CN116187563A - Sea surface temperature space-time intelligent prediction method based on fusion improvement variation modal decomposition - Google Patents
Sea surface temperature space-time intelligent prediction method based on fusion improvement variation modal decomposition Download PDFInfo
- Publication number
- CN116187563A CN116187563A CN202310173686.3A CN202310173686A CN116187563A CN 116187563 A CN116187563 A CN 116187563A CN 202310173686 A CN202310173686 A CN 202310173686A CN 116187563 A CN116187563 A CN 116187563A
- Authority
- CN
- China
- Prior art keywords
- surface temperature
- sea surface
- time
- prediction method
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 42
- 230000004927 fusion Effects 0.000 title claims abstract description 23
- 230000006872 improvement Effects 0.000 title claims abstract description 17
- 238000012360 testing method Methods 0.000 claims abstract description 11
- 238000000605 extraction Methods 0.000 claims abstract description 10
- 239000011159 matrix material Substances 0.000 claims description 35
- 230000006870 function Effects 0.000 claims description 12
- 239000013598 vector Substances 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 11
- 230000003993 interaction Effects 0.000 claims description 10
- 238000001228 spectrum Methods 0.000 claims description 9
- 238000012795 verification Methods 0.000 claims description 9
- 238000011156 evaluation Methods 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 6
- 230000009466 transformation Effects 0.000 claims description 6
- 230000015654 memory Effects 0.000 claims description 3
- KJONHKAYOJNZEC-UHFFFAOYSA-N nitrazepam Chemical compound C12=CC([N+](=O)[O-])=CC=C2NC(=O)CN=C1C1=CC=CC=C1 KJONHKAYOJNZEC-UHFFFAOYSA-N 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 abstract description 4
- 238000012549 training Methods 0.000 abstract description 4
- 238000013135 deep learning Methods 0.000 abstract 1
- 230000000694 effects Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 3
- 230000006641 stabilisation Effects 0.000 description 3
- 238000011105 stabilization Methods 0.000 description 3
- 230000007774 longterm Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000002123 temporal effect Effects 0.000 description 2
- 241000668842 Lepidosaphes gloverii Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000003631 expected effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 239000013535 sea water Substances 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Testing Resistance To Weather, Investigating Materials By Mechanical Methods (AREA)
Abstract
The invention discloses a sea surface temperature space-time intelligent prediction method for fusion improvement variation modal decomposition, which relates to the technical field of sea resource management and predicts the change of the future sea surface temperature according to the historical sea temperature, and comprises the following steps: acquiring sea surface temperature data in a history record; according to the sea surface temperature change in the history record, the data are put into an improved denoising module for denoising treatment, and a training set and a testing set are divided; carrying out space-time feature extraction on the training set by adopting a deep learning method, and fully acquiring the space dependence of sea surface temperature; characteristic data in the training set is used as input model parameters, and corresponding test results are obtained, so that denoising performance can be effectively improved, prediction precision and efficiency can be improved, and ocean resource utilization rate can be improved.
Description
Technical Field
The invention relates to the technical field of ocean resource management, in particular to a sea surface temperature space-time intelligent prediction method for fusion and improvement of variation modal decomposition.
Background
Accurate prediction of sea surface temperature (Sea Surface Temperature, SST) has important significance in the field of sea and meteorological, and due to the characteristics of nonlinearity, multiple noises and the like of sea surface temperature time sequences, a traditional machine learning model cannot fully learn the long-term change rule and the inherent characteristics of SST data, so that the prediction accuracy cannot be improved.
Deep learning models such as Long and Short-term Memory (LSTM) solve the problem of Long-scale dependence and improve SST prediction accuracy, however, the existing model regards each mark point as an independent individual, ignores the spatial interaction influence of the SST mark point, and has strong spatial transmission evolution characteristics along with seawater flow, so that SST prediction has the problems of multiple noise, non-stability and the like, and insufficient spatial characteristics of model extraction can lead to poor prediction results.
