CN111612274A - Tidal water level forecasting method based on space-time correlation - Google Patents

Tidal water level forecasting method based on space-time correlation Download PDF

Info

Publication number
CN111612274A
CN111612274A CN202010469480.1A CN202010469480A CN111612274A CN 111612274 A CN111612274 A CN 111612274A CN 202010469480 A CN202010469480 A CN 202010469480A CN 111612274 A CN111612274 A CN 111612274A
Authority
CN
China
Prior art keywords
water level
tidal
tide
level data
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.)
Granted
Application number
CN202010469480.1A
Other languages
Chinese (zh)
Other versions
CN111612274B (en
Inventor
朱贵强
胡勤友
梅强
杨春
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Maritime University
Original Assignee
Shanghai Maritime University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shanghai Maritime University filed Critical Shanghai Maritime University
Priority to CN202010469480.1A priority Critical patent/CN111612274B/en
Publication of CN111612274A publication Critical patent/CN111612274A/en
Application granted granted Critical
Publication of CN111612274B publication Critical patent/CN111612274B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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"
    • 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
    • G06N3/045—Combinations of networks
    • 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
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/067—Enterprise or organisation modelling
    • 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
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Abstract

A tidal water level forecasting method based on space-time correlation is characterized in that tidal water level data of a target port are predicted by utilizing tidal water level data of a relevant port which has correlation with tidal water level historical data of the target port to form a tidal water level data set of the target port, a convolutional neural network is utilized to extract spatial features from the tidal water level data, a gated cyclic unit neural network is utilized to extract time features from the tidal water level data, and a bidirectional gated cyclic unit neural network is utilized to train and verify the tidal water level data with the extracted outdated empty features to generate a short-time tidal water level forecasting model, so that the tidal water level of the target port is forecasted in a short time. The invention solves the problem that the historical data of the tidal water level of the port is less and the tidal water level can not be forecasted or is inaccurate, is convenient for the port and the piloting mechanism to reasonably arrange and dispatch, improves the production service efficiency and ensures the navigation safety.

