CN116611580A - Ocean red tide prediction method based on multi-source data and deep learning - Google Patents
Ocean red tide prediction method based on multi-source data and deep learning Download PDFInfo
- Publication number
- CN116611580A CN116611580A CN202310739974.0A CN202310739974A CN116611580A CN 116611580 A CN116611580 A CN 116611580A CN 202310739974 A CN202310739974 A CN 202310739974A CN 116611580 A CN116611580 A CN 116611580A
- Authority
- CN
- China
- Prior art keywords
- data
- red tide
- cnn
- lstm
- prediction model
- 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 35
- 238000013135 deep learning Methods 0.000 title claims abstract description 20
- 230000007613 environmental effect Effects 0.000 claims abstract description 29
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 9
- 238000011156 evaluation Methods 0.000 claims abstract description 8
- 238000012544 monitoring process Methods 0.000 claims abstract description 5
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 238000012360 testing method Methods 0.000 claims abstract description 5
- 238000005457 optimization Methods 0.000 claims abstract description 3
- 238000012549 training Methods 0.000 claims description 15
- ATNHDLDRLWWWCB-AENOIHSZSA-M chlorophyll a Chemical compound C1([C@@H](C(=O)OC)C(=O)C2=C3C)=C2N2C3=CC(C(CC)=C3C)=[N+]4C3=CC3=C(C=C)C(C)=C5N3[Mg-2]42[N+]2=C1[C@@H](CCC(=O)OC\C=C(/C)CCC[C@H](C)CCC[C@H](C)CCCC(C)C)[C@H](C)C2=C5 ATNHDLDRLWWWCB-AENOIHSZSA-M 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 239000013535 sea water Substances 0.000 claims description 8
- 229930002868 chlorophyll a Natural products 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- 238000005065 mining Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 6
- 229930002875 chlorophyll Natural products 0.000 claims description 5
- 235000019804 chlorophyll Nutrition 0.000 claims description 5
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 claims description 4
- 229910052760 oxygen Inorganic materials 0.000 claims description 4
- 239000001301 oxygen Substances 0.000 claims description 4
- 229920006395 saturated elastomer Polymers 0.000 claims description 4
- 239000000654 additive Substances 0.000 claims description 3
- 230000000996 additive effect Effects 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000005855 radiation Effects 0.000 claims description 3
- 230000001419 dependent effect Effects 0.000 claims description 2
- 230000001502 supplementing effect Effects 0.000 claims description 2
- 238000001514 detection method Methods 0.000 abstract description 4
- 230000000875 corresponding effect Effects 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012851 eutrophication Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000000790 scattering method Methods 0.000 description 1
- 238000012731 temporal analysis Methods 0.000 description 1
- 238000000700 time series analysis Methods 0.000 description 1
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"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
-
- 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
Abstract
The invention relates to a ocean red tide prediction method based on multi-source data and deep learning. On one hand, remote sensing data, water quality buoy monitoring data, meteorological data and manual detection data are combined, and on the other hand, a deep learning method based on CNN-LSTM is adopted to predict red tide by utilizing multiple characteristic factors of time sequences. The process comprises the following steps: a. constructing a data set, integrating and preprocessing environmental factor data; b. analyzing the correlation between the environmental factors and the red tide by adopting the Pearson coefficient, and analyzing the correlation between the multi-environmental factor combination and the red tide by utilizing the complex correlation coefficient; c. constructing a CNN-LSTM prediction model, and carrying out parameter optimization adjustment on the trained CNN-LSTM prediction model; d. and testing the model by using the multi-environment factors of different combinations, and performing performance evaluation. The method combines remote sensing image data and water quality detection data, increases the data quantity, improves the complexity of a data set and improves the accuracy of red tide forecasting.
Description
Technical Field
The invention relates to the field of marine ecological environment, in particular to the field of marine water quality monitoring, in particular to a marine red tide prediction method based on multi-source data and deep learning.
Background
Red tide is taken as a disaster which can greatly harm the marine environment, and brings potential safety hazard to all coastal cities. The occurrence of red tide can influence ecological balance, so that ocean pollution is caused, the red tide is monitored in real time as soon as possible, the development condition of the red tide is fully mastered, and the occurrence of the red tide is accurately predicted. Traditional empirical, statistical, and numerical model analysis methods are increasingly difficult to adapt to large-scale data scenarios. And the k-nearest neighbor and random forest algorithm, the back scattering method, the chlorophyll anomaly method and other machine learning methods for exploring the probability of generating red tide phenomenon through image detection. Due to the limited expressive power for features in large data volumes and complex scenarios. The traditional method has low prediction accuracy, low model complexity and insufficient applicability.