Disclosure of Invention
In order to solve the technical problems, the invention provides A, which comprises the following steps
S1, acquiring historical sea surface temperature data of a target point, and dividing the historical sea surface temperature data of the target point into a verification set and a test set;
s2, carrying out normalization processing on the data, decomposing the normalized data into K modal components through an IVMD module, inputting the modal components with correlation greater than 0.5 into a spatial feature extraction module, and using the rest modal components for signal reconstruction;
s3, establishing a weighted adjacent matrix, inputting the weighted adjacent matrix and high-correlation modal components into a GCN layer to extract space interaction features, and then inputting data into a time feature extraction module; capturing the time dynamics of each component after decomposition through an LSTM network layer, and then inputting the extracted space-time characteristics into a prediction module;
s4, inputting the components for extracting the space-time characteristics and the reconstruction components into a prediction module, and performing weighted summation and inverse normalization operation on each output prediction component after the prediction module predicts the space-time characteristics through a Dense layer to obtain a prediction result;
and S5, judging the accuracy of the current predicted sea surface temperature data according to the evaluation index, and finally, weather prediction is carried out on the sea resources according to the predicted sea surface temperature data.
The technical scheme of the invention is as follows:
further, in step S1, the historical sea surface temperature data of the target point is divided into an independent verification set and a test set by adopting a random sampling method, the verification set accounts for 90% of the historical sea surface temperature data of the target point, and the test set accounts for 10% of the historical sea surface temperature data of the target point.
In the foregoing sea surface temperature space-time intelligent prediction method with fusion improved variation modal decomposition, in step S2, the IVMD decomposition method is utilized to decompose and reduce noise on data, and the method comprises the following sub-steps
S2.1, obtaining an analysis signal from a target data set through Hilbert transformation, calculating a single-side spectrum of the analysis signal, and modulating a central band to a corresponding base band through multiplying the single-side spectrum by an operator;
s2.2, calculating square norms of demodulation bases, and calculating bandwidths of each modal component;
s2.3, continuously updating each component and the center frequency thereof by using an alternate direction multiplier method to obtain saddle points of the unconstrained model;
s2.4, determining the optimal decomposition modal number according to the center frequency analysis;
s2.5, optimizing the updating step length according to the average absolute error between the denoising time sequence and the original sequence, and simplifying the updating step length into residual index minimization for calculation;
s2.6, calculating the Pelson correlation coefficient of the decomposed decomposition mode and the original data.
In the foregoing sea surface temperature space-time intelligent prediction method of fusion improved variation modal decomposition, in step S2.1, a target data set is subjected to hilbert transformation to obtain an analysis signal, a single-side spectrum of the analysis signal is calculated, and a central band is modulated to a corresponding baseband by multiplying the single-side spectrum with an operator:
wherein δ (t) represents the dirac function, j represents the imaginary part, t represents the time variable, v k (t) represents the component of the decomposed kth modality,
an exponential signal representing the original signal;
in step S2.2, the square norm of the demodulation base is calculated, and the bandwidth of each modal component is calculated:
wherein f represents an initial signal, which is the number of decomposed modes; { u k }、{ω k The components and center frequencies of the decomposed kth modality are represented respectively; delta (t) represents a dirac function, j represents an imaginary part, x represents a convolution operation, u k (t) represents the component of the decomposed kth modality,
an exponential signal representing the original signal, < >>
Representing the partial derivative operation, t representing the time variable.
In the foregoing sea surface temperature space-time intelligent prediction method for fusion improvement variation modal decomposition, in step S2.3, a lagrangian multiplier τ (t) and a second order penalty factor α are introduced to find an optimal solution for constraint variation problem, and the extended lagrangian expression is as follows:
where alpha is a quadratic penalty factor, lambda is a Lagrange multiplier,<·>for the inner product operation, the method comprises the steps of,
representing the square of the L2 norm.
