CN114139448B - Method, system, medium, terminal and application for optimizing sea-based observation network station layout - Google Patents
Method, system, medium, terminal and application for optimizing sea-based observation network station layout Download PDFInfo
- Publication number
- CN114139448B CN114139448B CN202111434536.0A CN202111434536A CN114139448B CN 114139448 B CN114139448 B CN 114139448B CN 202111434536 A CN202111434536 A CN 202111434536A CN 114139448 B CN114139448 B CN 114139448B
- Authority
- CN
- China
- Prior art keywords
- station
- observation
- optimal
- sea
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Testing Or Calibration Of Command Recording Devices (AREA)
Abstract
The invention belongs to the technical field of sea-based observation network station information processing, and discloses a method, a system, a medium, a terminal and application for optimizing sea-based observation network station layout. Combining an EOF reconstruction method with a K central point clustering algorithm, simultaneously adding proportion control factors, and optimizing by combining actual conditions of different ocean numerical modes to obtain an optimal station under the comprehensive condition of a plurality of sets of ocean numerical modes. The method has short calculation time, the conventional data assimilation and CNOP usually need several days or even longer time to complete the calculation, and the method only needs tens of minutes; the method can be fully verified based on reanalysis data or post-reporting results, and can be developed based on a plurality of sets of reanalysis data or post-reporting data, so that mode dependence is avoided to the maximum extent; compared with the existing EOF reconstruction method which only considers the optimal station observed by a single variable such as sea surface height or sea surface temperature, the method can simultaneously consider three-dimensional temperature, salinity and sea current to give the optimal station.
Description
Technical Field
The invention belongs to the technical field of information processing of a site of a base observation network, and particularly relates to a method, a system, a medium, a terminal and application for optimizing the layout of a site of a sea-based observation network.
Background
At present, the existing sea-based observation network observation station optimization methods have three types:
(1) The data assimilation method comprises the steps that a marine assimilation numerical prediction mode is operated, a station with the largest prediction accuracy improvement effect is screened and predicted to serve as an optimal station, and the method has two defects that firstly, the method depends on the prediction performance of the numerical prediction mode, for example, the reliability of the optimal station obtained by the poor numerical prediction mode is poor, secondly, the calculated amount is huge, the method must depend on an ultra-computation platform, the calculation cost is high, and the transportability is poor, so that the method is difficult to transplant from one ultra-computation platform to another ultra-computation platform;
(2) The CNOP method finds the area with the fastest error increase by operating a numerical prediction mode, and takes the area as the selection of the optimal observation station, and has two disadvantages, namely, the CNOP method depends on the prediction performance of the numerical prediction mode, and the CNOP method pays more attention to specific target observation and has insufficient consideration on the coverage capability of the whole sea area observation;
(3) According to the traditional EOF reconstruction method, the observation elements of the whole observation sea area can be reconstructed based on the data observed by the station through the EOF reconstruction method, but only a single observation element field on a two-dimensional level and a horizontal single depth level, such as sea surface height, 10 m sea temperature and the like, can not be processed by three-dimensional variables.
To solve the above technical problems, prior art 1: a Model-Based Assessment and Design of a pathological Indian Ocean Mooring Array, based on EOF function, includes the following steps:
firstly, carrying out EOF decomposition on a 20 ℃ isotherm depth field and a mixed layer depth field to respectively obtain main modes of the two; projecting a series of observation data onto dominant Empirical Orthogonal Functions (EOFs), and sequencing importance of observation positions to generate a theoretical optimal observation array; and counting the most observed arrays of the two variables by adopting an artificially divided grid with fixed size, compiling a relative frequency graph of the observed positions and determining a uniform observed array. However, the technical drawback stored in prior art 1 is that the EOF reconstruction method only considers a single variable, the integration effect depends on the horizontal grid resolution, and is affected by human subjective factors.
Prior art 2: a multi-beam submarine topography correction method based on sound velocity profile inversion is characterized by comprising the following steps:
step one, constructing an objective function;
constructing a target function based on the sounding consistency of a plurality of ping corresponding beams in the overlapping area of adjacent strips;
determining an initial temperature;
setting the initial solution, the value range of the EOF coefficient and a neighborhood function;
calculating to obtain the value range of the EOF time coefficient according to the first feature vectors and the actually measured SVP, wherein the neighborhood function enables the generated candidate solution alpha to spread over a solution space, and the neighborhood function is given;
step four, setting an inner circulation termination criterion;
judging whether the inner circulation reaches a thermal equilibrium state or not by using the weighted average value of the target function;
step five, determining a temperature cooling function;
a temperature cooling function was constructed as follows:
T k+1＝βT k，β∈[0.5，0.99]；
step six, determining an outer loop termination criterion;
the external circulation termination condition is that no acceptable candidate solution exists at a plurality of continuous temperatures;
step seven, searching an optimal solution;
setting a memory function, storing the minimum value of the target function at each temperature, and finding out the minimum value of the global target function;
step eight, correcting distorted submarine topography;
and (4) performing multi-beam data processing by using the inverted SVP, and correcting the distorted submarine topography.