In recent years, with the development of deep learning, more and more deep neural networks have been developed, they have been intensively studied in many fields, and have had an extraordinary effect on the solution of a series of problems. To solve the above problem of limited expression of machine learning features, recent patents have combined prediction of red tide occurrence with deep learning. The Chinese patent with publication number of CN112365093A discloses a multi-feature factor red tide prediction model based on GRU deep learning, which comprises the steps of preprocessing collected data, analyzing the relevance of feature influence factors and red tide, constructing the GRU prediction model, training the model by utilizing multiple combination feature factors, evaluating the performance of the model and the like. When the method works, the multi-combination characteristic factors are used as input variables, and the probability of red tide occurrence is output through the trained GRU model. The model effectively combines deep learning, predicts the occurrence probability of the red tide by utilizing multiple combination characteristic factors, has a good expression effect on characteristics, and improves the accuracy of the red tide occurrence prediction, but the model has low data complexity, low applicability and rising space of prediction accuracy. Another example is chinese patent publication No. CN112084716a, which discloses a red tide prediction and early warning method based on eutrophication comprehensive evaluation, which includes data acquisition and correlation processing, and the obtained correlated data establishes decision tree, establishes and trains neural network prediction model, and the like. When the method is used, a high confidence coefficient data set and a low confidence coefficient data set are used for inputting a neural network prediction model, and a final prediction result is obtained; and sending out corresponding early warning information according to the final prediction result. The method can improve the defects of the prior art and improve the accuracy of red tide development trend prediction. But training data is small in scale, the model convergence speed is low, and the precision is still to be improved.
At present, no prediction method is available, which can adapt to a large-scale data scene and has high precision, and therefore, a ocean red tide prediction model based on multi-source data and deep learning is designed. Red tide predictions are made using time series analysis of multi-source data. The method has high prediction accuracy, can be suitable for large-scale scenes, improves the prediction accuracy and improves the model complexity. Has practical significance and good application scene.
Disclosure of Invention
The invention aims to provide a ocean red tide prediction method based on multi-source data and deep learning, which adopts a Pearson coefficient to analyze the correlation between environmental factors and red tide occurrence, utilizes a complex correlation coefficient to analyze the correlation between multi-environmental factor combination and red tide occurrence, constructs a CNN-LSTM neural network model, trains the model by utilizing multi-environmental factors of different combinations, and constructs an ocean red tide prediction model based on multi-source data and deep learning.
In order to achieve the above purpose, the technical scheme of the invention is as follows: a ocean red tide prediction method based on multi-source data and deep learning comprises the following steps:
a. constructing a data set, integrating and preprocessing the data;
b. analyzing the correlation between the environmental factors and the red tide by adopting the Pearson coefficient, and analyzing the correlation between the multi-environmental factor combination and the red tide by utilizing the complex correlation coefficient;
c. constructing a CNN-LSTM prediction model, and training the CNN-LSTM prediction model by utilizing different combination multi-environment factors;
d. and testing the trained CNN-LSTM prediction model to obtain a prediction result, and performing performance evaluation.
In an embodiment of the present invention, the step a specifically includes:
a1. extracting remote sensing data, and inverting the surface temperature of the sea water and the chlorophyll concentration by adopting a window splitting algorithm formula (1) and a wave band ratio method formula (2) respectively;
T s ＝A 0 +A 1 T 31 -A 2 T 32 (1)
t in s Represents the surface temperature of seawater, T 31 、T 32 At 31 st and 32 nd band radiation brightness temperature A 0 、A 1 、A 2 Is a window splitting algorithm parameter;
Chl-a＝a*(B NIR B RED ) 2 +b*(B NIR B RED )+c(2)
wherein Chl-a represents chlorophyll a concentration, B NIR 、B RED For the near infrared band and the infrared band, a, b and c are parameters to be solved;
a2. processing the missing data of the monitoring data, and supplementing the missing data by adopting an interpolation method;
a3. integrating the remote sensing extracted data with the monitored environmental factor data of the time sequence, and performing normalization processing;
wherein x is original characteristic factor data, x min Is the minimum value of the feature factor data, x max And x' is the data after normalization processing and is the maximum value of the feature factor data.