In the foregoing sea surface temperature space-time intelligent prediction method of fusion improved variation modal decomposition, in step S2.5, the update step length is optimized according to the average absolute error between the denoising time sequence and the original sequence, and is simplified to be calculated by minimizing the residual index:
wherein ,
for the kth modal component after decomposition of the ith signal, f i The original signal is the ith signal, N is the length of the decomposed signal, and K is the number of decomposed modes;
in step S2.6, pearson correlation coefficients of the decomposed decomposition modality and the raw data are calculated:
wherein ,xi To decompose the mode, y i As the original signal is meant to be a signal,
and />
Is the sample mean.
In the foregoing sea surface temperature space-time intelligent prediction method based on fusion and modification of variation modal decomposition, in step S3, the GCN layer is set to 2 layers, and the LSTM network layer is also set to 2 layers.
In the foregoing sea surface temperature space-time intelligent prediction method based on fusion and modification of variation modal decomposition, in step S3, the calculation method for extracting spatial interaction features by the GCN layer is as follows:
in the spectrogram convolution layer, the structural properties of the graph are reflected by the graph convolution laplacian matrix L,
wherein D represents a degree matrix, A represents an adjacency matrix, I N Is a unit matrix;
the GCN layer implements convolution operations with linear operators defining diagonals in the fourier domain,
g θ *x＝Ug θ U T x
wherein ,gθ The method is characterized in that the method comprises the steps of representing a convolution kernel, x represents the extracted features in data, and U consists of feature vectors of L; when the number of adjacent points is more than 10, the method adopts a chebyshev polynomial approximation solution,
wherein ,
representing the sum of the adjacency matrix and the identity matrix, +.>
Is->
Degree matrix of (H) (l) Representing the feature matrix of the first layer, W (l) Represents the weight matrix of the first layer, σ represents the activation function.
In the foregoing sea surface temperature space-time intelligent prediction method for fusion and improvement of variation modal decomposition, in step S3, the calculation method for extracting time features by the LSTM network layer is as follows:
f t ＝σ(W xf x t +W hf h t-1 +b f )
i t ＝σ(W xi x t +W hi h t-1 +b i )
o t ＝σ(W xo x t +W ho h t-1 +b o )
wherein ,xt Is the input vector, i t Is the input state in time step t, f t Is the forget state in time step t, o t Is the output state in time step t, h t (h t -1) is the hidden state in time step t (t-1), c t (c t -1) is the state of the cell in time step t (t-1);
adding nonlinearity at the top of the three gates in the form of tanh and sigmoid activation functions sigma; w (W) xf 、W hf 、W xi 、W hi 、W xo 、W ho 、W xc and Whc B represents weight vectors corresponding to the input vector, the forget gate, the input gate, the memory unit and the output gate respectively f 、b i 、b c B o Are all the bias variables, and the bias variables,
ha is a matrixAnd d, a dacard product.
In the foregoing sea surface temperature space-time intelligent prediction method for fusion improvement variation modal decomposition, in step S5, the evaluation indexes are set as root mean square error, average absolute error and average absolute percentage error, and the calculation formula is as follows:
wherein n is the number of samples, y i The actual value is represented by a value that is,
representing the predicted value.
The beneficial effects of the invention are as follows:
in the invention, the IVMD is adopted to perform noise reduction and stabilization treatment on the time series data, so that the noise reduction performance is improved, and the data becomes more stable; the decomposed high-correlation mode is put into a depth model for space-time feature extraction, and the rest mode components are used for signal reconstruction; the GCN utilizes a marker dot diagram network structure, can fully extract highly significant interaction characteristics in a space domain, and combines LSTM capturing time dynamics, so that model prediction accuracy is effectively improved, and resource utilization is improved.
Drawings
FIG. 1 is an overall flow chart of a predictive model in an embodiment of the invention;
FIG. 2 is an exploded block diagram of an improved variation mode in an embodiment of the invention;
FIG. 3 is a schematic diagram showing the effect of the signal stabilization process according to the embodiment of the present invention;
FIG. 4 is a graph comparing the prediction fit of an embodiment of the present invention to an existing model.
