrimination result; if the first discrimination result is greater than, determine that the applied discrimination model is the first discrimination model; if the first discrimination result is not greater than, determine that the applied discrimination model is the second discrimination model.

[0110] The using the application discrimination model to perform abnormal prediction processing on the pre-processed data to obtain an abnormal meteorological probability value includes:

[0111] The application discrimination model is used to calculate and process the data matrix A corresponding to the pre-processed data to obtain the abnormal meteorological probability value.

[0112] The time-frequency transformation process can be implemented by using Gabor transformation.

[0113] The expression of the entropy normalization process is:

[0114] b ij =-a ij log2a ij ,

[0115] Among them, b ij is the element in the i-th row and j-th column of the entropy matrix;

[0116] The determination threshold f may be [2.5, 1.5].

[0117] The data collection information is the time when the data is collected.

[0118] The method for removing outliers from a wind direction data sequence includes: calculating the modulus and divergence of the gradient value of each data in the wind direction data sequence, and determining whether the quotient of the modulus and divergence of the gradient value is greater than a set threshold value; if so, determining that the data is an outlier and removing the data.

[0119] The outlier elimination and smoothing process include: filling missing values, smoothing noise data, smoothing or deleting outlier points; the smoothed noise data is firstly determined to obtain noise data, and then the noise data is smoothed according to the data before and after the noise data; the noise data is a value whose value is less than the detection sensitivity of the sensor of the observed data, or greater than the measurement upper limit of the sensor of the observed data. The outlier point can be determined by the Kalman filtering method. The filling value of the missing value can be determined by averaging the measured values ​​within a certain sampling interval before and after the missing value.

[0120] The acquisition time alignment process comprises:

[0121] Set a unified data collection time;

[0122] When the data collection time interval of a certain type of data sequence of the multi-source meteorological and oceanographic data set is greater than the time interval of the unified data collection time, interpolation processing is performed on adjacent data to obtain the observed data value at the unified data collection time, and the observed data value is added to the type of data sequence of the multi-source meteorological and oceanographic data set;

[0123] When the data collection time interval of a certain type of data sequence of the multi-source meteorological and oceanographic data set is less than the time interval of the unified data collection time, a number of data in the data sequence are sampled to obtain data consistent with the unified data collection time, thereby realizing the update of the data sequence of the type.

[0124] The normalization process comprises:

[0125] Normalization is to map data with different value ranges to a specified value range. Its mathematical expression is:

[0126]

[0127] Among them, a is a data value in the data set, a max is the maximum value of all data in the data set, a min is the minimum value of all data in the data set, a 归一化 is the value of data a after normalization;

[0128] In a second aspect of the embodiment of the present application, a device for predicting abnormalities in a meteorological and marine environment is disclosed, the device comprising:

[0129] A memory storing executable program code;

[0130] a processor coupled to the memory;

[0131] The processor calls the executable program code stored in the memory to execute the abnormal prediction method of the meteorological and marine environment.

[0132] In a third aspect of an embodiment of the present application, a computer storable medium is disclosed, wherein the computer storage medium stores computer instructions, and when the computer instructions are called, they are used to execute the abnormality prediction method of the meteorological and marine environment.

[0133] In a fourth aspect of the embodiments of the present application, an information data processing terminal is disclosed, and the information data processing terminal is used to implement the abnormal prediction method of the meteorological and oceanic environment.