In the prior art 2, the EOF reconstruction method only considers a single variable, and cannot simultaneously consider three-dimensional temperature, salinity and ocean current, so that the accuracy of the optimal station is poor.
Through the above analysis, the problems and defects of the prior art are as follows:
(1) The prior art has large calculated amount, needs multiple machines to complete the operation and has high calculation cost;
(2) The prior art has long calculation time and low calculation efficiency;
(3) In the prior art, in analyzing data or reporting results, the artificial subjective dependence is strong, so that the processed data information has large deviation;
(4) The existing EOF reconstruction method only considers single variables such as sea surface height or optimal station position observed by sea surface temperature, so that the obtained information has poor referential property.
The difficulty in solving the above problems and defects is:
(1) The calculation amount is reduced, and the difficulty is high;
(2) The calculation time is shortened, and the difficulty is high;
(3) The influence of subjective factors needs to be reduced, and the difficulty is moderate;
(4) From two-dimensional univariate to three-dimensional multivariable, the effective integration is required, and the difficulty is high.
The significance for solving the problems and the defects is as follows:
(1) The calculation amount and the calculation cost are effectively reduced, the technology can realize single-machine operation and can be completed on a personal computer without the limitation of hardware equipment;
(2) The calculation time is greatly shortened, and the method is quick and efficient;
(3) The influence of human subjective factors is eliminated, so that the data result is more objective;
(4) The technology can comprehensively consider a plurality of three-dimensional marine variables through improvement, so that the designed observation array can effectively capture the change characteristics of various variables, the applicability and the universality are enhanced, and the reference significance is achieved.
Disclosure of Invention
In order to overcome the problems in the related art, the disclosed embodiments of the present invention provide a method, system, medium, terminal and application for optimizing the layout of sea-based observation network sites.
The technical scheme is as follows: a method for optimizing the site layout of a sea-based observation network comprises the following steps:
combining an EOF reconstruction method with a K central point clustering algorithm, simultaneously adding proportion control factors, and optimizing by combining actual conditions of different ocean numerical modes to obtain an optimal station under the comprehensive condition of a plurality of sets of ocean numerical modes.
In one embodiment, the method for optimizing the site layout of the sea-based observation network comprises the following steps:
(1) Performing data extraction on L different ocean numerical modes according to uniform resolution, and selecting the temperature, salinity and ocean current data of each mode in an analysis sea area to generate a set W (L, V, I, j, Z); then based on an EOF reconstruction method, respectively carrying out station optimization on the single two-dimensional variable to obtain an optimal station set S (L, V, Z) of each variable at different depths, wherein elements S (L, V, Z) are optimal observation station distribution of a V (V is more than or equal to 1 and less than or equal to V) variable of a L (L is more than or equal to 1 and less than or equal to L) mode on a Z (Z is more than or equal to 1 and less than or equal to Z) layer, V is the number of variables, and Z is the number of depth layers;
(2) Before K central point clustering is carried out, obtaining a self-adaptive weight matrix by utilizing observation data, model simulation data and an L1 norm; the influence factors of the weight are from the approximation degree of temperature, salinity and ocean current data in different ocean numerical modes relative to a real observation data standard set; integrating a set S (L, V, Z) obtained by an EOF reconstruction method with a self-adaptive weight matrix to obtain a weighted observation station set S' (L, V, Z);
(3) And dividing the observation station position set S' (L, V and Z) into K clusters by adopting a K central point clustering algorithm, and taking the central point of each cluster as the finally designed observation station position.
In an embodiment, before performing K-center clustering, the step (2) obtains an adaptive weight matrix by using the observation data, the model simulation data, and the L1 norm, and includes:
1) Firstly, taking the average value of one or more kinds of observation data as observation standard data, namely, taking the average value as the true value of ocean data, and recording the true value as O (V, I, J, Z), wherein I is the number of lattice points on each latitude, and J is the number of lattice points on each longitude;
2) Respectively calculating each mode data W (l, V, I, J, Z) & gtY l＝1，2，...，L The error E (V, I, J, Z) relative to the observation standard data O (V, I, J, Z) can be root mean square error, absolute error, relative error or correlation coefficient; then, the spatial average of the errors is calculated and all the modes are combined to obtain a set E 0 (L，V)；
3) And calculating an adaptive weight matrix B by using the L1 norm, wherein the following formula is satisfied:
where the L1 norm is the sum of the absolute values of the elements of the vector, e.g., the L1 norm of matrix B is:
(3) Multiplying the observation station set S (L, V, Z) obtained through EOF analysis optimization with each element in the weight matrix B (V, L) respectively to obtain a weighted observation station set S' (L, V, Z), namely:
S′(l，v，z)＝S(l，v，z)×B(v，l)，
wherein L is more than or equal to 1 and less than or equal to L, V is more than or equal to 1 and less than or equal to V, and Z is more than or equal to 1 and less than or equal to Z.