Detailed Description
The sea surface temperature space-time intelligent prediction method for fusion improved variation modal decomposition provided by the embodiment, as shown in figure 1, comprises the following steps of
S1, acquiring historical sea surface temperature data of a target point, and dividing the historical sea surface temperature data of the target point into a verification set and a test set;
the random sampling method is adopted to divide the historical sea surface temperature data of the target point into an independent verification set and a test set, wherein the verification set accounts for 90% of the historical sea surface temperature data of the target point, and the test set accounts for 10% of the historical sea surface temperature data of the target point.
S2, carrying out normalization processing on the data, decomposing the normalized data into K modal components through an improved variation modal decomposition (Improved Variational Mode Decomposition, IVMD) module, inputting the modal components with high correlation (namely, correlation is more than 0.5) into a spatial feature extraction module, and using the rest modal components for signal reconstruction.
As shown in FIG. 2, the IVMD decomposition method is used to decompose and reduce noise of the data, and the effect of the IVMD module signal stabilization is shown in FIG. 3, which comprises the following steps
S2.1, obtaining an analysis signal by Hilbert transformation of a target data set, calculating a single-side spectrum of the analysis signal, modulating a central band to a corresponding baseband by multiplying the analysis signal with an operator,
wherein δ (t) represents the dirac function, j represents the imaginary part, t represents the time variable, v k (t) represents the component of the decomposed kth modality,
an exponential signal representing the original signal;
s2.2, calculating square norms of demodulation bases, calculating bandwidths of each modal component,
wherein f represents an initial signal, which is the number of decomposed modes; { u k }、{ω k The components and center frequencies of the decomposed kth modality are represented respectively; delta (t) represents a dirac function, j represents an imaginary part, x represents a convolution operation, u k (t) represents the component of the decomposed kth modality,
an exponential signal representing the original signal, < >>
Representing partial derivative operation, t representing time variable;
s2.3, continuously updating each component and the center frequency thereof by using an alternate direction multiplier method to obtain saddle points of the unconstrained model; wherein { u } k The decomposed IMF component is represented by { omega } k -representing the center frequency of each component; in order to find the optimal solution of the constraint variation problem, a Lagrangian multiplier τ (t) and a second order penalty factor α are introduced, and the extended Lagrangian expression is as follows:
where alpha is a quadratic penalty factor, lambda is a Lagrange multiplier,<·>for the inner product operation, the method comprises the steps of,
represents the square of the L2 norm;
s2.4, determining the optimal decomposition modal number according to the center frequency analysis;
s2.5, optimizing the update step length according to the average absolute error between the denoising time sequence and the original sequence, simplifying the optimization to be calculated by minimizing the residual error index,
wherein ,
for the kth modal component after decomposition of the ith signal, f i The original signal is the ith signal, N is the length of the decomposed signal, and K is the number of decomposed modes;
s2.6, calculating the Pelson correlation coefficient of the decomposed decomposition mode and the original data,
wherein ,xi To decompose the mode, y i As the original signal is meant to be a signal,
and />
Is the sample mean.
S3, establishing a weighted adjacent matrix, inputting the weighted adjacent matrix and high-correlation modal components into a GCN layer to extract space interaction features, setting the GCN layer to be 2 layers, and then inputting data into a time feature extraction module; the temporal dynamics of the decomposed components are captured by the LSTM network layer, which is set to 2 layers, and then the extracted spatio-temporal features are input to the prediction module.
The calculation method for extracting the space interaction characteristics by the GCN layer comprises the following steps:
in the spectrogram convolution layer, the best reflection of the structural properties of the graph is the graph convolution laplace matrix L,
wherein D represents a degree matrix, A represents an adjacency matrix, I N Is a unit matrix;
unlike the convolution operation of classical convolution operators, GCN implements the convolution operation with linear operators defining diagonals in the fourier domain,
g θ *x＝Ug θ U T x
wherein ,gθ The method is characterized in that the method comprises the steps of representing a convolution kernel, x represents the extracted features in data, and U consists of feature vectors of L; when the number of adjacent points is larger than 10, namely the graph structure is larger, the method adopts the Chebyshev polynomial of the cutting ratio to approximate solving,
wherein ,
representing the sum of the adjacency matrix and the identity matrix, +.>
Is->
Degree matrix of (H) (l) Representing the feature matrix of the first layer, W (l) Represents the weight matrix of the first layer, σ represents the activation function.
