CN112906955B - Tidal data fitting method based on multi-source data driving - Google Patents
Tidal data fitting method based on multi-source data driving Download PDFInfo
- Publication number
- CN112906955B CN112906955B CN202110159143.7A CN202110159143A CN112906955B CN 112906955 B CN112906955 B CN 112906955B CN 202110159143 A CN202110159143 A CN 202110159143A CN 112906955 B CN112906955 B CN 112906955B
- Authority
- CN
- China
- Prior art keywords
- data
- tide
- matrix
- input
- neural network
- 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.)
- Active
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
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C13/00—Surveying specially adapted to open water, e.g. sea, lake, river or canal
- G01C13/002—Measuring the movement of open water
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/02—Instruments for indicating weather conditions by measuring two or more variables, e.g. humidity, pressure, temperature, cloud cover or wind speed
-
- 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
-
- 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
Abstract
The invention discloses a target point tidal data fitting method based on multi-source data driving, which comprises the following steps: collecting multi-source data; performing relevance sorting and information screening on the multi-source data; establishing a fitting model based on a radial basis function neural network; and carrying out network dynamic adjustment based on the sliding window. The invention utilizes the actually measured tide information and the hydrometeorological information of nearby point positions, the information fully reflects the influence of the hydrometeorological factors on the tide, and the radial basis function neural network based on partial least squares regression can realize accurate high-dimensional nonlinear fitting and can obtain more accurate tide forecast. Under the condition that certain sea area point locations are not suitable for arranging permanent tide observation stations, the invention provides accurate tide fitting results for the point locations by using tide observation point data near the sea area. Under the condition that the tide information of the tide observation point location is lost, the accurate fitting data of the observation point location can be obtained through the tide data of the nearby point location.
Description
Technical Field
The invention relates to a tide forecasting method, in particular to a tide data fitting method based on multi-source data driving.
Background
Tidal phenomena refer to the periodic movement of seawater under the tidal forces of celestial bodies (mainly the moon and the sun). The tidal phenomenon is caused by the tidal force caused by the fluctuation of seawater due to the combined action of the gravitational force of the moon and the centrifugal force. The prime movers of the earth tide, the sea tide and the gas tide are caused by different gravitations of the earth in the day and the month, and the three have mutual influence. Because the moon is closer to the earth than the sun, the ratio of the moon to the solar tidal power is 11: and 5, for the ocean, Taiyin tide is more obvious than solar tide. Elastic-plastic tidal deformation of the crust at the bottom of the ocean can cause corresponding sea tides, i.e. for sea tides, the influence of the ground tide effect exists; the migration of sea water quality caused by sea tide changes the load born by the earth crust, so that the earth crust can be repeatedly bent. The tide is above the tide and acts on the sea surface to cause additional vibration, which further complicates the tidal change.
The tide generation is related to the sun and the moon, and also corresponds to the traditional lunar calendar in China. At the first time of the lunar calendar, namely the lunar, the sun and the moon are on one side of the earth, so that the maximum tidal attraction force is obtained, so that the 'big tide' is caused; when the lunar phase is the upper chord and the lower chord, namely the first eight and twenty-three times of the lunar calendar, the solar induced tide force and the lunar induced tide force mutually offset a part, so that the 'small tide' occurs, so that the situation that the quan has the saying that the first fifteen big tides and the first eighty three places see the beach is provided in the rural adage. In addition, the rise tide occurs every day, and the rise tide occurs every day, because the moon moves by 13 degrees upwards on the celestial sphere every day, and the total time is about 50 minutes, namely, the time of the day before the moon and the day after the moon (1 taiyin day is 50 minutes compared with 24 hours) is delayed by about 50 minutes (the tide occurs in the day after the moon and the tide generally occurs twice every day), the rise tide time is also delayed by about 50 minutes every day. A plurality of tide computing methods (tide pushing moments) such as an eighth tide computing method, which are obtained by summarizing the labor people in China over a hundred years are one example of the tide computing methods, and a concise formula is as follows:
high tide time is 0.8h x [ lunar calendar date-1 (or 16) ] + high tide gap.