In an embodiment, the method for optimizing the site layout of the sea-based observation network further includes the following steps:
s1, respectively carrying out Station optimization on a single two-dimensional variable based on an EOF reconstruction method to obtain a set { Station _ Temperature | depth = i, station _ sales | depth = i, and Station _ Velocity | depth = i } of optimal stations of variables (Temperature, salinity and ocean current) at different depths;
s2, obtaining a clustering result of a set { Station _ Temperature | depth = i, station _ sales | depth = i, and Station _ Velocity | depth = i } of the set optimal Station by using a K-center clustering algorithm, wherein the clustering result is used as the optimal Station under the condition of multiple variables;
and S3, reporting results by using different ocean numerical modes, obtaining the set { Station | model = i } of the optimal Station, and obtaining the optimal Station under the comprehensive condition of a plurality of sets of ocean numerical modes by using the K-center clustering method in the second step.
In an embodiment, the step S1 specifically includes:
projecting the observation data onto dominant Empirical Orthogonal Functions (EOFs) to construct an analysis model, wherein the dominant Empirical Orthogonal Functions (EOFs) are generated by long-time observation data; the analytical model is as follows:
wherein, w a Is the n-dimensional column vector of the mapping field, n is the number of grid points; w is an n-dimensional column vector, is a time average, M is an n x M matrix,
column i of M is the ith major EOF, w i EOF M is the number of EOFs used; c is an m-dimensional column vector containing the amplitude or weighting coefficients of the EOF;
determining a coefficient c by performing least square fitting of a series of observed values on the EOFs, and generating a single time point; by calculating a least squares solution of the linear system (2.3);
HMc＝w o (2.3)
the analysis given by (2.1) is a linear combination of the time average plus the EOFs, where the amplitude of the EOFs is determined by (2.3) through the vector c;
the columns of the matrix M are mutually orthogonal, and the columns of the matrix M are normalized to obtain a standard orthogonal matrix
Satisfies the following conditions:
only a few observations are relevant to the number of grid points,
is not orthogonal, restricted by the observation position formally represented by the operator H; defining an observation array H that will generate reliable analyses from the analytical model (2.1) and that will be if and only if it satisfies the condition (2.5)An optimal array;
according to definition, the condition number that all the eigenvalues of the identity matrix I are 1, I is also 1; starting from n observation positions, independently excluding one observation point at a time, and obtaining the rest
The condition numbers are calculated, the observation position with the minimum condition number is eliminated, leaving (n-1) observation positions; until each remaining observation is independently excluded, leaving a desired number of observation positions; the final configuration of the observation locations is treated as an optimal array.
In an embodiment, the step S2 specifically includes:
the optimal array for three-dimensional temperature and salinity observation variables obtained by the EOF reconstruction method is inconsistent in the positions of observation stations at different depths; and dividing similar elements into closely related clusters by using a K central point clustering algorithm, so that different arrays are consistent in the vertical direction aiming at different variables.
Another object of the present invention is to provide a sea-based observation network site layout optimization system for implementing the method for sea-based observation network site layout optimization, the sea-based observation network site layout optimization system including:
a single two-dimensional variable optimal station set acquisition module used for respectively carrying out station optimization on single two-dimensional variables based on an EOF reconstruction method to obtain an optimal station set of the variables at different depths
{Station_Temperature|depth＝i，Station_Salinity|depth＝i，Station_Velocity|depth＝i}；
A multivariable optimal station set acquisition module for obtaining a set of optimal stations of the set of optimal stations in the variables at different depths by using a K-center clustering algorithm
{ Station _ Temperature | depth = i, station _ sales | depth = i, and Station _ Velocity | depth = i } as an optimal Station in a multivariate situation;
and the multiple sets of ocean numerical mode optimal Station position acquisition modules are used for reporting results after different ocean numerical modes are utilized to obtain a set { Station | model = i } of the optimal Station positions, and obtaining the optimal Station positions under the comprehensive condition of the multiple sets of ocean numerical modes by utilizing a K-center clustering method.
It is another object of the present invention to provide a program storage medium for receiving user input, the stored computer program causing an electronic device to perform the method for sea-based observational website layout optimization.
Another object of the present invention is to provide an information data processing terminal including a memory and a processor, the memory storing a computer program, the computer program, when executed by the processor, causing the processor to execute the method of sea-based observation site layout optimization.
The invention also aims to provide a sea-based observation network observation station system which utilizes the optimized method layout of the sea-based observation network station layout.