[0134] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for predicting abnormalities in meteorological and marine environments, characterized in that: include: S1, obtaining a multi-source meteorological and oceanographic data set; the multi-source meteorological and oceanographic data set includes a collection data sequence of multiple meteorological parameters at the same location; S2, preprocessing the multi-source meteorological and oceanographic data set to obtain a preprocessed data set; S3, performing identification processing on the preprocessed data set and the discriminant model set to obtain an applied discriminant model; S4, using the application discrimination model to perform abnormal prediction processing on the preprocessed data set to obtain an abnormal meteorological probability value; The discriminant model set includes a first discriminant model and a second discriminant model; The first discriminant model comprises: Obtain the data matrix A; Performing singular value decomposition on the data matrix A to obtain an orthogonal matrix U, a diagonal matrix P and an orthogonal matrix V; the singular value decomposition includes: A=UPV; For each row vector of the data matrix, using the data collection information corresponding to the row vector as an independent variable and the data of the row vector as a dependent variable, autoregression-sliding average modeling is performed to obtain a regression model for each row vector; Using the regression model of each row vector, the mean of the row vectors with the same row number of the orthogonal matrix U is calculated to obtain the weight value of each row vector; the weight values ​​of all row vectors are combined to obtain a weight vector; the weight vector includes m data; For each element of the data matrix A, entropy normalization is performed to obtain the entropy matrix B; Perform correlation calculation on the data matrix and entropy matrix to obtain the eigenvector; The calculation expression of the association calculation is: Among them, b ij is the element in the i-th row and j-th column of the entropy matrix B, a ij is the element in the i-th row and j-th column of the data matrix A, where the dimension of the data matrix A is m×n, c i is the i-th element of the eigenvector; Performing weighted sum processing on the feature vector using the weight vector to obtain an abnormal weather probability value; The performing identification processing on the preprocessed data set and the discriminant model set to obtain an applied discriminant model includes: The preprocessed data set is represented as a data matrix A; the row vector of the data matrix A is a collection data sequence of a meteorological parameter at the same location; The data matrix A is subjected to singular value decomposition to obtain an orthogonal matrix U, a diagonal matrix P and an orthogonal matrix V; the singular value decomposition includes: A=UPV; Calculate the mean e1 and variance e2 of all diagonal elements of the diagonal matrix P; Calculate the quotient of mean e1 and variance e2, e1 / e2; Determine whether the quotient is greater than a preset discrimination threshold f to obtain a first discrimination result; if the first discrimination result is greater than, determine that the applied discrimination model is the first discrimination model; if the first discrimination result is not greater than, determine that the applied discrimination model is the second discrimination model.

2. The abnormal prediction method of the meteorological and marine environment according to claim 1, characterized in that: The preprocessing of the multi-source meteorological and oceanographic data set to obtain a preprocessed data set includes: S21, performing outlier removal and smoothing processing on the multi-source meteorological and oceanographic dataset to obtain a first dataset; S22, performing acquisition time alignment processing on the first data set to obtain a second data set; S23, performing numerical normalization processing on the second data set to obtain a preprocessed data set.

3. The abnormal prediction method of the meteorological and marine environment according to claim 1, characterized in that: The multi-source meteorological and oceanographic data set includes a temperature data sequence, a potential height data sequence, a meridional wind speed data sequence, a latitudinal wind speed data sequence, a relative humidity data sequence, a pressure height data sequence, a wind direction data sequence and a dew point temperature data sequence collected at the same location.

4. The abnormal prediction method of the meteorological and marine environment according to claim 1, characterized in that: The second discriminant model comprises: Obtain the data matrix A; Performing time-frequency transformation processing on each row vector of the data matrix A to obtain a corresponding time-frequency sequence; Using all the time-frequency sequences, a time-frequency matrix is ​​constructed; the row vector of the time-frequency matrix is ​​the time-frequency sequence; Performing cross-correlation calculation on all time-frequency sequences to obtain a cross-correlation matrix; the element in the i-th row and j-th column of the cross-correlation matrix is ​​the normalized cross-correlation value between the i-th time-frequency sequence and the j-th time-frequency sequence; The time-frequency matrix is ​​subjected to a first eigenvalue extraction process to obtain a first eigenvector; the expression for the first eigenvalue extraction process is: Among them, m and n represent the row dimension and column dimension of the time-frequency matrix respectively, and z + is the first eigenvector, z i + represents the i-th data of the first eigenvector, z ij Represents the element in the i-th row and j-th column of the time-frequency matrix; The time-frequency matrix is ​​subjected to a second eigenvalue extraction process to obtain a second eigenvector; the expression for the second eigenvalue extraction process is: Among them, z - is the second eigenvector, z i - Represents the i-th data of the second eigenvector; The diagonal elements of the cross-correlation matrix are used to perform a weighted summation of the differences between the first eigenvector and the second eigenvector to obtain an abnormal weather probability value.

5. The abnormal prediction method of the meteorological and marine environment according to claim 4, characterized in that: The expression of the difference weighted summation process is: Among them, t is the probability value of abnormal weather, d i is the i-th diagonal element of the cross-correlation matrix.

6. A device for predicting abnormalities in meteorological and marine environments, characterized in that: The device comprises: A memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the abnormal prediction method for the meteorological and marine environment according to any one of claims 1 to 5.

7. A computer storable medium, characterized in that: The computer storage medium stores computer instructions, and when the computer instructions are called, they are used to execute the abnormal prediction method for the meteorological and marine environment according to any one of claims 1 to 5.

8. An information data processing terminal, characterized in that: The information data processing terminal is used to implement the abnormal prediction method of the meteorological and marine environment as described in any one of claims 1 to 5.

Images