Description

Tidal water level forecasting method based on space-time correlation
Technical Field
The invention relates to a tidal water level forecasting method based on space-time correlation.
Background
Accurate tidal forecast information plays an important role for coastal engineers, port authorities, maritime management institutions, shipping practitioners, marine accident rescue institutions, and the like. At present, a lot of methods for tidal forecast exist, but most of the methods are used for carrying out short-term tidal forecast on the current port based on a large amount of historical tidal data of the current port, little or no historical data of some minor ports are not considered, and the minor ports also need the short-term tidal forecast to provide references for activities such as production, life and the like.
Disclosure of Invention
The invention provides a tidal water level forecasting method based on space-time correlation, which solves the problem that the tidal water level forecasting cannot be carried out or is inaccurate due to less historical data of the tidal water level of a port, is convenient for reasonably arranging and dispatching the tidal water level of the port and a piloting mechanism, improves the production service efficiency and ensures the sailing safety.
In order to achieve the above object, the present invention provides a tidal water level forecasting method based on space-time correlation, comprising the following steps:
predicting tide water level data of the target port by using tide water level data of a relevant port having correlation with the tide water level historical data of the target port to form a tide water level data set of the target port;
the method comprises the steps of extracting spatial features from tide water level data by using a convolutional neural network, extracting time features from the tide water level data by using a gated circulation unit neural network, and training and verifying the tide water level data with the extracted outdated empty features by using a bidirectional gated circulation unit neural network to generate a short-time tide water level forecasting model so as to realize short-time forecasting of the tide water level of a target port.
The method for forming the tidal level data set of the target port comprises the following steps:
step S1, obtaining tide water level data;
step S2, supplementing missing tide water level data;
step S3, calculating Kendall correlation coefficients of tide level data of the target port and the surrounding ports;
step S4, formulating data;
the method comprises the steps of taking a target port as a prediction point, taking tide water level data of the target port as a dependent variable, determining the tide water level data of the port with a Kendell correlation coefficient of the tide water level data of the target port being more than or equal to 0.8 as an independent variable, mapping the tide water level data as the independent variable and the dependent variable onto a one-dimensional vector, and expressing the one-dimensional spatial information vector at the same moment as:
Xs＝(X1，X2，…，Xs)
combining the one-dimensional spatial information vectors at different time points into a matrix as follows:
Figure BDA0002513813490000021
wherein s is different ports, and t is time;
step S5, data normalization;
the data set X formulated in step S4 was normalized using the pandas library and numpy library in python:
Figure BDA0002513813490000022
wherein, the tide level time sequence is represented by X, X' represents the number within (-1, 1) after X is normalized, min (X) represents the minimum value in the tide level data, and max (X) represents the maximum value in the tide level data.
And (4) supplementing missing tide water level data by adopting an interpolation method.
Dividing the data in the tide level data set of the target port, wherein the first 80% is used as a training set, the second 20% is used as a verification set, and the 10% is used as a test set.
The method for generating the forecast model of the short-term tidal water level comprises the following steps:
s1, extracting the spatial features of the tide water level data by using a Convolutional Neural Network (CNN);
step S2, time feature extraction is carried out on the tide water level data with the extracted space features by using a gate control circulation unit GRU neural network;
and step S3, training and verifying the tide water level data with the extracted spatial characteristics and temporal characteristics by using a Bi-directional gating circulation unit Bi-GRU neural network, and establishing a short-time tide forecasting model to realize end-to-end prediction output.
The spatial feature extraction method comprises the following steps:
filling operation;
performing convolution operation;
G(i)＝F(Aw+B)
w is a filtering weight of a node, B is a deviation, A is a value of an input node, G (i) is spatial feature data after convolution, F is an activation function, and a ReLu activation function is used;
and (4) performing a pooling operation.
Using mean square error MSE as the loss function of the short-time tide forecasting model during training;
Figure BDA0002513813490000031
wherein, ytIs the observed value of tidal water level,
Figure BDA0002513813490000032
Is the predicted value of tidal water level.
Performing depth optimization neural network during the short-time tide forecasting model training by using Adam algorithm;
m＝β1·m+(1-β1)·dx
Figure BDA0002513813490000033
Figure BDA0002513813490000034
wherein, β1And β2Is two constants, x is the updated parameter, dxIs the derivative vector of x, m is used to store the first order matrix, and v is used to store the second order matrix.
After each training, verifying the verification set by using the short-term tide forecasting model and calculating the precision of the short-term tide forecasting model, and when the precision is not lower than 97 percent or the training frequency reaches 2000 times, stopping the training and storing the short-term tide forecasting model;
the accuracy calculation formula is as follows:
Figure BDA0002513813490000035
wherein, ytIs the observed value of tidal water level,
Figure BDA0002513813490000036
Is the predicted value of tidal water level.
Testing the short-time tide forecasting model by using a test set, and evaluating the short-time tide forecasting model by using a Mean Absolute Error (MAE), a Mean Absolute Percentage Error (MAPE), a Root Mean Square Error (RMSE) and a Correlation Coefficient (CC);
Figure BDA0002513813490000041
Figure BDA0002513813490000042
Figure BDA0002513813490000043
Figure BDA0002513813490000044
wherein, ytIs the observed value of tidal water level,
Figure BDA0002513813490000045
Is a tidal water level predicted value,
Figure BDA0002513813490000046
Is the average value of the observed tidal water level,
Figure BDA0002513813490000047
is the predicted average of tidal water levels.