By combining all the technical schemes, the invention has the advantages and positive effects that:
(1) The invention has small calculation amount, can complete the operation by a single machine and has low calculation cost; the conventional data assimilation method needs a companion mode, has a large calculation amount, needs to be carried out on a high-performance computing cluster, and cannot be operated on a personal computer or a workstation.
(2) The method has short calculation time, the conventional data assimilation and CNOP usually need several days or even longer time to complete the calculation, and the method only needs tens of minutes;
(3) The method can be fully verified based on reanalysis data or post-reporting results, and can be developed based on a plurality of sets of reanalysis data or post-reporting data, so that mode dependence is avoided to the maximum extent;
(4) Compared with the existing EOF reconstruction method which only considers the optimal station observed by a single variable such as sea surface height or sea surface temperature, the method can simultaneously consider three-dimensional temperature, salinity and sea current to give the optimal station. The invention can simultaneously consider a plurality of variables needing important observation, greatly saves resources and cost and has universality.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the disclosure.
Fig. 1 is a flowchart of a method for optimizing a site layout of a sea-based observation network according to an embodiment of the present invention.
Fig. 2 is a schematic diagram of a sea-based observation network site layout optimization system according to an embodiment of the present invention.
In the figure: 1. a single two-dimensional variable optimal station set acquisition module; 2. a multivariable optimal station set acquisition module; 3. and a plurality of sets of ocean numerical mode optimal station acquisition modules.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention more comprehensible, embodiments accompanying figures are described in detail below. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein, but rather should be construed as broadly as the present invention is capable of modification in various respects, all without departing from the spirit and scope of the present invention.
As shown in fig. 1, a method for optimizing a site layout of a sea-based observation network according to an embodiment of the present invention includes:
s101, based on an EOF reconstruction method, station optimization is respectively carried out on single two-dimensional variables, and a set { Station _ Temperature | depth = i, station _ sales | depth = i, and Station _ Velocity | depth = i } of optimal stations of variables (Temperature, salinity, ocean current) in different depths is obtained.
S102, obtaining a clustering result of the set { Station _ Temperature | depth = i, station _ sales | depth = i, and Station _ Velocity | depth = i } of the set optimal Station by using a K-center clustering algorithm, wherein the clustering result is used as the optimal Station under the condition of multivariate.
S103, reporting results after different ocean numerical modes, obtaining the set { Station | model = i } of the optimal Station, and obtaining the optimal Station under the condition of comprehensively considering a plurality of sets of ocean numerical modes by using the K-center clustering method in the second step, thereby avoiding the dependence of the screened optimal Station on different modes to the maximum extent.
In a preferred embodiment of the present invention, step S1 specifically includes the following for a single two-dimensional variable:
the purpose of the analytical model is to reconstruct a grid two-dimensional map of key ocean variables from long-term observation data and array locations. The specific operation is to project the observed data onto a dominant Empirical Orthogonal Function (EOFs), i.e. using the dominant empirical orthogonal function of the variables as a set of basis functions, whereas the EOFs are generated from long-term observed data. Each assay was as follows:
wherein w a Is the n-dimensional column vector of the mapping field, n is the number of grid points;
is an n-dimensional column vector, is a time average, M is an n x M matrix,
column i of M is the ith major EOF, w i EOF M is the number of used primary EOFs; c is an m-dimensional column vector containing the amplitude or weighting coefficients of the EOF.
To generate a single point-in-time analysis, the coefficient c must be determined by a least squares fit of a series of observations to the EOFs. Formally, this is achieved by computing the least squares solution of the linear system (2.3).
HMc＝w o (2.3)
In fact, the analysis given by (2.1) is a linear combination of time average plus EOFs, where the amplitude of the EOFs is determined by (2.3) through the vector c. Similarly, the analysis from the higher level assimilation system is simply a linear combination of the background field plus the ensemble members.
The ability of the analytical models (2.1) - (2.3) to determine the coefficients in c depends on the effect of the observation mapping onto the EOFs and, more specifically, on their degree of discrimination between the different EOFs. Through EOF analysis, the columns (EOFs basis functions) of the matrix M are mutually orthogonal, and if the column of M is standardized by the invention, a standard orthogonal matrix is obtained
Satisfies the following conditions:
only a few observations are actually relevant to the number of grid points, so
Is not generally orthogonal, and depends on the observation position formally represented by the operator H. Here, the invention defines an observation array H that will generate a reliable analysis from the analytical model (2.1), and is considered to be the optimal array if and only if it satisfies the condition (2.5). It is noted that other criteria may be used to define the optimal array.