In step S3, the calculation method for extracting the time features by the LSTM network layer is as follows:
f t ＝σ(W xf x t +W hf h t-1 +b f )
i t ＝σ(W xi x t +W hi h t-1 +b i )
o t ＝σ(W xo x t +W ho h t-1 +b o )
wherein ,xt Is the input vector, i t Is the input state in time step t, f t Is the forget state in time step t, o t Is the output state in time step t, h t (h t -1) is the hidden state in time step t (t-1), c t (c t -1) is the state of the cell in time step t (t-1);
adding nonlinearity at the top of the three gates in the form of tanh and sigmoid activation functions sigma; w (W) xf 、W hf 、W xi 、W hi 、W xo 、W ho 、W xc and Whc B represents weight vectors corresponding to the input vector, the forget gate, the input gate, the memory unit and the output gate respectively f 、b i 、b c B o Are all the bias variables, and the bias variables,
is the Hadamard product of the matrix.
Workflow of GCN: and taking the decomposed mode as input, converting the data into three dimensions, inputting the three dimensions into a GCN layer for spatial feature extraction, and setting the layer numbers to 256 and 128 respectively.
Workflow of LSTM: the extracted spatial feature data is input to the LSTM layer to capture temporal dynamics of SST data, and then the spatio-temporal feature data is input to the fully connected layer prediction output.
S4, inputting the components for extracting the space-time characteristics and the reconstruction components into a prediction module, and performing weighted summation and inverse normalization operation on each output prediction component after the prediction module predicts the space-time characteristics through a Dense layer to obtain a prediction result;
s5, judging the accuracy of the current predicted sea surface temperature data according to the evaluation index, and finally, weather prediction is carried out on the sea resources according to the predicted sea surface temperature data;
to verify the validity of the model proposed in this embodiment, using Root Mean Square Error (RMSE), mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE) as evaluation indexes, the calculation formula is as follows:
wherein n is the number of samples, y i The actual value is represented by a value that is,
representing the predicted value.
In order to verify the validity and applicability of each module of the model of this embodiment, on the basis of the same data set, RMSE, MAE and MAPE are used as reference indexes to compare the prediction effects of the baseline model and the existing model, the experimental results are shown in table 1, and the prediction fit pair of this embodiment and the existing model is shown in fig. 4.
Table 1 evaluation index value of existing model
Compared with the SVR baseline model, the model of the embodiment reduces 77.31%, 78.40% and 74.02% in terms of RMSE, MAE and MAPE indexes, has high training speed, but cannot fully learn the long-term change rule and the inherent characteristics of SST data, and the predicted result cannot reach the expected effect.
Compared with STL-LSTM and EMD-GRU, the model of the embodiment has the advantages that the RMSE and the MAE are respectively reduced by 54.91%, 52.85% and 58.02% and 54.01%, and the MAPE is reduced by 57.47% and 53.74%, mainly because the model ignores the space interaction effect of the mark points; compared with CNN-LSTM, the RMSE, MAE and MAPE of the model of the embodiment are respectively reduced by 50.27%, 51.50% and 52.42%, because the influence of noise of data on prediction accuracy is not considered; compared with VMD-CNN-LSTM, the RMSE, MAE and MAPE of the model of the embodiment are reduced by 19.75%, 20.42% and 18.39%, respectively, mainly because the model does not consider the denoising performance of the decomposition module and omits the interaction effect of the mark points, so that the prediction effect cannot be improved.
Experiments show that the prediction index of the model of the embodiment is superior to that of the existing model, and the stability and effectiveness of the model for predicting SST data are proved.
In addition to the embodiments described above, other embodiments of the invention are possible. All technical schemes formed by equivalent substitution or equivalent transformation fall within the protection scope of the invention.