From the above formula, one high tide time in the day can be calculated, and for the normal semi-solar tide sea area, the value is added or subtracted by 12 minutes and 25 minutes (or added or subtracted by 12 minutes and 24 minutes for convenience of calculation), so that the other high tide time can be calculated. When the value is added or subtracted by 6 minutes, 12 minutes can obtain the time of low tide, namely the time of low tide. However, due to the complexity of the movement of the moon and the sun, a big tide may be delayed for one or several days sometimes, a climax between taiyin days often lags behind the time of the moon in the upper or middle day or the lower middle day for one or several hours, and tides occur once in a taiyin day in some places. Therefore, the rising tide and the falling tide are different in time and interval every day.
Since the tide is affected by many factors, periodic factors such as tidal forces, and non-periodic factors such as wind, air pressure, coastal characteristics, precipitation, inclination of the lunar orbit, etc. The traditional tidal forecast method is a harmonic analysis method. According to the method, the parameters of each tide in the model are calculated through statistics and analysis of long-term tide data of the tide station, and long-term tide forecast is obtained based on the established tide mathematical model.
The forecasting precision of the traditional harmonic analysis method is not only influenced by the number of the tide divisions, but also cannot be analyzed by the influence of non-periodic factors, so that the forecasting precision is low.
At present, three methods are generally adopted for actually reporting and forecasting tides at a target point, wherein one method is to directly set a measuring point at the target point to actually report tides and forecast the tides based on actually reported data; secondly, based on the long-term tidal observation data of the point location, performing data analysis by using a harmonic analysis method, and accordingly giving a tidal forecast of the point location; and thirdly, carrying out tide forecast on the target point by using the tide data of other points according to the correlation between the tide data of other points and the tide data of the target point.
The following problems exist in the tidal forecast using the above three methods:
(1) the first method has a problem in that many empty areas of tidal data exist in the current coastal sea area. The tide information in the blank area is complemented, which cannot be realized only by actual observation. Because the blank area is large, huge manpower and material resources are required to be invested, and the systematic real-time tide actual prediction and tide prediction cannot be carried out at each required point position.
In addition, in some areas, such as areas with busy sailing, the actual prediction and forecast of tide need to be carried out in the areas due to sailing safety reasons, but because the sailing is busy, the system and long-term tide observation cannot be implemented, and only temporary tide observation can be carried out in a channel design stage or in a time period with low channel navigation density.
(2) The second method, when using tidal analysis, requires long-term actual tidal data to determine the parameters of each partial tide due to the large number of partial tides. However, as described above, the long-term and systematic tidal observation requires enormous manpower and material resources, and it is impossible to perform the long-term tidal observation at every desired site.
Moreover, the model of the method using the harmonic coefficient cannot reflect non-periodic time-varying factors, such as the influence of the hydrometeorology factors on the tide, so that the forecasting precision of the method on the tide is limited, and especially the forecasting error is large under the condition that the hydrometeorology factors are severely changed.
(3) In the third method, the tide at the target point is estimated according to the forecast data of the tide at other tide points, and the current example of the method is the estimation of the attached port tide in the tide table of Chinese English edition. The method has the main problems that the calculation precision is low, and the tide situation of the main harbor according to the method is mainly carried out according to a harmonic analysis method, so that the influence of factors such as hydrometeorology and the like on the tide can not be reflected.
Disclosure of Invention
In order to solve the problems in the prior art, the invention aims to design a tidal data fitting method based on multi-source data driving, which has high prediction precision.
In order to achieve the above purpose, the basic idea of the invention is as follows: aiming at a target point position which is not suitable for establishing a permanent observation station, the tide and hydrographic meteorological element data of other nearby point positions and the precise mapping of the tide and hydrographic meteorological element data of the point position are established through temporarily observed tide data of the target point position and tide and hydrographic meteorological element data of other nearby point positions. And after the temporary observation is finished, fitting the tide of the point by the tide and the hydrological meteorological element data of other points.
The technical scheme of the invention is as follows: a target point tidal data fitting method based on multi-source data driving comprises the following steps:
A. collecting multi-source data
Collecting multi-source data as standby information through various sensors arranged at a target point and nearby points, wherein the multi-source data comprises tide data, meteorological data and hydrological data; the tide data comprises measured tide data of a target point location and measured tide data of nearby point locations; the meteorological data comprise measured data of air pressure, wind power, wind direction and air temperature of nearby point positions; the hydrological data comprise measured data of water temperature, precipitation and salinity of nearby point sites; the target point location is a point location which needs to be subjected to tide actual report and forecast but cannot be subjected to long-term tide observation and tide observation of a system due to condition limitation; the nearby point location is a point location which is conditionally observed for long-term and systematic tides near the target point location;
B. relevance sorting and information screening of multi-source data
Performing relevance sorting and information screening on multi-source data by a relation analysis method for calculating a partial relevance coefficient, selecting partial data of a nearby point location with strong relevance with the tide data of a target point location as the input of a fitting model, and taking the tide data of the target point location as the output of the fitting model;
C. building fitting model based on radial basis function neural network
B, according to the input data and the output data selected in the step B, training and establishing a fitting model based on a radial basis function neural network, calculating mapping from a network hidden layer to an output layer by using a partial least square method, and fitting a mapping relation of multi-source data with strong correlation with the target point position tidal data to the target point position current tidal data;
the training process of the orthogonal least square method of the radial basis function neural network comprises the following steps:
setting N training samples, wherein each sample comprises an input training sample and an output training sample; taking each input training sample as the center c of a hidden node; find the jth input training sample xjFor the ith center ciThe formula is as follows:
wherein sigmaiThe default width is 1 for the width of the ith hidden node; calculating the response between input training samples according to the formula to finally obtain an NxN response matrix phi;
the response matrix Φ is subjected to an orthogonal trigonometric decomposition using Gram-Schmit law as follows:
Φ＝WU，
where u is an NxN order upper triangular matrix with a diagonal element of 1, and W is an NxN matrix with columns WiOrthogonal;
defining an error reduction rate erriAs a measure for measuring the contribution of the hidden node to the output:
wherein
trace is the trace of the matrix;
selection of erriThe vector with large numerical value is used as the center of the hidden node; giving a network training precision threshold rho, and selecting err with the maximum value at each stepiTaking the corresponding vector as a hidden node, performing Gram-Schmidt orthogonalization on the remaining vector in the previous step in the next step, and calculating erriStopping training until the following set precision conditions are satisfiedRefining:
wherein rho is defaulted to 0.05; thus obtaining M through trainingSThe center of each hidden node forms a radial basis function neural network; the obtained connection matrix from the hidden layer to the output layer of the radial basis function neural network is obtained by solving a partial least square method;
the input vector is linked with the hidden layer through an input layer of the radial basis function neural network, and each node in the hidden layer represents a radial basis function with the same dimension as the input dimension. Let input vector x have n independent variables, output vector y have l dependent variables, and the sample set has m samples. Thus, the input X is an m × n dimensional matrix and the output Y is an m × l dimensional matrix.
In the partial least square method, a Gaussian function is used as a radial basis function to carry out nonlinear transformation, and a hidden layer response matrix A is obtained. Wherein the elements in the ith row and the jth column are:
wherein xiFor the ith sample, | | · | | | represents the Euclidean distance, cjAnd σjRespectively the center and width of the jth basis function. And cj＝xjThe resulting matrix a is a symmetric matrix and the elements on the main diagonal are all 1. The response matrix a is an m × m dimensional square matrix, independent of the input dimension, and is determined only by the number of samples.
A partial least squares regression operation is performed between the response matrix a and the output matrix Y. After the principal component matrix T is extracted, the response matrix A and the output matrix Y are respectively projected on the principal component matrix T, and the radial basis function neural network based on partial least squares regression is obtained as follows:
Y＝TR+F＝AWR+F
wherein T is a principal component matrix of A, and is m × nTA dimension matrix; w is A conversion matrixIs m × nTA dimension matrix; r is nTA regression coefficient matrix of x l dimensions; f is a residual matrix of dimension m x l.
And training the radial basis function neural network based on partial least squares regression by using the actually measured tidal data of the target point location as output and using the tidal data, hydrological data and meteorological data of nearby point locations as input, and performing generalized application on the obtained neural network. When the method is applied, actually measured tide data, hydrological data and meteorological data of nearby point locations are used as input, and the output obtained by the neural network is the obtained tide information of the target point location.
D. Dynamic network adjustment based on sliding window
Under the condition that the target point position is updated with data, in order to reflect the latest mapping relation between the tide and hydrometeorology data of nearby point positions and the tide data of the target point position, a sliding window is established, and a fitting model based on a radial basis function neural network is dynamically adjusted by utilizing input and output data updated in real time.
The sliding window is a first-in first-out data sample sequence with fixed width, when a new group of input-output data is received, the new data group is added into the sliding window, the earliest group of data is shifted out of the sliding window, and the sliding window W at the moment t is usedSDExpressed as:
WSD＝[(pt-L+1，qt-L+1)，…，(pt，qt)]，
where L is the width of the sliding window.
The real-time dynamics of the mapping relationship is reflected by the input-output data sets within the sliding window, i.e. by the input matrix P and the corresponding output vector q, respectively:
in the formula, npIs the dimension of the input matrix.
And respectively using the input matrix P and the corresponding output vector q as the input and the output of a radial basis function neural network based on partial least squares regression, training and dynamically adjusting the neural network, and fitting the target point position tidal data by using the updated neural network.
Compared with the prior art, the invention has the following beneficial effects:
1. although the traditional forecasting method based on harmonic analysis can reflect the influence of celestial body tidal attraction and give stable tidal forecasting, the traditional forecasting method cannot reflect the influence of factors such as time-varying aperiodic hydrological weather and the like, so that the condition of low forecasting precision occurs, and a larger forecasting error is easy to occur under the condition of large influence of external environmental factors. The invention utilizes the actually measured tide information and the hydrometeorological information of nearby point positions, the information fully reflects the influence of the hydrometeorological factors on the tide, and the radial basis function neural network based on partial least squares regression can realize accurate high-dimensional nonlinear fitting and can obtain more accurate tide forecast.
2. Under the condition that certain sea area point locations are not suitable for arranging permanent tide observation stations, the invention provides accurate tide fitting results for the point locations by using tide observation point data near the sea area. Only a temporary point location observation device is needed to be established, and the temporary observation device can be removed after the mapping relation between the surrounding point location information and the point location tide data is established.
3. Under the condition that the tide information of the tide observation point location is lost, the accurate fitting data of the observation point location can be obtained through the tide data of the nearby point location.
4. The invention can be popularized and establishes a tide observation and forecasting system covering a certain sea area.
Drawings
FIG. 1 is a neural network training input-output model.
FIG. 2 is a neural network generalized input-output model.
Fig. 3 is a flow chart of the present invention.
FIG. 4 is a schematic view of the tidal forecast of the present invention at the target site.
FIG. 5 is a schematic view of the tidal at the target site of harmonic analysis forecasts.
Detailed Description
The invention is further described below with reference to the accompanying drawings. As shown in figure 1 of the drawings, in which,
example (c): the actual tide report is needed at a certain target point in the center of a main channel for entering a port at a certain coastal port, but the navigation safety of a ship can be influenced by arranging a fixed tide measuring device. In this case, the method provided in fig. 1-3 of the present invention may be adopted to set a temporary tidal measurement device at the target site for a period of time to measure the tidal data before the runway is deployed, and simultaneously measure the tidal data and the hydrometeorological data of nearby sites to establish a mapping relationship between the tidal, hydrological, and meteorological data of nearby sites and the tidal data of the target site. After the temporary tidal measurement device of the target point is removed, the tidal data of the target point can be acquired by generalization of the established mapping by using tidal, hydrological and meteorological data of nearby points.
Setting the tide data of the target point position as H at the time t0(t), tide, air pressure, wind power, wind direction, air temperature, water temperature, precipitation and salinity information of R nearby point locations are used as alternative inputs, partial information with strong correlation with the tide data of the target point location is selected as the input of a fitting model by calculating partial correlation coefficients of the tide, hydrological and meteorological data of the R nearby point locations and the tide data of the target point location, and the tide data H of the target point location is used as the input of the fitting model0And (t) as output, establishing a radial basis function neural network based on a partial least square method as a fitting model.
After the model is established, the measured tide, hydrological data and meteorological data of nearby point locations are used as input, the tide of the target point location is generalized, the generalized result is compared with the measured tide data of the target point location, and the result is shown in fig. 4. As can be seen from fig. 4, the fitting result has higher generalization accuracy.
Correspondingly, when the tide at the point is forecasted by the traditional harmonic analysis method, the forecast result is as shown in fig. 5, and the forecast result has a larger deviation from the actual measurement result.
Comparing the results shown in fig. 4 and fig. 5, it can be seen that by the tidal data fitting method based on multi-source data driving provided by the invention, the tidal information of the target point location can be obtained by accurately fitting the measured information of the tide and the hydrometeorology of the nearby point location, and the fitting accuracy is higher than that of the traditional harmonic analysis method.
The present invention is not limited to the embodiment, and any equivalent idea or change within the technical scope of the present disclosure is to be regarded as the protection scope of the present invention.
Claims (1)
1. A target point position tidal data fitting method based on multi-source data driving is characterized in that: the method comprises the following steps:
A. collecting multi-source data
Collecting multi-source data as standby information through various sensors arranged at a target point and nearby points, wherein the multi-source data comprises tide data, meteorological data and hydrological data; the tide data comprises measured tide data of a target point location and measured tide data of a nearby point location; the meteorological data comprise measured data of air pressure, wind power, wind direction and air temperature of nearby point positions; the hydrological data comprise measured data of water temperature, precipitation and salinity of nearby point locations; the target point location is the point location which needs to be subjected to tide actual report and forecast but cannot be subjected to long-term and system tide observation due to condition limitation; the nearby point location is a point location which is conditionally observed for long-term and systematic tides near the target point location;
B. relevance sorting and information screening of multi-source data
Performing relevance sorting and information screening on multi-source data by a relation analysis method for calculating a partial relevance coefficient, selecting partial data of a nearby point location with strong relevance with the tide data of a target point location as the input of a fitting model, and taking the tide data of the target point location as the output of the fitting model;
C. building fitting model based on radial basis function neural network
B, according to the input data and the output data selected in the step B, training and establishing a fitting model based on a radial basis function neural network, calculating mapping from a network hidden layer to an output layer by using a partial least square method, and fitting a mapping relation of multi-source data with strong correlation with the target point position tidal data to the target point position current tidal data;
the training process of the orthogonal least square method of the radial basis function neural network comprises the following steps:
setting N training samples, wherein each sample comprises an input training sample and an output training sample; taking each input training sample as the center c of a hidden node; finding the jth input training sample xjFor the ith center ciThe formula is as follows:
wherein sigmaiThe default width is 1 for the width of the ith hidden node; calculating the response between input training samples according to the formula to finally obtain an NxN response matrix phi;
the response matrix Φ is subjected to an orthogonal trigonometric decomposition using Gram-Schmit law as follows:
Φ＝WU，
where u is an NxN order upper triangular matrix with a diagonal 1, and W is an NxN matrix with columns WiOrthogonal;
defining an error reduction rate erriAs a measure for measuring the contribution of the hidden node to the output:
wherein
trace is the trace of the matrix;
selection of erriVector with large value asThe center of the hidden node; giving a network training precision threshold rho, and selecting err with the maximum value at each stepiTaking the corresponding vector as a hidden node, orthogonalizing the rest vectors in the previous step in Gram-Schmidt mode in the next step, and calculating erriAnd stopping training until the following set precision conditions are met:
wherein rho is defaulted to 0.05; thus obtaining M through trainingSThe center of each hidden node forms a radial basis function neural network; the obtained connection matrix from the hidden layer to the output layer of the radial basis function neural network is obtained by solving a partial least square method;
the input vector is linked with the hidden layer through an input layer of the radial basis function neural network, each node in the hidden layer represents a radial basis function, and the dimension of the radial basis function is the same as that of the input; setting an input vector x to have n independent variables, an output vector y to have l dependent variables, and a sample set to have m samples; thus, the input X is an m × n dimensional matrix, and the output Y is an m × l dimensional matrix;
in a partial least square method, a Gaussian function is used as a radial basis function to carry out nonlinear transformation to obtain a hidden layer response matrix A; wherein the elements in the ith row and the jth column are:
wherein xiFor the ith sample, | | · | | | represents the Euclidean distance, cjAnd σjThe center and width of the jth basis function, respectively; and cj＝xjTherefore, the obtained matrix A is a symmetric matrix, and the elements on the main diagonal are all 1; the response matrix A is an m multiplied by m dimensional square matrix, is irrelevant to the input dimension and is only determined by the number of samples;
performing partial least squares regression operation between the response matrix A and the output matrix Y; after the principal component matrix T is extracted, the response matrix A and the output matrix Y are respectively projected on the principal component matrix T, and the radial basis function neural network based on partial least squares regression is obtained as follows:
Y＝TR+F＝AWR+F
wherein T is a principal component matrix of A, and is m × nTA dimension matrix; w is a transformation matrix of m × nTA dimension matrix; r is nTA regression coefficient matrix of x l dimensions; f is a residual error matrix with dimension of m multiplied by l;
training a radial basis function neural network based on partial least squares regression by using actually-measured tidal data of a target point location as output and using tidal data, hydrological data and meteorological data of nearby point locations as input, and performing generalized application on the obtained neural network; when the method is applied, actually measured tide data, hydrological data and meteorological data of nearby point locations are used as input, and the output obtained by the neural network is the obtained tide information of the target point location;
D. network dynamic adjustment based on sliding window
Under the condition that the data is updated at the target point, in order to reflect the latest mapping relation between the tide and hydrometeorology data of nearby points and the tide data of the target point, a sliding window is established, and a fitting model based on a radial basis function neural network is dynamically adjusted by utilizing the input and output data updated in real time;
the sliding window is a first-in first-out data sample sequence with fixed width, when a new group of input-output data is received, the new data group is added into the sliding window, the earliest group of data is shifted out of the sliding window, and the sliding window W at the moment t is usedSDExpressed as:
WSD＝[(pt-L+1，qt-L+1)，…，(pt，qt)]，
wherein L is the width of the sliding window;
the real-time dynamics of the mapping relationship is reflected by the input-output data sets within the sliding window, i.e. by the input matrix P and the corresponding output vector q, respectively:
in the formula, npIs the dimension of the input matrix;
and respectively using the input matrix P and the corresponding output vector q as the input and the output of a radial basis function neural network based on partial least squares regression, training and dynamically adjusting the neural network, and fitting the target point position tidal data by using the updated neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110159143.7A CN112906955B (en) | 2021-02-04 | 2021-02-04 | Tidal data fitting method based on multi-source data driving |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110159143.7A CN112906955B (en) | 2021-02-04 | 2021-02-04 | Tidal data fitting method based on multi-source data driving |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112906955A CN112906955A (en) | 2021-06-04 |
| CN112906955B true CN112906955B (en) | 2022-06-21 |
Family
ID=76122623
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110159143.7A Active CN112906955B (en) | 2021-02-04 | 2021-02-04 | Tidal data fitting method based on multi-source data driving |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112906955B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114969664B (en) * | 2022-06-01 | 2023-03-21 | 广州市城市规划勘测设计研究院 | Water level correction method, device, equipment and medium |
| CN116702095B (en) * | 2023-06-01 | 2024-04-05 | 大连海事大学 | Modularized marine ship motion attitude real-time forecasting method |
| CN116822336B (en) * | 2023-06-01 | 2024-04-09 | 大连海事大学 | Multi-model combined tide forecasting method |
| CN117518298A (en) * | 2023-12-29 | 2024-02-06 | 大连海事大学 | Tidal combination forecasting method with self-adaptive model structure |
| CN117894158A (en) * | 2024-03-15 | 2024-04-16 | 江苏省气象台 | Cold and tide disaster risk pre-assessment method based on intelligent grid air temperature prediction |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102109619A (en) * | 2009-12-23 | 2011-06-29 | 栗小东 | System and method for predicting typhoon surge based on artificial intelligence |
| 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 |
| CN106339355A (en) * | 2016-08-31 | 2017-01-18 | 哈尔滨工程大学 | Carrier landing short-term track online predicting method of carrier-based aircraft |
| KR101866380B1 (en) * | 2017-02-14 | 2018-06-12 | 인하대학교 산학협력단 | Method for extracting occurrence date of meteo-tsunami by analazing and handling missing section of tidal data |
-
2021
- 2021-02-04 CN CN202110159143.7A patent/CN112906955B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102109619A (en) * | 2009-12-23 | 2011-06-29 | 栗小东 | System and method for predicting typhoon surge based on artificial intelligence |
| 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 |
| CN106339355A (en) * | 2016-08-31 | 2017-01-18 | 哈尔滨工程大学 | Carrier landing short-term track online predicting method of carrier-based aircraft |
| KR101866380B1 (en) * | 2017-02-14 | 2018-06-12 | 인하대학교 산학협력단 | Method for extracting occurrence date of meteo-tsunami by analazing and handling missing section of tidal data |
Non-Patent Citations (2)
| Title |
|---|
| Sequential learning radial basis function network for real-time tidal level predictions;尹建川等;《Ocean Engineering》;20121016;第57卷;第49-55页 * |
| 径向基函数神经网络及其在船舶运动控制中的应用研究;尹建川;《万方数据》;20080825;正文第1-94页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112906955A (en) | 2021-06-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112906955B (en) | Tidal data fitting method based on multi-source data driving | |
| Solari et al. | The wind forecast for safety management of port areas | |
| Landberg et al. | Short-term prediction of local wind conditions | |
| CN110020462B (en) | Method for fusing meteorological data and generating numerical weather forecast | |
| Chu et al. | South China sea wind-wave characteristics. part I: validation of WAVEWATCH-III using TOPEX/Poseidon data | |
| Alvarez et al. | Comparison of different wind products and buoy wind data with seasonality and interannual climate variability in the southern Bay of Biscay (2000–2009) | |
| CN107688906B (en) | Multi-method fused transmission line meteorological element downscaling analysis system and method | |
| Donelan et al. | On the growth rate of wind‐generated waves | |
| Carvalho et al. | Comparison between CCMP, QuikSCAT and buoy winds along the Iberian Peninsula coast | |
| Halliwell Jr et al. | OSSE impact analysis of airborne ocean surveys for improving upper-ocean dynamical and thermodynamical forecasts in the Gulf of Mexico | |
| CN104915534A (en) | Deformation analysis and decision-making method of electric power tower based on sequence learning | |
| CN113128758B (en) | Maximum wave height forecasting system constructed based on offshore buoy wave observation data | |
| CN112946784B (en) | SuperDARN radar convection diagram short-term forecasting method based on deep learning | |
| Zhang et al. | Analysis of 50-year wind data of the southern Baltic Sea for modelling coastal morphological evolution–a case study from the Darss-Zingst Peninsula | |
| Roberts et al. | A regression-based approach for cool-season storm surge predictions along the New York–New Jersey coast | |
| CN113704693A (en) | High-precision effective wave height data estimation method | |
| CN115310706A (en) | Typhoon path intelligent forecasting method | |
| Bore et al. | Determination of environmental conditions relevant for the ultimate limit state at an exposed aquaculture location | |
| Bailey | Wind resources for offshore wind farms: Characteristics and assessment | |
| Pore | The relation of wind and pressure to extratropical storm surges at Atlantic City | |
| Foley et al. | A statistical framework to combine multivariate spatial data and physical models for hurricane surface wind prediction | |
| Camarasa et al. | Rainfall–runoff modelling of ephemeral streams in the Valencia region (eastern Spain) | |
| CN115204073A (en) | Tide prediction method based on kriging interpolation | |
| Qi et al. | Establishment and tests of EnOI assimilation module for WAVEWATCH III | |
| Nemes | Statistical analysis of wind speed profile: a case study from Iasi Region, Romania |
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 |