For any application, the observation array cannot be guaranteed to meet the requirement of (2.5). Thus, in practice, the present invention contemplates configuring the observation array (i.e., defining H) so that
As close to I as possible. To quantize the matrix product
The present invention takes into account its condition number (ratio of maximum eigenvalue to minimum eigenvalue, condition number of 2-norm). By definition, all eigenvalues of the identity matrix I are 1, so the condition number of I is also 1, so the present invention needs to find a given ^ greater than or equal to>
The position of the minimum condition number. The invention starts with n observation positions, one observation point at a time being excluded independently, so that the remaining part is->
The condition number may be calculated. Eventually the observation location with the minimum condition number is eliminated, leaving (n-1) observation locations. This process is then repeated, with each remaining observation being independently excluded until the desired number of observation positions remain. The final configuration of the observation locations is considered to be the optimal array for this particular configuration of the analysis system.
In a preferred embodiment of the present invention, step S2 specifically includes the following steps:
the EOF-based method gives an optimal array for three-dimensional observation variables such as temperature and salinity, but the observation site locations at different depths are not consistent. Therefore, it is necessary to integrate the different arrays so that they are uniform for different variables in the vertical direction. The present invention uses the K-center clustering algorithm introduced in Park and Junand et al (2008), which is capable of dividing similar elements into closely related clusters.
The K center point clustering algorithm is a common clustering algorithm, and the method can reduce the influence of certain isolated data on the clustering process by a calculation mode with the shortest sum of distances, so that the final effect is closer to real division.
Suppose that N points should be divided into K (K < N) clusters, where K ultimately requires the number of sites to be observed. The specific algorithm comprises the following steps:
(1) Initializing K center points: randomly selecting K points from all the optimal position sets as the central points of all the clusters;
(2) N points are classified into K clusters. First, manhattan distance is used as a measure of dissimilarity. The manhattan distance between point i and point j is given by:
d(i，j)＝|x i -x j |+|y i -y j |. (2.6)
secondly, classifying the N points into K clusters by distributing each point to the nearest initial center;
(3) The positions of the K center points are updated. In each cluster, sequentially selecting points according to the sequence, calculating the sum of distances from the point to all the points in the current cluster, and selecting the point with the minimum sum of the distances as a new central point;
(4) Repeating steps (2) and (3) until the sum of the distances of the K clusters varies by less than 10 -4 m, which are considered to be no longer changing, and then the K center points are considered to be the final integrated array observation positions.
In a preferred embodiment of the present invention, step S3 further comprises: because the simulation effects of different ocean numerical modes aiming at physical quantities such as temperature, salinity, ocean current and the like are different, or a certain variable needs to be observed in practical design in a key mode, the method can be reasonably optimized by combining an EOF reconstruction method with a K central point clustering algorithm and adding specific gravity control factors in combination with the practical conditions of different ocean numerical modes. The specific operation is as follows:
(1) If the invention comprehensively considers L ocean numerical modes, data extraction is carried out on different ocean numerical modes according to uniform resolution, and a set W (L, V, I, J, Z) is generated by selecting the temperature, salinity and ocean current data of each mode in the analysis sea area; and then respectively carrying out station optimization on the single two-dimensional variable based on an EOF reconstruction method to obtain an optimal station set S (L, V, Z) of each variable at different depths, wherein elements S (L, V, Z) are the optimal observed station position distribution of the V (V is more than or equal to 1 and less than or equal to V) variable of the L (L is more than or equal to 1 and less than or equal to L) mode on the Z (Z is more than or equal to 1 and less than or equal to Z) layer, V is the number of variables, and Z is the number of depth layers.
(2) Before clustering, the invention utilizes observation data, model simulation data and L1 norm to obtain an objective adaptive weight matrix. The influence factors of the weight mainly come from the approximation degree of temperature, salinity and ocean current data in different ocean numerical modes relative to a real observation data standard set. The specific calculation process is as follows:
1) Firstly, taking the average value of one or more kinds of observation data as observation standard data, namely, taking the average value as the true value of ocean data, and recording the true value as O (V, I, J, Z), wherein I is the number of lattice points on each latitude, and J is the number of lattice points on each longitude;
2) Respectively calculating each mode data W (l, V, I, J, Z) & gtY l＝1，2，...，L The error E (V, I, J, Z) relative to the observation standard data O (V, I, J, Z) may be a root mean square error, an absolute error, a relative error, a correlation coefficient, or the like. Then, the spatial average of the errors is calculated and all the modes are combined to obtain a set E 0 (L，V)；
3) The adaptive weight matrix B is calculated using the L1 norm, i.e. the following equation is satisfied:
where the L1 norm is the sum of the absolute values of the elements of the vector, e.g., the L1 norm of matrix B is:
(3) Multiplying the observation station position set S (L, V, Z) obtained through EOF analysis optimization with each element in the weight matrix B (V, L) respectively to obtain a weighted observation station position set S' (L, V, Z), namely:
S′(l，v，z)＝S(l，v，z)×B(v，l)，
wherein L is more than or equal to 1 and less than or equal to L, V is more than or equal to 1 and less than or equal to V, and Z is more than or equal to 1 and less than or equal to Z. And dividing the observation site into K clusters by adopting a K central point clustering algorithm, and regarding the central point of each cluster as a finally designed observation site, wherein the sites are consistent for different observation variables in the vertical direction.