The present invention also provides a processing apparatus comprising: a processor adapted to execute various program codes; and a data storage device adapted to store a plurality of program codes; the program code is adapted to be loaded and executed by a processor to implement the method for tidal level forecasting based on spatiotemporal correlations.
According to the tidal water level prediction method and device, the tidal water level data of the target port can be predicted in a space-time mode by utilizing the tidal water level data of the port with high correlation with the historical tidal water level data of the target port, the problem that the tidal water level prediction cannot be carried out or is inaccurate due to the fact that the historical tidal water level data of the port is few is solved, the port and a piloting mechanism can conveniently arrange and schedule reasonably, the production service efficiency is improved, and the sailing safety is guaranteed.
Drawings
FIG. 1 is a schematic diagram of a tidal water level forecasting method based on space-time correlation provided by the invention.
FIG. 2 is a flow chart of the formation of a tidal level data set for a target port.
FIG. 3 is a schematic diagram of generating a forecast model of short tidal water level according to an embodiment of the present invention.
Fig. 4 is a schematic diagram of the positions of a target port and its associated ports in an embodiment of the present invention.
Detailed Description
The preferred embodiment of the present invention will be described in detail below with reference to fig. 1 to 4.
As shown in FIG. 1, the present invention provides a tidal water level forecasting method based on space-time correlation, comprising the following steps:
the tide water level data preprocessing module predicts tide water level data of a target port by utilizing tide water level data of a relevant port with higher relevance to tide water level historical data of the target port to form a tide water level data set of the target port;
the tide water level forecasting module utilizes the convolutional neural network to extract spatial characteristics from the tide water level data, applies the gate control cycle unit neural network to extract time characteristics from the tide water level data with the extracted spatial characteristics to generate tide water level data with obvious space-time characteristics, and applies the bidirectional gate control cycle unit neural network to learn and train the tide water level data with the extracted space-time characteristics to generate a short-time tide water level forecasting model so as to realize short-time forecasting of the tide water level of the target port.
As shown in fig. 2, the method for forming the tidal level data set of the target port specifically includes:
and step S1, acquiring tide water level data.
And (3) obtaining the integral tide water level data of the target port and ports around the target port by using a tide water level data recorder sensor.
And step S2, filling up the missing tide water level data.
And (4) carrying out interpolation and completion on the missing integral point tide water level data by using an interpolation method.
And step S3, calculating a correlation coefficient.
And calculating Kendall (Kendall) correlation coefficients of the tidal level data of the target port and the whole point of the port around the target port.
And step S4, formulating data.
The method comprises the steps of taking a target port as a prediction point, taking tidal water level data of the target port as a dependent variable, determining the tidal water level data of the port with high correlation (the correlation coefficient is more than or equal to 0.8) with the tidal water level data of the target port as an independent variable, mapping the tidal water level data as the independent variable and the dependent variable onto a one-dimensional vector, and expressing the one-dimensional spatial information vector at the same moment as follows:
Xs＝(X1，X2，…，Xs)
combining the one-dimensional spatial information vectors at different time points into a matrix as follows:
Figure BDA0002513813490000061
wherein s is different ports and t is time.
And step S5, normalizing the data.
The data set X formulated in step S4 was normalized using the pandas library and numpy library in python, and the calculation method was as follows:
Figure BDA0002513813490000062
wherein, the tide level time sequence is represented by X, X' represents the number within (-1, 1) after X is normalized, min (X) represents the minimum value in the tide level data, and max (X) represents the maximum value in the tide level data.
And step S6, dividing the data set.
The first 80% of the data set X' obtained in step S5 is used as a training set, the second 20% is 10% as a verification set, and the 10% is used as a test set.
As shown in fig. 1, the method for generating a model for forecasting tidal water level in short time specifically includes the following steps:
and step S1, spatial feature extraction.
And (3) performing spatial feature extraction on the tide water level data by using a Convolutional Neural Network (CNN).
And step S2, extracting time characteristics.
And (3) applying a Gate controlled circulation Unit (GRU) neural network to extract the time characteristics of the tide water level data with the extracted spatial characteristics.
And step S3, training a model.
And (3) learning and training the tide water level data with the extracted spatial characteristics and temporal characteristics by using a bidirectional gating circulation Unit (Bi-GRU) neural network, and establishing a short tide forecasting model to realize end-to-end prediction output.
In an embodiment of the present invention, the tidal water level forecast is performed for the upper-sea residual mountain (31 ° 25.5 'N, 122 ° 14.1' E), and the following steps are performed:
step 1, obtaining tide water level data.
As shown in fig. 4, tidal water level data recorder sensors were used to obtain the tidal water level data at the whole points of shanghai residual (31 ° 25.5 'N, 122 ° 14.1' E), and the surrounding reed harbors (30 ° 50.0 'N, 121 ° 50.0' E), zhongtong (31 ° 6.8 'N, 121 ° 54.2' E), rime (31 ° 23.5 'N, 121 ° 30.5' E), jinshan (30 ° 45.0 'N, 121 ° 22.0' E), huangpu (31 ° 14.0 'N, 121 ° 29.0' E), gao (31 ° 19.8 'N, 121 ° 33.5' E), chongming (31 ° 32.0 'N, 121 ° 38.0' E) from 0 at 1 month 1 year 2020 to 23 year 2 month 29 days 23 years, respectively.
And 2, filling up missing tide water level data.
And (4) carrying out interpolation and completion on the missing integral point tide water level data by using a cubic spline interpolation method.
And 3, calculating a correlation coefficient.
The Kendall (Kendall) correlation coefficients of the tidal level data of sanshan and surrounding reed harbor, Zhongtong, Erime, Jinshan, Huangpu, Gao bridge and Chongming whole point are calculated to be 0.826, 0.814, 0.278, 0.488, 0.089, 0.38 and 0.341 respectively, and the correlation of the tidal level data of Zhongtang, Luguang and sanshan is the highest.
And 4, formulating data.
The whole tide level data of the Chinese dredging, the reed harbor and the sanshan are formulated, namely:
Figure BDA0002513813490000071
wherein s is1Representative of medium dredging, s2Represents Luchao harbor, s3Representing a sanshan, t is time.
And 5, normalizing the data.
The data set formulated in step 4 was normalized using the pandas library and numpy library in python, as follows:
Figure BDA0002513813490000072
wherein, the tide level time sequence is represented by X, X' represents the number within (-1, 1) after X is normalized, min (X) represents the minimum value in the tide level data, and max (X) represents the maximum value in the tide level data.
And 6, dividing the data set.
And (5) taking the first 80% of the data set in the step 5 as a training set, taking the last 20% as a verification set, and taking 10% as a test set.
And step 7, padding operation.
As shown in fig. 3, in order to be able to put the convolutional neural network in a matrix form, a zero dimension is added, that is, a padding operation in the convolutional neural network is performed, and the original data is changed into a small matrix of 3 × 3.
And 8, performing convolution operation.
As shown in fig. 3, since it is a simple time series, each element in the matrix is processed using one-dimensional convolution, the convolution kernel size is 2 × 2, and the step size is 1. And a one-dimensional convolution kernel filter is used as a convolution layer, convolution information of a local perception domain is obtained through a sliding filter, unit nodes generated in each step slide on vectors, and local features are aggregated into global features. After convolution the data becomes a 2 x 2 matrix.
G(i)＝F(Aw+B)
Wherein w is a filtering weight of a node, B is a deviation, A is a value of an input node, G (i) is spatial feature data after convolution, and F is an activation function which uses a ReLu activation function.
And 9, performing pooling operation.
As shown in fig. 3, the pooling filter is applied to the spatial feature data g (i) after convolution operation as a pooling layer, effectively reducing the size of the feature map to reduce the parameter and learning burden, filtering out some unnecessary spatial feature information during pooling to obtain more abstract tidal level data spatial features, in order not to miss important information, using mean pooling, the pooling kernel size is 2 × 2,step length is 1, the characteristic sequence G (i) generated in the previous step is reduced to half of the original size through mean pooling, namely the characteristic sequence G (i) is changed into 1 × 1, and a data vector after convolution and pooling is represented as Ct＝(C1，C2，…，Ct)。
And step 10, extracting space-time characteristics.
As shown in FIG. 3, the convolved and pooled tidal water level data C with obvious spatial characteristicstInputting the time characteristic into a GRU module for time characteristic extraction to generate tide water level data CG with obvious space-time characteristicst＝(CG1，CG2，…，CGt) This is taken as the input of the Bi-GRU module.
And step 11, training and verifying the model.
The model is trained and validated using a training set and a validation set, as shown in FIG. 3. The mean square error MSE (mean Squared error) is used as the loss function when the short-term tidal forecast model is trained, which can accurately describe the difference between the true value and the predicted value.
The MSE expression is as follows:
Figure BDA0002513813490000091
wherein, ytIs the observed value of tidal water level,
Figure BDA0002513813490000092
Is the predicted value of tidal water level.
And carrying out deep optimization on the neural network during training by using an Adam algorithm. The Adam algorithm is an update of the RMSProp algorithm by integrating momentum, is an optimizer based on first-order gradients, has very high computational efficiency, and requires little memory.
The core part of the Adam algorithm is as follows:
m＝β1·m+(1-β1)·dx
Figure BDA0002513813490000093
Figure BDA0002513813490000094
wherein, β1And β2Is two constants, x is the updated parameter, dxIs the derivative vector of x, m is used to store the first order matrix, and v is used to store the second order matrix.
After each training, the short-term tide forecasting model is used for verifying the verification set and calculating the precision of the short-term tide forecasting model, and when the precision is not lower than 97 percent or the training times reach 2000 times, the training is stopped and the model is saved. The accuracy calculation formula is as follows:
Figure BDA0002513813490000095
wherein, ytIs the observed value of tidal water level,
Figure BDA0002513813490000096
Is the predicted value of tidal water level.
And step 12, testing the tidal forecasting method based on the space-time correlation.
And (4) testing the tide forecasting method based on the space-time correlation by using the short-time tide forecasting model saved in the step (11).
The tide forecasting method based on the space-time correlation is evaluated by adopting an average absolute error MAE, an average absolute percentage error MAPE, a root mean square error RMSE and a correlation coefficient CC, and the specific calculation formula is as follows:
Figure BDA0002513813490000097
Figure BDA0002513813490000101
Figure BDA0002513813490000102
Figure BDA0002513813490000103
wherein, ytIs the observed value of tidal water level,
Figure BDA0002513813490000104
Is a tidal water level predicted value,
Figure BDA0002513813490000105
Is the average value of the observed tidal water level,
Figure BDA0002513813490000106
is the predicted average of tidal water levels.
By using the tidal prediction method based on the space-time correlation, which is disclosed by the invention, experiments are carried out by taking the above-sea hedge as an example, and the experimental result shows that the average absolute error MAE is as follows: 0.5602cm, mean absolute percent error MAPE: 0.32%, root mean square error RMSE: 0.7879cm, and a correlation coefficient CC of 99.99%.
According to the tidal water level prediction method and device, the tidal water level data of the target port can be predicted in a space-time mode by utilizing the tidal water level data of the port with high correlation with the historical tidal water level data of the target port, the problem that the tidal water level prediction cannot be carried out or is inaccurate due to the fact that the historical tidal water level data of the port is few is solved, the port and a piloting mechanism can conveniently arrange and schedule reasonably, the production service efficiency is improved, and the sailing safety is guaranteed.
While the present invention has been described in detail with reference to the preferred embodiments, it should be understood that the above description should not be taken as limiting the invention. Various modifications and alterations to this invention will become apparent to those skilled in the art upon reading the foregoing description. Accordingly, the scope of the invention should be determined from the following claims.