Claims (10)
1. A sea surface temperature space-time intelligent prediction method for fusion improvement of variation modal decomposition is characterized by comprising the following steps: comprises the following steps
S1, acquiring historical sea surface temperature data of a target point, and dividing the historical sea surface temperature data of the target point into a verification set and a test set;
s2, carrying out normalization processing on the data, decomposing the normalized data into K modal components through an IVMD module, inputting the modal components with correlation greater than 0.5 into a spatial feature extraction module, and using the rest modal components for signal reconstruction;
s3, establishing a weighted adjacent matrix, inputting the weighted adjacent matrix and high-correlation modal components into a GCN layer to extract space interaction features, and then inputting data into a time feature extraction module; capturing the time dynamics of each component after decomposition through an LSTM network layer, and then inputting the extracted space-time characteristics into a prediction module;
s4, inputting the components for extracting the space-time characteristics and the reconstruction components into a prediction module, and performing weighted summation and inverse normalization operation on each output prediction component after the prediction module predicts the space-time characteristics through a Dense layer to obtain a prediction result;
and S5, judging the accuracy of the current predicted sea surface temperature data according to the evaluation index, and finally, weather prediction is carried out on the sea resources according to the predicted sea surface temperature data.
2. The sea surface temperature space-time intelligent prediction method for fusion improvement variation modal decomposition according to claim 1, wherein the sea surface temperature space-time intelligent prediction method is characterized by comprising the following steps of: in the step S1, the historical sea surface temperature data of the target point is divided into an independent verification set and a test set by adopting a random sampling method, the verification set accounts for 90% of the historical sea surface temperature data of the target point, and the test set accounts for 10% of the historical sea surface temperature data of the target point.
3. The sea surface temperature space-time intelligent prediction method for fusion improvement variation modal decomposition according to claim 1, wherein the sea surface temperature space-time intelligent prediction method is characterized by comprising the following steps of: in the step S2, the IVMD decomposition method is utilized to decompose and reduce noise of the data, and the method comprises the following steps of
S2.1, obtaining an analysis signal from a target data set through Hilbert transformation, calculating a single-side spectrum of the analysis signal, and modulating a central band to a corresponding base band through multiplying the single-side spectrum by an operator;
s2.2, calculating square norms of demodulation bases, and calculating bandwidths of each modal component;
s2.3, continuously updating each component and the center frequency thereof by using an alternate direction multiplier method to obtain saddle points of the unconstrained model;
s2.4, determining the optimal decomposition modal number according to the center frequency analysis;
s2.5, optimizing the updating step length according to the average absolute error between the denoising time sequence and the original sequence, and simplifying the updating step length into residual index minimization for calculation;
s2.6, calculating the Pelson correlation coefficient of the decomposed decomposition mode and the original data.
4. The sea surface temperature space-time intelligent prediction method for fusion improvement variation modal decomposition according to claim 3, wherein the sea surface temperature space-time intelligent prediction method is characterized by comprising the following steps of: in the step S2.1, the target data set is subjected to hilbert transformation to obtain an analysis signal, a single side spectrum of the analysis signal is calculated, and the central band is modulated to a corresponding baseband by multiplying the single side spectrum with an operator:
wherein δ (t) represents the dirac function, j represents the imaginary part, t represents the time variable, v k (t) represents the component of the decomposed kth modality,
an exponential signal representing the original signal;
in step S2.2, the square norm of the demodulation base is calculated, and the bandwidth of each modal component is calculated:
wherein f represents an initial signal, which is the number of decomposed modes; { u k }、{ω k The components and center frequencies of the decomposed kth modality are represented respectively; delta (t) represents a dirac function, j represents an imaginary part, x represents a convolution operation, u k (t) represents the component of the decomposed kth modality,
an exponential signal representing the original signal, < >>
Representing partial derivative operation, t representsTime variable.
5. The sea surface temperature space-time intelligent prediction method for fusion improvement variation modal decomposition according to claim 3, wherein the sea surface temperature space-time intelligent prediction method is characterized by comprising the following steps of: in the step S2.3, a lagrangian multiplier τ (t) and a second order penalty factor α are introduced to find an optimal solution for the constraint variation problem, where the extended lagrangian expression is as follows:
wherein alpha is a quadratic penalty factor, lambda is a Lagrangian multiplier, and lambda is an inner product operation,
representing the square of the L2 norm.