In an embodiment of the present invention, the step b specifically includes:
b1. inputting processed environmental factor data including saturated dissolved oxygen, pH, chlorophyll a concentration, water temperature, salinity, turbidity, tide, wind speed u component, wind speed v component and air temperature;
b2. and (3) analyzing the correlation between each environmental factor and the occurrence of red tide by using the Pearson coefficient, wherein the calculation formula is as follows:
wherein X represents an environmental factor, Y represents the occurrence of red tide, cov (X, Y) is covariance between the two, and sigma X and sigma Y are standard deviations of X and Y respectively;
b3. and analyzing the correlation between the multi-environment characteristic factor combinations of different combinations and the occurrence of red tide by using complex correlation coefficients, wherein the calculation formula is as follows:
wherein y represents the occurrence of red tide, and y is the result obtained by regression of y on all environmental factors x.
In an embodiment of the present invention, the step c specifically includes:
c1. carrying out various different combinations on the processed characteristic environment factors;
c2. constructing a CNN-LSTM prediction model, wherein the CNN is used for mining local characteristics of environment variable data, the calculation mode is shown as a formula (6), the LSTM is used for mining time sequence dependent characteristics of an environment factor time sequence, and the calculation process is obtained by a series of formulas (7) - (11):
wherein M is j Is a set of the input maps that are to be mapped,for the output of the j-th group of data of layer l, < >>For the output of the ith data,weight for the ith data of the jth group,/->The additive bias is mapped for the j-th group of data at the first layer, and for different output maps, the input maps are convolved into different kernels;
f t ＝σ(W f *[h t-1 ,x t ]+b f )(7)
i t ＝σ(W i *[h t-1 ,x t ]+b i )(8)
C t ＝tanh*(W c *[h t-1 ,x t ]+b c )(9)
O t ＝σ(W o *[h t-1 ,x t ]+b o )(10)
h t ＝O t *tanh(C t )(11)
f t i is a forgetful door t For unit input gates, h t For the current cell output, C t Representing a memory cell, x t Representing element inputs, σ represents a sigmoid function, and tanh and σ are used as activation functions in the structure, W i 、W f 、W c 、W o Is a recursive weight matrix, b i 、b f 、b c 、b o Is a corresponding bias term;
c3. selecting an activation function and a loss function of a CNN-LSTM prediction model, and setting an optimizer super-parameter and a model super-parameter of the CNN-LSTM prediction model;
c4. fitting training is carried out on the environmental parameters, training data are confirmed, and repeated training optimization is carried out on the CNN-LSTM prediction model.
In an embodiment of the present invention, the step d specifically includes:
d1. performing performance evaluation on the trained CNN-LSTM prediction model, and verifying the stability and prediction accuracy of the model;
d2. and comparing and verifying the optimized CNN-LSTM prediction model by using different CNN-LSTM prediction models, different data sets corresponding to the same CNN-LSTM prediction model and multi-characteristic environmental factors of different combinations.
Compared with the prior art, the invention has the following beneficial effects: the invention discloses a ocean red tide prediction method based on multi-source data and deep learning. And adopting a deep learning method based on CNN-LSTM to predict red tide by utilizing multiple characteristic factors. The method combines remote sensing image data and water quality detection data, increases the data quantity, improves the complexity of a data set and the convergence speed of a model, and improves the accuracy of red tide forecasting.
Drawings
FIG. 1 is a general flow chart of an implementation of the present invention.
FIG. 2 is a flow chart of the inverse sea water surface temperature of the split window algorithm.
Detailed Description
The technical scheme of the invention is specifically described below with reference to the accompanying drawings.
Referring to fig. 1-2, one embodiment of the present invention includes the steps of:
a. constructing a data set, integrating and preprocessing the data;
b. the pearson coefficient is adopted to analyze the correlation between the environmental factors and the occurrence of the red tide, the environmental factors with relatively large correlation are reserved, and the complex correlation coefficient is utilized to analyze the correlation between the environmental factors and the occurrence of the red tide;
c. constructing a CNN-LSTM prediction model, and training the model by utilizing different combined multi-environment factors;
d. and testing the trained CNN-LSTM prediction model to obtain a prediction result, and performing performance evaluation.