Claims (10)

1. A tidal water level forecasting method based on space-time correlation is characterized by comprising the following steps:
predicting tide water level data of the target port by using tide water level data of a relevant port having correlation with the tide water level historical data of the target port to form a tide water level data set of the target port;
the method comprises the steps of extracting spatial features from tide water level data by using a convolutional neural network, extracting time features from the tide water level data by using a gated circulation unit neural network, and training and verifying the tide water level data with the extracted outdated empty features by using a bidirectional gated circulation unit neural network to generate a short-time tide water level forecasting model so as to realize short-time forecasting of the tide water level of a target port.
2. The tidal level forecasting method based on space-time correlation as claimed in claim 1, wherein the method for forming the tidal level data set of the target port comprises:
step S1, obtaining tide water level data;
step S2, supplementing missing tide water level data;
step S3, calculating Kendall correlation coefficients of tide level data of the target port and the surrounding ports;
step S4, formulating data;
the method comprises the steps of taking a target port as a prediction point, taking tide water level data of the target port as a dependent variable, determining the tide water level data of the port with a Kendell correlation coefficient of the tide water level data of the target port being more than or equal to 0.8 as an independent variable, mapping the tide water level data as the independent variable and the dependent variable onto a one-dimensional vector, and expressing the one-dimensional spatial information vector at the same moment as:
Xs＝(X1，X2，…，Xs)
combining the one-dimensional spatial information vectors at different time points into a matrix as follows:
Figure FDA0002513813480000011
wherein s is different ports, and t is time;
step S5, data normalization;
the data set X formulated in step S4 was normalized using the pandas library and numpy library in python:
Figure FDA0002513813480000021
wherein, the tide level time sequence is represented by X, X' represents the number within (-1, 1) after X is normalized, min (X) represents the minimum value in the tide level data, and max (X) represents the maximum value in the tide level data.
3. The tidal level forecasting method based on space-time correlation according to claim 2, characterized in that the data in the tidal level data set of the target port are divided, the first 80% is taken as a training set, the last 20% is taken as a verification set, and the 10% is taken as a test set.
4. The method for tidal level forecast based on spatiotemporal correlation as claimed in claim 3, wherein said method for generating a model of short tidal level forecast comprises the steps of:
s1, extracting the spatial features of the tide water level data by using a Convolutional Neural Network (CNN);
step S2, time feature extraction is carried out on the tide water level data with the extracted space features by using a gate control circulation unit GRU neural network;
and step S3, training and verifying the tide water level data with the extracted spatial characteristics and temporal characteristics by using a Bi-directional gating circulation unit Bi-GRU neural network, and establishing a short-time tide forecasting model to realize end-to-end prediction output.
5. The tidal water level forecasting method based on space-time correlation as claimed in claim 4, wherein the spatial feature extraction method comprises:
filling operation;
performing convolution operation;
G(i)＝F(Aw+B)
w is a filtering weight of a node, B is a deviation, A is a value of an input node, G (i) is spatial feature data after convolution, F is an activation function, and a ReLu activation function is used;
and (4) performing a pooling operation.
6. The method of space-time correlation based tidal water level forecast of claim 5, wherein Mean Square Error (MSE) is used as a loss function when training said short-time tidal forecast model;
Figure FDA0002513813480000022
wherein, ytIs the observed value of tidal water level,
Figure FDA0002513813480000023
Is the predicted value of tidal water level.
7. The method of space-time correlation based tidal water level forecast of claim 6, wherein Adam's algorithm is used to perform depth-optimized neural network during the training of said short-time tidal forecast model;
m＝β1·m+(1-β1)·dx
Figure FDA0002513813480000031
Figure FDA0002513813480000032
wherein, β1And β2Is two constants, x is the updated parameter, dxIs the derivative vector of x, m is used to store the first order matrix, and v is used to store the second order matrix.
8. The method for tidal water level forecast based on spatio-temporal correlation as claimed in claim 7, wherein after each training, said short tidal forecast model is applied to verify the verification set and calculate the precision of the short tidal forecast model, when the precision is not less than 97%, or the number of training times reaches 2000, the training is stopped and the short tidal forecast model is saved;
the accuracy calculation formula is as follows:
Figure FDA0002513813480000033
wherein, ytIs the observed value of tidal water level,
Figure FDA0002513813480000034
Is the predicted value of tidal water level.
9. The method of claim 8, wherein the short term tidal forecast model is tested using a test set, and the short term tidal forecast model is evaluated using a Mean Absolute Error (MAE), a mean percent absolute error (MAPE), a Root Mean Square Error (RMSE), and a Correlation Coefficient (CC);
Figure FDA0002513813480000035
Figure FDA0002513813480000036
Figure FDA0002513813480000037
Figure FDA0002513813480000041
wherein, ytIs the observed value of tidal water level,
Figure FDA0002513813480000042
Is a tidal water level predicted value,
Figure FDA0002513813480000043
Is the average value of the observed tidal water level,
Figure FDA0002513813480000044
is the predicted average of tidal water levels.
10. A processing apparatus, comprising: a processor adapted to execute various program codes; and a data storage device adapted to store a plurality of program codes; wherein the program code is adapted to be loaded and executed by a processor to implement the method of tidal level forecast based on spatiotemporal correlation of any of claims 1-9.
CN202010469480.1A 2020-05-28 2020-05-28 Tidal water level forecasting method based on space-time correlation Active CN111612274B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010469480.1A CN111612274B (en) 2020-05-28 2020-05-28 Tidal water level forecasting method based on space-time correlation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010469480.1A CN111612274B (en) 2020-05-28 2020-05-28 Tidal water level forecasting method based on space-time correlation

Publications (2)

Publication Number Publication Date
CN111612274A true CN111612274A (en) 2020-09-01
CN111612274B CN111612274B (en) 2023-09-12

Family

ID=72196668

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010469480.1A Active CN111612274B (en) 2020-05-28 2020-05-28 Tidal water level forecasting method based on space-time correlation

Country Status (1)

Country Link
CN (1) CN111612274B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113077110A (en) * 2021-04-21 2021-07-06 国家海洋信息中心 GRU-based harmonic residual segmented tide level prediction method
CN113506000A (en) * 2021-07-15 2021-10-15 上海海事大学 Method for calculating rainstorm overflow risk of urban catchment area
CN116542392A (en) * 2023-05-24 2023-08-04 珠江水利委员会珠江水利科学研究院 Real-time forecasting method and system for tidal range water level and readable storage medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102214262A (en) * 2010-04-02 2011-10-12 上海海洋大学 Tide predicting method
CN104376230A (en) * 2014-12-03 2015-02-25 大连海事大学 Tidal prediction method
CN105956709A (en) * 2016-05-12 2016-09-21 大连海事大学 GUI based modular support vector machine tide forecasting method
KR101754030B1 (en) * 2016-06-09 2017-07-05 군산대학교산학협력단 Tide Level Data Processing Method And Device Of Thereof
CN108694482A (en) * 2018-07-27 2018-10-23 西南石油大学 Based on fractal theory and improved least square method supporting vector machine tide flow velocity prediction technique

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102214262A (en) * 2010-04-02 2011-10-12 上海海洋大学 Tide predicting method
CN104376230A (en) * 2014-12-03 2015-02-25 大连海事大学 Tidal prediction method
CN105956709A (en) * 2016-05-12 2016-09-21 大连海事大学 GUI based modular support vector machine tide forecasting method
KR101754030B1 (en) * 2016-06-09 2017-07-05 군산대학교산학협력단 Tide Level Data Processing Method And Device Of Thereof
CN108694482A (en) * 2018-07-27 2018-10-23 西南石油大学 Based on fractal theory and improved least square method supporting vector machine tide flow velocity prediction technique

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
ANIBAL FLORES;HUGO TITO;DEYMOR CENTTY: "Improving Gated Recurrent Unit Predictions with Univariate Time Series Imputation Techniques" *
张泽国;尹建川;柳成;张心光;: "基于SAPSO-BP网络模型的港口潮汐实时预报", no. 01 *
李佳文;王景升: "基于深度学习的短时交通流预测" *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113077110A (en) * 2021-04-21 2021-07-06 国家海洋信息中心 GRU-based harmonic residual segmented tide level prediction method
CN113506000A (en) * 2021-07-15 2021-10-15 上海海事大学 Method for calculating rainstorm overflow risk of urban catchment area
CN113506000B (en) * 2021-07-15 2024-04-02 上海海事大学 Method for calculating storm overflow risk of urban water collecting area
CN116542392A (en) * 2023-05-24 2023-08-04 珠江水利委员会珠江水利科学研究院 Real-time forecasting method and system for tidal range water level and readable storage medium
CN116542392B (en) * 2023-05-24 2024-02-13 珠江水利委员会珠江水利科学研究院 Real-time forecasting method and system for tidal range water level and readable storage medium

Also Published As

Publication number Publication date
CN111612274B (en) 2023-09-12

Similar Documents

Publication Publication Date Title
US11334407B2 (en) Abnormality detection system, abnormality detection method, abnormality detection program, and method for generating learned model
CN109978228B (en) PM2.5 concentration prediction method, device and medium
CN111612274A (en) Tidal water level forecasting method based on space-time correlation
US20150254554A1 (en) Information processing device and learning method
CN107153874A (en) Water quality prediction method and system
KR101938829B1 (en) Method and apparatus for learning neural network to predict occurrence of sea fog
CN112712209B (en) Reservoir warehousing flow prediction method and device, computer equipment and storage medium
Braakmann-Folgmann et al. Sea level anomaly prediction using recurrent neural networks
CN111199270A (en) Regional wave height forecasting method and terminal based on deep learning
US11487266B2 (en) Method for recognizing contingencies in a power supply network
CN112668238B (en) Rainfall processing method, rainfall processing device, rainfall processing equipment and storage medium
CN111091196A (en) Passenger flow data determination method and device, computer equipment and storage medium
CN111242351A (en) Tropical cyclone track prediction method based on self-encoder and GRU neural network
CN112084864A (en) Model optimization method and device, electronic equipment and storage medium
CN116542389A (en) Truck fuel consumption prediction method, system, equipment and storage medium
CN113253336B (en) Earthquake prediction method and system based on deep learning
CN117520784A (en) Groundwater level multi-step prediction method based on convolution attention long-short-term neural network
CN112966815A (en) Target detection method, system and equipment based on impulse neural network
CN110852415B (en) Vegetation index prediction method, system and equipment based on neural network algorithm
CN115545334B (en) Land utilization type prediction method and device, electronic equipment and storage medium
CN115618714B (en) Solar radiation probability forecasting method and related equipment
US20220327335A1 (en) Controlling asynchronous fusion of spatio-temporal multimodal data
KR102388292B1 (en) Radiation parameterization system for correcting uncertainty of radiation parameterization emulator and operating method thereof
KR20200092469A (en) Apparatus and method for providing fashion coordination knowledge based on a neural network having explicit memory
Lønnum Deep Learning Metocean Simulation and its Applications in Marine Simulation-based Design

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
GR01 Patent grant
GR01 Patent grant