6. The sea surface temperature space-time intelligent prediction method for fusion improvement variation modal decomposition according to claim 3, wherein the sea surface temperature space-time intelligent prediction method is characterized by comprising the following steps of: in the step S2.5, the update step is optimized according to the average absolute error between the denoising time sequence and the original sequence, and is simplified to be calculated by minimizing the residual index:
wherein ,
for the kth modal component after decomposition of the ith signal, f i The original signal is the ith signal, N is the length of the decomposed signal, and K is the number of decomposed modes;
in step S2.6, pearson correlation coefficients of the decomposed decomposition modality and the raw data are calculated:
wherein ,xi To decompose the mode, y i As the original signal is meant to be a signal,
and />
Is the sample mean.
7. The sea surface temperature space-time intelligent prediction method for fusion improvement variation modal decomposition according to claim 1, wherein the sea surface temperature space-time intelligent prediction method is characterized by comprising the following steps of: in the step S3, the GCN layer is set to 2 layers, and the LSTM network layer is also set to 2 layers.
8. The sea surface temperature space-time intelligent prediction method for fusion improvement variation modal decomposition according to claim 1, wherein the sea surface temperature space-time intelligent prediction method is characterized by comprising the following steps of: in the step S3, the calculation method for extracting the spatial interaction features by the GCN layer is as follows:
in the spectrogram convolution layer, the structural properties of the graph are reflected by the graph convolution laplacian matrix L,
wherein D represents a degree matrix, A represents an adjacency matrix, I N Is a unit matrix;
the GCN layer implements convolution operations with linear operators defining diagonals in the fourier domain,
g θ *x＝Ug θ U T x
wherein ,gθ The method is characterized in that the method comprises the steps of representing a convolution kernel, x represents the extracted features in data, and U consists of feature vectors of L; when the number of adjacent points is more than 10, the method adopts a chebyshev polynomial approximation solution,
wherein ,
representing the sum of the adjacency matrix and the identity matrix, +.>
Is->
Degree matrix of (H) (l) Representing the feature matrix of the first layer, W (l) Represents the weight matrix of the first layer, σ represents the activation function.
9. The sea surface temperature space-time intelligent prediction method for fusion improvement variation modal decomposition according to claim 1, wherein the sea surface temperature space-time intelligent prediction method is characterized by comprising the following steps of: in the step S3, the calculation method for extracting the time features by the LSTM network layer is as follows:
f t ＝σ(W xf x t +W hf h t-1 +b f )
i t ＝σ(W xi x t +W hi h t-1 +b i )
o t ＝σ(W xo x t +W ho h t-1 +b o )
wherein ,xt Is the input vector, i t Is the input state in time step t, f t Is the forget state in time step t, o t Is the output state in time step t, h t (h t -1) is the hidden state in time step t (t-1), c t (c t -1) is the state of the cell in time step t (t-1);
adding nonlinearity at the top of the three gates in the form of tanh and sigmoid activation functions sigma; w (W) xf 、W hf 、W xi 、W hi 、W xo 、W ho 、W xc and Whc B represents weight vectors corresponding to the input vector, the forget gate, the input gate, the memory unit and the output gate respectively f 、b i 、b c B o Are all the bias variables, and the bias variables,
is the Hadamard product of the matrix.