Further, the step a specifically includes:
a1. the environment factor data of the red tide satellite remote sensing picture is extracted, the sea water surface temperature and chlorophyll concentration are inverted by mainly adopting the following formula, the sea water surface temperature is inverted by utilizing a window splitting algorithm (the formula is as follows (1)), and the chlorophyll concentration is inverted by utilizing a wave band ratio method (the formula is as follows (2)).
T s ＝A 0 +A 1 T 31 -A 2 T 32 (1)
T in s Represents the surface temperature of seawater, T 31 、T 32 At 31 st and 32 nd band radiation brightness temperature A 0 、A 1 、A 2 Is a window splitting algorithm parameter.
Chl-a＝a*(B NIR B RED ) 2 +b*(B NIR B RED )+c(2)
Wherein Chl-a represents chlorophyll a concentration, B NIR 、B RED For the near infrared band and the infrared band, a, b and c are parameters to be solved.
a2. The missing data in the water quality monitoring data is subjected to interpolation treatment, and the missing data is supplemented by mainly adopting a GAN-based time sequence interpolation method.
a3. Integrating the remote sensing extraction data with the monitored environmental factor data of the time sequence, and carrying out normalization processing.
Wherein x is original characteristic factor data, x min Is the minimum value of the feature factor data, x max And x' is the data after normalization processing and is the maximum value of the feature factor data.
Further, the step b specifically includes:
b1. inputting processed environmental factor data including saturated dissolved oxygen, pH, chlorophyll a concentration, water temperature, salinity, turbidity, tide, wind speed u component, wind speed v component, air temperature, etc.
b2. The pearson coefficient is utilized to analyze the correlation between each environmental factor including saturated dissolved oxygen, pH, chlorophyll a concentration, water temperature, salinity and the like and red tide occurrence, and the calculation formula is as follows:
wherein X represents an environmental factor, Y represents the occurrence of red tide, cov (X, Y) is covariance between the two, and σx and σy are standard deviations of X and Y, respectively.
b3. And analyzing the correlation between the multi-environment characteristic factor combinations of different combinations and the occurrence of red tide by using complex correlation coefficients, wherein the calculation formula is as follows:
wherein y represents the occurrence of red tide, and y is the result obtained by regression of y on all environmental factors x.
Further, the step c specifically includes:
c1. and (3) carrying out various combinations on the processed characteristic environment factors to be used as the input of a model, splitting a data set, and dividing the data set into a training set, a verification set and a test set, wherein the ratio of the training set to the verification set is 3:1:1.
c2. Constructing a CNN-LSTM network model, mining local features of various environmental variable data by using CNN, and finally flattening the extracted feature information to obtain feature input which can be used for the LSTM model, wherein the LSTM is responsible for calculating the dependence among various environmental factors of a time sequence, mining the time features of the data, and the calculation process is obtained by a series of formulas ((7) - (11)):
wherein M is j Is a set of the input maps that are to be mapped,for the output of the j-th group of data of layer l, < >>For the output of the ith data,weight for the ith data of the jth group,/->The additive bias is mapped for the j-th group of data at the first layer, and for different output maps, the input maps are convolved into different kernels;
f t ＝σ(W f *[h t-1 ,x t ]+b f )(7)
i t ＝σ(W i *[h t-1 ,x t ]+b i )(8)
C t ＝tanh*(W c *[h t-1 ,x t ]+b c )(9)
O t ＝σ(W o *[h t-1 ,x t ]+b o )(10)
h t ＝O t *tanh(C t )(11)
f t i is a forgetful door t For unit input gates, h t For the current cell output, C t Representing a memory cell, x t Representing element inputs, σ represents a sigmoid function, and tanh and σ are used as activation functions in the structure, W i 、W f 、W c 、W o Is a recursive weight matrix, b i 、b f 、b c 、b o Is the corresponding bias term.
c3. Determining an activation function and a loss function of a network model, setting an optimizer super-parameter and a model super-parameter of the network model, selecting a training model through a grid search method, determining the range of each parameter while ensuring the convergence of the model, and selecting.
c4. The final value of the network model super-parameters is determined by continuous selective training of the parameters.
Further, the step d specifically includes:
d1. and performing performance evaluation on the trained network model, verifying the stability and prediction accuracy of the model, and adopting root mean square deviation and average absolute error as indexes.
d2. And comparing and verifying the optimized network model by utilizing different network models, different data sets corresponding to the same model and multi-characteristic environment factors of different combinations.
The above is a preferred embodiment of the present invention, and all changes made according to the technical solution of the present invention belong to the protection scope of the present invention when the generated functional effects do not exceed the scope of the technical solution of the present invention.
Claims (5)
1. The ocean red tide prediction method based on multi-source data and deep learning is characterized by comprising the following steps of:
a. constructing a data set, integrating and preprocessing the data;
b. analyzing the correlation between the environmental factors and the red tide by adopting the Pearson coefficient, and analyzing the correlation between the multi-environmental factor combination and the red tide by utilizing the complex correlation coefficient;
c. constructing a CNN-LSTM prediction model, and training the CNN-LSTM prediction model by utilizing different combination multi-environment factors;
d. and testing the trained CNN-LSTM prediction model to obtain a prediction result, and performing performance evaluation.
2. The ocean red tide prediction method based on multi-source data and deep learning according to claim 1, wherein the step a specifically comprises:
a1. extracting remote sensing data, and inverting the surface temperature of the sea water and the chlorophyll concentration by adopting a window splitting algorithm formula (1) and a wave band ratio method formula (2) respectively;
T s ＝A 0 +A 1 T 31 -A 2 T 32 (1)
t in s Represents the surface temperature of seawater, T 31 、T 32 At 31 st and 32 nd band radiation brightness temperature A 0 、A 1 、A 2 Is a window splitting algorithm parameter;
Chl-a＝a*(B NIR B RED ) 2 +b*(B NIR B RED )+c(2)
wherein Chl-a represents chlorophyll a concentration, B NIR 、B RED For the near infrared band and the infrared band, a, b and c are parameters to be solved;
a2. processing the missing data of the monitoring data, and supplementing the missing data by adopting an interpolation method;
a3. integrating the remote sensing extracted data with the monitored environmental factor data of the time sequence, and performing normalization processing;
wherein x is original characteristic factor data, x min Is the minimum value of the feature factor data, x max And x' is the data after normalization processing and is the maximum value of the feature factor data.
3. The ocean red tide prediction method based on multi-source data and deep learning according to claim 1, wherein the step b specifically comprises:
b1. inputting processed environmental factor data including saturated dissolved oxygen, pH, chlorophyll a concentration, water temperature, salinity, turbidity, tide, wind speed u component, wind speed v component and air temperature;
b2. and (3) analyzing the correlation between each environmental factor and the occurrence of red tide by using the Pearson coefficient, wherein the calculation formula is as follows:
wherein X represents an environmental factor, Y represents the occurrence of red tide, cov (X, Y) is covariance between the two, and sigma X and sigma Y are standard deviations of X and Y respectively;
b3. and analyzing the correlation between the multi-environment characteristic factor combinations of different combinations and the occurrence of red tide by using complex correlation coefficients, wherein the calculation formula is as follows:
wherein y represents the occurrence of red tide, and y is the result obtained by regression of y on all environmental factors x.
4. The ocean red tide prediction method based on multi-source data and deep learning according to claim 1, wherein the step c specifically comprises:
c1. carrying out various different combinations on the processed characteristic environment factors;
c2. constructing a CNN-LSTM prediction model, wherein the CNN is used for mining local characteristics of environment variable data, the calculation mode is shown as a formula (6), the LSTM is used for mining time sequence dependent characteristics of an environment factor time sequence, and the calculation process is obtained by a series of formulas (7) - (11):
wherein M is j Is a set of the input maps that are to be mapped,for the output of the j-th group of data of layer l, < >>For the output of the ith data, +.>Weight for the ith data of the jth group,/->The additive bias is mapped for the j-th group of data at the first layer, and for different output maps, the input maps are convolved into different kernels;
f t ＝σ(W f *[h t-1 ,x t ]+b f )(7)
i t ＝σ(W i *[h t-1 ,x t ]+b i )(8)
C t ＝tanh*(W c *[h t-1 ,x t ]+b c )(9)
O t ＝σ(W o *[h t-1 ,x t ]+b o )(10)
h t ＝O t *tanh(C t )(11)
f t i is a forgetful door t For unit input gates, h t For the current cell output, C t Representing a memory cell, x t Representing element inputs, σ represents a sigmoid function, and tanh and σ are used as activation functions in the structure, W i 、W f 、W c 、W o Is a recursive weight matrix, b i 、b f 、b c 、b o Is a corresponding bias term;
c3. selecting an activation function and a loss function of a CNN-LSTM prediction model, and setting an optimizer super-parameter and a model super-parameter of the CNN-LSTM prediction model;
c4. fitting training is carried out on the environmental parameters, training data are confirmed, and repeated training optimization is carried out on the CNN-LSTM prediction model.
5. The ocean red tide prediction method based on multi-source data and deep learning according to claim 1, wherein the step d specifically comprises:
d1. performing performance evaluation on the trained CNN-LSTM prediction model, and verifying the stability and prediction accuracy of the model;
d2. and comparing and verifying the optimized CNN-LSTM prediction model by using different CNN-LSTM prediction models, different data sets corresponding to the same CNN-LSTM prediction model and multi-characteristic environmental factors of different combinations.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310739974.0A CN116611580A (en) | 2023-06-21 | 2023-06-21 | Ocean red tide prediction method based on multi-source data and deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310739974.0A CN116611580A (en) | 2023-06-21 | 2023-06-21 | Ocean red tide prediction method based on multi-source data and deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116611580A true CN116611580A (en) | 2023-08-18 |
Family
ID=87676602
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310739974.0A Pending CN116611580A (en) | 2023-06-21 | 2023-06-21 | Ocean red tide prediction method based on multi-source data and deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116611580A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117131365A (en) * | 2023-10-24 | 2023-11-28 | 自然资源部第二海洋研究所 | Red tide prediction method, system and medium based on sea pneumatic force field data |
-
2023
- 2023-06-21 CN CN202310739974.0A patent/CN116611580A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117131365A (en) * | 2023-10-24 | 2023-11-28 | 自然资源部第二海洋研究所 | Red tide prediction method, system and medium based on sea pneumatic force field data |
| CN117131365B (en) * | 2023-10-24 | 2024-02-13 | 自然资源部第二海洋研究所 | Red tide prediction method, system and medium based on sea pneumatic force field data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109272146B (en) | Flood prediction method based on deep learning model and BP neural network correction | |
| CN112149316A (en) | Aero-engine residual life prediction method based on improved CNN model | |
| CN110070226A (en) | Photovoltaic power prediction technique and system based on convolutional neural networks and meta learning | |
| CN112731309A (en) | Active interference identification method based on bilinear efficient neural network | |
| CN115099500B (en) | Water level prediction method based on weight correction and DRSN-LSTM model | |
| CN111242377A (en) | Short-term wind speed prediction method integrating deep learning and data denoising | |
| CN110738355A (en) | urban waterlogging prediction method based on neural network | |
| CN114169445A (en) | Day-ahead photovoltaic power prediction method, device and system based on CAE and GAN hybrid network | |
| CN112749663B (en) | Agricultural fruit maturity detection system based on Internet of things and CCNN model | |
| CN112557826A (en) | Ship electric power system fault diagnosis method | |
| CN116611580A (en) | Ocean red tide prediction method based on multi-source data and deep learning | |
| CN111242351A (en) | Tropical cyclone track prediction method based on self-encoder and GRU neural network | |
| CN114154619A (en) | Ship track prediction method based on CNN and BILSTM | |
| CN116229380A (en) | Method for identifying bird species related to bird-related faults of transformer substation | |
| Zhang et al. | Advance prediction of coastal groundwater levels with temporal convolutional and long short-term memory networks | |
| CN113642255A (en) | Photovoltaic power generation power prediction method based on multi-scale convolution cyclic neural network | |
| CN112507881A (en) | sEMG signal classification method and system based on time convolution neural network | |
| CN112508106A (en) | Underwater image classification method based on convolutional neural network | |
| CN115272776B (en) | Hyperspectral image classification method based on double-path convolution and double attention and storage medium | |
| CN111275025A (en) | Parking space detection method based on deep learning | |
| Feng et al. | Graph convolution based spatial-temporal attention lstm model for flood forecasting | |
| Wang et al. | Filling gaps in significant wave height time series records using bidirectional gated recurrent unit and cressman analysis | |
| CN115828758A (en) | Seawater three-dimensional prediction method and system based on improved firework algorithm optimization network | |
| Lu et al. | A deep belief network based model for urban haze prediction | |
| CN115630612A (en) | Software measurement defect data augmentation method based on VAE and WGAN |
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 |