As shown in fig. 2, the sea-based observation network site layout optimization system provided in the embodiment of the present invention includes:
the single two-dimensional variable optimal Station set acquisition module 1 is configured to perform Station optimization on a single two-dimensional variable respectively based on an EOF reconstruction method to obtain a set of optimal stations { Station _ Temperature | depth = i, station _ salience | depth = i, and Station _ Velocity | depth = i } of variables (Temperature, salinity, ocean current) at different depths.
The multivariate optimal Station position set obtaining module 2 is configured to obtain a clustering result of a set { state _ Temperature | depth = i, state _ salience | depth = i, and state _ Velocity | depth = i } of the set optimal Station positions by using a K-center clustering algorithm, and use the clustering result as the optimal Station position in a multivariate situation.
And the multiple sets of ocean numerical mode optimal Station position acquisition modules 3 are used for reporting results after different ocean numerical modes are utilized, obtaining a set { Station | model = i } of the optimal Station positions, and obtaining the optimal Station positions under the condition of comprehensively considering the multiple sets of ocean numerical modes by utilizing the K-center clustering method in the second step, so that the dependence of the screened optimal Station positions on different modes is avoided to the maximum extent.
The effects of the present invention will be further described below with reference to specific experiments.
The optimal array given by the EOF-based method is inconsistent in observation station positions of different modes and different variables at different depths. Therefore, the K central point clustering algorithm is adopted to reasonably integrate different arrays, so that different variables aiming at different ocean numerical mode data in the vertical direction are consistent, and the method has guiding significance for arrangement and design of actual observation stations.
Therefore, the invention can avoid the dependence of the screened optimal station on different modes to the maximum extent by adding the weight factor control between the EOF reconstruction method and the clustering, and obtain the observation station distribution meeting the specific design requirement to the maximum extent.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice in the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.
It will be understood that the present disclosure is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure should be limited only by the attached claims.
Claims (8)
1. A method for optimizing the layout of a site of a sea-based observation network is characterized by comprising the following steps:
combining an EOF reconstruction method with a K central point clustering algorithm, simultaneously adding a proportion control factor, and optimizing by combining actual conditions of different ocean numerical modes to obtain an optimal station under the comprehensive condition of a plurality of sets of ocean numerical modes;
the method for optimizing the site layout of the sea-based observation network comprises the following steps:
(1) According to the uniform resolution, carrying out data extraction on L different ocean numerical modes, and selecting the temperature, salinity and ocean current data of each mode in the analysis sea area to generate a set W (L, V, I, J, Z); then based on an EOF reconstruction method, respectively carrying out station optimization on the single two-dimensional variable to obtain an optimal station set S (L, V, Z) of variables at different depths, wherein elements S (L, V, Z) are the optimal observation station distribution of the V variable of the L mode on the Z layer, L is more than or equal to 1 and less than or equal to L, V is more than or equal to 1 and less than or equal to V, Z is more than or equal to 1 and less than or equal to Z, V is the number of variables, and Z is the number of depth layers;
(2) Before K central point clustering is carried out, obtaining a self-adaptive weight matrix by utilizing observation data, model simulation data and an L1 norm; the influence factors of the weight are from the approximation degree of temperature, salinity and ocean current data in different ocean numerical modes relative to a real observation data standard set; integrating the set S (L, V, Z) obtained by an EOF reconstruction method with a self-adaptive weight matrix to obtain a weighted observation station set S' (L, V, Z);
(3) Dividing an observation station position set S' (L, V and Z) into K clusters by adopting a K central point clustering algorithm, and taking the central point of each cluster as a finally designed observation station position;
in step (2), before K center point clustering is performed, an adaptive weight matrix is obtained by using observation data, model simulation data and an L1 norm, and the method includes:
1) Firstly, taking the average value of one or more kinds of observation data as observation standard data, namely, taking the average value as the true value of ocean data, and recording the true value as O (V, I, J, Z), wherein I is the number of lattice points on each latitude, and J is the number of lattice points on each longitude;
2) Respectively calculating each mode data W (l, V, I, J, Z) & gtY l＝1，2，...，L Relative to the error E (V, I, J, Z) of the observation standard data O (V, I, J, Z), adopting root mean square error, absolute error, relative error or correlation coefficient; then, the spatial average of the errors is calculated and all the modes are combined to obtain a set E 0 (L，V)；
3) And calculating an adaptive weight matrix B by using the L1 norm, wherein the following formula is satisfied:
wherein, L1 norm vector points to the sum of the absolute values of each element, and L1 norm of matrix B is:
(3) Multiplying the observation station set S (L, V, Z) obtained through EOF analysis optimization with each element in the weight matrix B (V, L) respectively to obtain a weighted observation station set S' (L, V, Z), namely:
S′(l，ｖ，z)＝S(l，v，z)×B(v，l)，
wherein L is more than or equal to 1 and less than or equal to L, V is more than or equal to 1 and less than or equal to V, and Z is more than or equal to 1 and less than or equal to Z.
2. The method of claim 1, wherein the method of sea-based observational network site layout optimization further comprises the steps of:
s1, respectively carrying out Station optimization on a single two-dimensional variable based on an EOF reconstruction method to obtain a set of optimal stations of different depths of Temperature, salinity and ocean current variables { Station _ Temperature | depth = i, station _ Salinity | depth = i, and Station _ Velocity | depth = i };
s2, obtaining a clustering result of the set { Station _ Temperature | depth = i, station _ sales | depth = i and Station _ Velocity | depth = i } of the optimal Station by using a K-center clustering algorithm, wherein the clustering result is used as the optimal Station under the condition of multivariable;
and S3, reporting results by using different ocean numerical modes to obtain a set { Station | model = i } of the optimal Station, and obtaining the optimal Station under the comprehensive condition of a plurality of sets of ocean numerical modes by using a K-center clustering method in the second step.
3. The method for optimizing sea-based observation network site layout according to claim 2, wherein the step S1 specifically includes:
projecting the observation data to a dominant empirical orthogonal function EOFs (empirical orthogonal function) to construct an analysis model, wherein the dominant empirical orthogonal function EOFs is generated by long-time observation data; the analytical model is as follows:
wherein w a Is the n-dimensional column vector of the mapping domain, n is the number of grid points;
is an n-dimensional column vector, is a time average, M is an n x M matrix,
the ith column of M is the ith major EOF, w i EOF M is the number of EOFs used; c is an m-dimensional column vector containing the amplitude or weighting coefficients of the EOF;
determining a coefficient c by performing least square fitting of a series of observed values on the EOFs, and generating a single time point; by calculating a least squares solution of the linear system (2.3);
HMc＝w o (2.3)
the analysis given by (2.1) is a linear combination of the time average plus the EOFs, where the amplitude of the EOFs is determined by (2.3) through the vector c;
the columns of the matrix M are mutually orthogonal, and the columns of the matrix M are normalized to obtain a standard orthogonal matrix
Satisfies the following conditions:
only a few observations are relevant to the number of grid points,
is not orthogonal, restricted by the observation position formally represented by the operator H; defining an observation array H that will generate a reliable analysis from the analytical model (2.1), the optimal array if and only if it satisfies the condition (2.5);
by definition, the condition number for all eigenvalues of the identity matrix I being 1, I is also 1; starting from n observation positions, independently excluding one observation point at a time, and obtaining the rest
The condition numbers are calculated, the observation position with the minimum condition number is eliminated, leaving (n-1) observation positions; until each remaining observation is independently excluded, leaving a desired number of observation positions; the final configuration of the observation positions is taken as the optimal array.
4. The method for optimizing sea-based observation site layout according to claim 2, wherein the step S2 specifically includes:
the optimal array for three-dimensional temperature and salinity observation variables obtained by the EOF reconstruction method is inconsistent in the positions of observation stations at different depths; and dividing similar elements into closely related clusters by using a K central point clustering algorithm, so that different arrays are consistent in the vertical direction aiming at different variables.
5. A sea-based observation network site layout optimization system for implementing the method for sea-based observation network site layout optimization according to any one of claims 1 to 4, wherein the sea-based observation network site layout optimization system comprises:
a single two-dimensional variable optimal station set acquisition module used for respectively carrying out station optimization on single two-dimensional variables based on an EOF reconstruction method to obtain an optimal station set of the variables at different depths
{Station_Temperature|depth＝i，Station_Salinity|depth＝i，Station_Velocity|depth＝i}；
A multivariable optimal station set acquisition module for obtaining a set of optimal stations of the set of optimal stations in the variables at different depths by using a K-center clustering algorithm
{ Station _ Temperature | depth = i, station _ sales | depth = i, and Station _ Velocity | depth = i } as an optimal Station position in a multivariate situation;
and the multiple sets of ocean numerical mode optimal Station position acquisition modules are used for reporting results after different ocean numerical modes are utilized to obtain a set { Station | model = i } of the optimal Station positions, and obtaining the optimal Station positions under the comprehensive condition of the multiple sets of ocean numerical modes by utilizing a K-center clustering method.
6. A program storage medium for receiving user input, the stored computer program causing an electronic device to perform the method for site placement optimization for a sea-based observation network of any one of claims 1 to 4.
7. An information data processing terminal, characterized in that the information data processing terminal comprises a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the method for sea-based observation network site layout optimization according to any one of claims 1 to 4.
8. A sea-based observation network observation station system is characterized in that the sea-based observation network observation station system utilizes the method for optimizing the station position layout of the sea-based observation network according to any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111434536.0A CN114139448B (en) | 2021-11-29 | 2021-11-29 | Method, system, medium, terminal and application for optimizing sea-based observation network station layout |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111434536.0A CN114139448B (en) | 2021-11-29 | 2021-11-29 | Method, system, medium, terminal and application for optimizing sea-based observation network station layout |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114139448A CN114139448A (en) | 2022-03-04 |
| CN114139448B true CN114139448B (en) | 2023-04-18 |
Family
ID=80388969
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111434536.0A Active CN114139448B (en) | 2021-11-29 | 2021-11-29 | Method, system, medium, terminal and application for optimizing sea-based observation network station layout |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114139448B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116644832A (en) * | 2023-04-07 | 2023-08-25 | 自然资源部第二海洋研究所 | Optimized layout determining method for bay water quality monitoring station |
| CN116976509B (en) * | 2023-07-24 | 2024-05-07 | 自然资源部第一海洋研究所 | Quasi-synchronous large-scale multi-station offshore ocean current continuous observation method, system and application |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113297801A (en) * | 2021-06-15 | 2021-08-24 | 哈尔滨工程大学 | Marine environment element prediction method based on STEOF-LSTM |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103336995B (en) * | 2013-04-19 | 2016-04-13 | 国家电网公司 | The construction method of a kind of million kilowatt photovoltaic generation base light simultaneous measurement network |
| CN105718721A (en) * | 2016-01-15 | 2016-06-29 | 中国电力科学研究院 | Regional solar energy resource calculation method |
| CN106485351A (en) * | 2016-09-30 | 2017-03-08 | 浙江大学 | A kind of marine environmental monitoring erect-position layout optimization method |
| CN108981957B (en) * | 2018-05-31 | 2021-01-05 | 西北工业大学 | Underwater temperature field reconstruction method based on self-organizing neural network and empirical orthogonal function |
-
2021
- 2021-11-29 CN CN202111434536.0A patent/CN114139448B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113297801A (en) * | 2021-06-15 | 2021-08-24 | 哈尔滨工程大学 | Marine environment element prediction method based on STEOF-LSTM |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114139448A (en) | 2022-03-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114139448B (en) | Method, system, medium, terminal and application for optimizing sea-based observation network station layout | |
| Lichti et al. | Estimating utilization distributions with kernel versus local convex hull methods | |
| Esteves et al. | Competitive k-means, a new accurate and distributed k-means algorithm for large datasets | |
| Kissling et al. | Spatial autocorrelation and the selection of simultaneous autoregressive models | |
| CN111327377B (en) | Method, device, equipment and storage medium for field intensity prediction | |
| CN113329437A (en) | Wireless network signal propagation path loss prediction method and electronic equipment | |
| CN114677494B (en) | Method, device and equipment for calculating radar detection capability based on subdivision grids | |
| Portilla-Yandún et al. | On the statistical analysis of ocean wave directional spectra | |
| CN108344968A (en) | A kind of orthogonal propagation operator direction-finding method based on multimodal quantum cuckoo search mechanisms | |
| Zio et al. | Optimal spatial sampling schemes for environmental surveys | |
| González‐Abad et al. | Using explainability to inform statistical downscaling based on deep learning beyond standard validation approaches | |
| CN108566256B (en) | Construction method of frequency spectrum map | |
| CN107046448B (en) | MIMO OTA three-dimensional channel space characteristic reproduction method and device | |
| Oh et al. | Explorative navigation of mobile sensor networks using sparse Gaussian processes | |
| CN114880373B (en) | Soil sampling method, system, storage medium and electronic equipment | |
| Sreepathi et al. | Parallel multivariate spatio-temporal clustering of large ecological datasets on hybrid supercomputers | |
| JP7382097B2 (en) | Near-space atmospheric state simulation method and device based on quantum walk | |
| Shen et al. | A sparse manifold learning approach to robust indoor positioning based on Wi-Fi RSS fingerprinting | |
| CN115190587A (en) | WIFI position determination method and device, electronic equipment and storage medium | |
| CN114325877B (en) | Assessment method and device for weather forecast data | |
| Haskard et al. | Modelling non-stationary variance of soil properties by tempering an empirical spectrum | |
| Wang et al. | Spatial automatic subgroup analysis for areal data with repeated measures | |
| CN115685314A (en) | Seismic reservoir physical property parameter prediction method and device | |
| Suraci et al. | A meta-classifier approach for outlier identification in geodetic networks. | |
| Fan et al. | Effect of Spatial Decomposition on the Efficiency of k Nearest Neighbors Search in Spatial Interpolation |
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 |