10. The sea surface temperature space-time intelligent prediction method for fusion improvement variation modal decomposition according to claim 1, wherein the sea surface temperature space-time intelligent prediction method is characterized by comprising the following steps of: in the step S5, the evaluation index is set as a root mean square error, an average absolute error and an average absolute percentage error, and the calculation formula is as follows:
wherein n is the number of samples, y i The actual value is represented by a value that is,
representing the predicted value. />
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310173686.3A CN116187563A (en) | 2023-02-28 | 2023-02-28 | Sea surface temperature space-time intelligent prediction method based on fusion improvement variation modal decomposition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310173686.3A CN116187563A (en) | 2023-02-28 | 2023-02-28 | Sea surface temperature space-time intelligent prediction method based on fusion improvement variation modal decomposition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116187563A true CN116187563A (en) | 2023-05-30 |
Family
ID=86446029
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310173686.3A Pending CN116187563A (en) | 2023-02-28 | 2023-02-28 | Sea surface temperature space-time intelligent prediction method based on fusion improvement variation modal decomposition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116187563A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116976149A (en) * | 2023-09-22 | 2023-10-31 | 广东海洋大学 | Sea surface temperature prediction method |
| CN117151303A (en) * | 2023-09-12 | 2023-12-01 | 河海大学 | Ultra-short-term solar irradiance prediction method and system based on hybrid model |
| CN117151303B (en) * | 2023-09-12 | 2024-05-31 | 河海大学 | Ultra-short-term solar irradiance prediction method and system based on hybrid model |
-
2023
- 2023-02-28 CN CN202310173686.3A patent/CN116187563A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117151303A (en) * | 2023-09-12 | 2023-12-01 | 河海大学 | Ultra-short-term solar irradiance prediction method and system based on hybrid model |
| CN117151303B (en) * | 2023-09-12 | 2024-05-31 | 河海大学 | Ultra-short-term solar irradiance prediction method and system based on hybrid model |
| CN116976149A (en) * | 2023-09-22 | 2023-10-31 | 广东海洋大学 | Sea surface temperature prediction method |
| CN116976149B (en) * | 2023-09-22 | 2023-12-29 | 广东海洋大学 | Sea surface temperature prediction method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Parameter estimation of underwater impulsive noise with the Class B model | |
| CN108696331B (en) | Signal reconstruction method based on generation countermeasure network | |
| CN105260998A (en) | MCMC sampling and threshold low-rank approximation-based image de-noising method | |
| CN111785329A (en) | Single-cell RNA sequencing clustering method based on confrontation automatic encoder | |
| CN109740057B (en) | Knowledge extraction-based enhanced neural network and information recommendation method | |
| Lewis et al. | Modeling long-range dependence, nonlinearity, and periodic phenomena in sea surface temperatures using TSMARS | |
| CN112183742B (en) | Neural network hybrid quantization method based on progressive quantization and Hessian information | |
| CN113852432A (en) | RCS-GRU model-based spectrum prediction sensing method | |
| CN111931983A (en) | Precipitation prediction method and system | |
| CN116187563A (en) | Sea surface temperature space-time intelligent prediction method based on fusion improvement variation modal decomposition | |
| CN111523644A (en) | Intermediate-long term runoff prediction method based on LSFL combined model | |
| CN106528679A (en) | Time series analysis method based on multilinear autoregression model | |
| CN117636183A (en) | Small sample remote sensing image classification method based on self-supervision pre-training | |
| CN111310996B (en) | User trust relation prediction method and system based on graph self-coding network | |
| Feng et al. | An expedient DDF-based implementation of perfectly matched monolayer | |
| Djurić et al. | Sequential parameter estimation of time-varying non-Gaussian autoregressive processes | |
| CN111401155B (en) | Image recognition method of residual error neural network based on implicit Euler jump connection | |
| CN114707348A (en) | Partial least square Kriging model-assisted efficient global optimization method for aviation speed reducer | |
| CN114764577A (en) | Lightweight modulation recognition model based on deep neural network and method thereof | |
| Pan et al. | Specific radar emitter identification using 1D-CBAM-ResNet | |
| CN113688774B (en) | Advanced learning-based high-rise building wind induced response prediction and training method and device | |
| CN114189876B (en) | Flow prediction method and device and electronic equipment | |
| CN113987909B (en) | Oilpaper insulation aging prediction method, device, computer equipment and storage medium | |
| CN117849901A (en) | Method and system for remote sensing estimation of atmospheric parameters taking scene representativeness into consideration | |
| Xie | Statistical analysis based on Lake Michigan fish acoustic data using LASSO method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |