AU2021106032A4 - Method for constructing meteorological element universal function - Google Patents
Method for constructing meteorological element universal function Download PDFInfo
- Publication number
- AU2021106032A4 AU2021106032A4 AU2021106032A AU2021106032A AU2021106032A4 AU 2021106032 A4 AU2021106032 A4 AU 2021106032A4 AU 2021106032 A AU2021106032 A AU 2021106032A AU 2021106032 A AU2021106032 A AU 2021106032A AU 2021106032 A4 AU2021106032 A4 AU 2021106032A4
- Authority
- AU
- Australia
- Prior art keywords
- function
- regression model
- new
- function group
- value
- 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
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/047—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators the criterion being a time optimal performance criterion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Mathematical Analysis (AREA)
- Environmental & Geological Engineering (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Automation & Control Theory (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Computational Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Ecology (AREA)
- Environmental Sciences (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Atmospheric Sciences (AREA)
- Operations Research (AREA)
- Probability & Statistics with Applications (AREA)
- Biodiversity & Conservation Biology (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Complex Calculations (AREA)
Abstract
A method for constructing a meteorological element universal function includes
filtering a basis function group; linearizing the basis function group to obtain a linear
function group; building a training set for an evaporation duct area; using the training set
5 to train the linear function group to obtain a regression model according to a LS-SVM
algorithm; and according to the regression model, solving an empirical coefficient of the
basis function group to obtain the new meteorological element universal function. After
linearizing the basis function group, the linear kernel-based LS-SVM algorithm is used to
train the linear function using the training set to obtain the regression model, reducing the
10 amount of calculation in the solution, and improving the accuracy of evaporation duct
monitoring and prediction.
1-7
1/5
101
filtering a basis function group
linearizing the basis function group to obtain linear 102
function group
103
building a training set for an evaporation duct area
using the training set to train the linear function group - 104
to obtain a regression model according to a LS-SVM
algorithm
107
testing the regression model
according to the regression model, solving an empiical 105
coefficient of the basis function group to obtain the new
meteorological eement universal function
108
optimizing a parameterof LS-SVM through the mind
evolution algorithm
FIG. 1
Description
1/5
101 filtering a basis function group
linearizing the basis function group to obtain linear 102
function group
103 building a training set for an evaporation duct area
using the training set to train the linear function group - 104 to obtain a regression model according to a LS-SVM algorithm 107
testing the regression model
according to the regression model, solving an empiical 105 coefficient of the basis function group to obtain the new meteorological eement universal function
108 optimizing a parameterof LS-SVM through the mind evolution algorithm
FIG. 1
[0001] The present invention relates to the technical field of marine meteorological detection, specially relates to a method for constructing a meteorological element universal function.
[0002] Evaporation duct is a low-altitude atmospheric duct at sea (generally tens of meters in thickness). Electromagnetic waves of a certain frequency can propagate forward along the waveguide area with very low loss, and the efficiency of electromagnetic radiation sources breaks through the limit of the curvature of the earth, and the detection range or communication distance is significantly increased. Therefore, accurate monitoring and forecasting of the evaporation duct is of great significance to improving the performance of electronic equipment such as radar and communication. At present, the monitoring and forecasting of evaporation duct is a research hotspot at home and abroad, and the meteorological element universal function scheme is a key factor that affects the accuracy of the evaporation duct monitoring and forecasting methods.
[0003] The existing meteorological element universal function scheme is based on a large amount of measured atmospheric data, which has better universality. The influence of atmospheric stratification in the surface layer on the universal function scheme can be
described by the stability parameter . When takes a different value, the factors
affecting atmospheric turbulence will change, and the form of the universal function scheme will also change to a certain extent. The initial research on the meteorological
element universal function scheme concentrated on the -1<g <1 interval. For example,
a series of schemes proposed by Webb, Dyer, Businger, Hogstrom, etc., among which the most classic is the BD74 scheme proposed by Businger and Dyer, which is still widely used. In the study of momentum flux near the surface, Mahrt et al. found that the calculation
effect of BD74 and other schemes is not ideal under the strong stable condition of g >1.
As a result, a large number of researches on the meteorological element universal function schemes under strong stable conditions have been launched. The representative schemes include the HDB88 scheme designed by Holtslag et al., the BH91 scheme proposed by Beljaars et al. after the HDB88 model is revised, as well as Cheng and Brutsaer' s CB05
scheme of wind speed profile and temperature profile design under strong stable conditions.
;< -1 belongs to the free convection interval. There are two main ways to improve the
universal function scheme. One is to modify the index based on the BD74 scheme, and its representative is the new scheme designed by Carl et al. The other is to interpolate the scheme designed by Carl et al. and the BD74 scheme to obtain better robustness. Its representatives include the new universal function scheme proposed by Grachev et al. and the optimized COARE 3.0 algorithm proposed by Fairall et al, and AT2005 scheme designed by Akylas and Tombrou. Wilson also proposed a meteorological element universal function scheme under unstable conditions, imitating other statistics in the atmosphere (hereinafter referred to as the WS2000 scheme).
[0004] However, the application conditions of the various meteorological element universal function schemes are different. In order to further meet the requirements of the marine evaporation duct area guarantee, the present invention designs a new universal function construction method based on the existing meteorological element universal function scheme, which effectively improve the accuracy of marine evaporation duct monitoring and forecasting.
[0005] In view of the above technical problems in the prior art, the present invention provides a new type of meteorological element universal function construction method and construction system. The constructed new universal function can effectively improve the accuracy of evaporation duct monitoring and forecasting.
[0006] A method for constructing a meteorological element universal function, wherein, the method for constructing a meteorological element universal function includes: filtering a basis function group; linearizing the basis function group to obtain a linear function group; building a training set for an evaporation duct area; using the training set to train the linear function group to obtain a regression model according to a LS-SVM algorithm; and according to the regression model, solving an empirical coefficient of the basis function group to obtain the new meteorological element universal function.
[0007] Preferably, the method for constructing the new meteorological element universal function, comprising a method for testing the regression model: building a test set; testing the regression model or the new meteorological element universal function through the test set to obtain a test value; and calculating a root mean square error according to the test value and a corresponding real value in the test set, and the root mean square error is required to be less than a first threshold value.
[0008] Preferably, filtering the basis function group from following function schemes: BD74, BH91, WS2000, CG05 and AT2005.
[0009] Preferably, the basis function group includes a WS2000 function scheme:
Iq, Ph h )= (4)=(-y 1
Ph ()=1 qh 1 /3j / 2 -1/2)
[0010] comprising, 7, h and h re empirical coefficients, P()is a wind
speed universal function, is a Mourning-Obukhov parameters, m is a wind speed, and h is a temperature;
[0011] the linear function group is as following:
ryin 1=1-(pq,({) =y Y.12 =9, I - ((-.=),n 2/3
Y.2 = P. fl {> 0 Yh2 (P 1 =6h;(3)
[0012] comprising, x= and Y-1 are linear functions of Pm(;), Yhl is the
linearized function of h
[0013] Preferably, a method for solving the WS2000 function scheme, comprising: collecting data through a meteorological gradiometer; normalizing collected data, and building the training set and the test set; training the linear function group through the training set and obtaining the regression model according to a LS-SVM algorithm with a linear kernel and a preset penalty factor coefficient; predicting the test set by the regression model and calculating the root mean square error between a predicted value and a true value; evaluating the regression model by the root mean square error; and according to an evaluated regression model, solving the empirical coefficient of the WS2000 function scheme, and obtaining the new meteorological element universal function.
[0014] Preferably, the method for constructing the new meteorological element universal function includes a method of improving the new meteorological element universal function through a mind evolution algorithm: optimizing a parameter of LS-SVM through the mind evolution algorithm, and the parameter include a penalty factor coefficient.
[0015] Preferably, a method for optimizing the parameter of LS-SVM through the mind evolution algorithm, comprising:
[0016] setting an initial value, an upper limit and a lower limit of the penalty factor coefficient; according to the initial value, the upper limit and the lower limit, randomly generating a certain number of penalty factor coefficients; coding the penalty factor coefficients as individuals; taking the individual as the center, generating multiple new individuals, and letting the individuals and the new individuals form a subgroup; building a score function for the individuals; iterating the subgroup according to the mind evolution algorithm, and obtaining a subgroup with the highest score and a subgroup score; and decoding a individual with the highest score in the subgroup with the highest score to obtain an optimal penalty factor coefficient.
[0017] Preferably, a method for building an individual's score function, comprising:
[0018] decoding the individual and obtaining the penalty factor coefficient; substituting a decoded penalty factor coefficient into the LS-SVM algorithm, and using the training set to train the linear function group to obtain the regression model; using the training set to train the linear function group to obtain the regression model; using the regression model to predict the test set and obtaining the predicted value; calculating a mean square error based on the predicted value and the true value of the test set; and using a reciprocal of the mean square error as a value of a score function.
[0019] Preferably, through a binary coding, the penalty factor coefficient is coded as an individual, and an individual particle value can be expressed as the following:
.. at,} S,={tal,a2,r**, (5)
[0020] comprising: a,,(i=1,2,-,1) is a binary code string of the penalty factor coefficient, St is the value of a t-th generation particle, a length 1 of St depends on a parameter accuracy;
[0021] taking the individual as the center, generating new individuals:
X(1: num)= center(1: num)+0.5x (rand(1,num)x 2 -1) (7)
[0022] comprising: num is a length of a parameter encoding, X(1: num) is a new
individual, X(1: num) is a center individual,and rand(1,num) used to generate 1 xnum
matrix in a interval [0,1].
[0023] converting the new individual to the interval [0,1] by formula (8):
S(x1 )= 1 + 1e-'(8)
[0024] comprising: x, is a decimal value of a i-th bit string in a new individual X,
S(x) is a converted value, and then the binary individual is generated by a formula (9):
B 1, random S(x,) | 0, random>S(J (9)
[0025] comprising: Bi is a updated binary code value of Xi, a rand function generates
random numbers in the interval [0,1].
[0026] A construction system of the method for constructing the new meteorological element universal function, wherein, the new meteorological element universal function includes a filtering module, a linearization module, a training set selection module, a training module and an expansion module;
[0027] the screening module is used for screening basis function groups; the linearization module is used to linearize the basis function group to obtain the linear function group; the training set selection module is used to establish a training set of a evaporation waveguide region; the training module is used to train the linear function group based on the LS-SVM algorithm using the training set to obtain a regression model; and the expansion module is used to solve the empirical coefficients of the basis function group according to the regression model to obtain the new meteorological element universal function.
[0028] Compared with the prior art, the beneficial effects of the present invention are: after linearizing the basis function group, the linear kernel-based LS-SVM algorithm is used to train the linear function using the training set to obtain the regression model.
Solving the empirical coefficients of the basis function group according to the linearized relationship and the regression model, so as to solve the new meteorological element universal function, simplifying the process of solving the new meteorological element universal function, reducing the amount of calculation in the solution, and improving the accuracy of evaporation duct monitoring and forecasting.
[0029] FIG. 1 is a flowchart of the method for constructing a meteorological element universal function of the present invention;
[0030] FIG. 2 is a flow chart of the method for optimizing the parameters of LS-SVM through the mind evolution algorithm;
[0031] FIG. 3 is a graph of the convergence change of the initial high score subgroup;
[0032] FIG. 4 is the initial failure sub-group convergence change graph;
[0033] FIG. 5 is the logical block diagram of the construction system of the present invention.
[0034] In order to make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be described clearly and completely in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.
[0035] The present invention will be described in further detail below in conjunction with the accompanying drawings:
[0036] a method for constructing a meteorological element universal function, as shown in FIG. 1, comprising:
[0037] S10: filtering a basis function group. In a specific embodiment, the basis function group is selected from the following function schemes: BD74, BH91, WS2000, CG05, and AT2005, but not limited to this, where the function scheme is the prior art and
K1 will not be repeated in this application.
[0038] S102: linearizing the basis function group to obtain a linear function group. The existing basis functions are nonlinear. In order to facilitate the calculation of coefficients, the expressions are uniformly transformed into linear.
[0039] S103: building a training set for an evaporation duct area. The meteorological data of the marine evaporation duct area can be collected through the meteorological gradiometer, which is used to construct the training set and the test set.
[0040] S104: using the training set to train the linear function group to obtain a regression model according to a LS-SVM algorithm. In a specific embodiment, the linear kernel
k(x,xI)= xix is selected for solving. Least squares support vector machine (LS-SVM)
is an improved type of SVM algorithm, which uses equality constraints instead of inequality constraints, avoiding solving the quadratic rule problem in the dual space in the SVM, and has the convenience of solving.
[0041] S105: according to the regression model, solving an empirical coefficient of the basis function group to obtain the new meteorological element universal function.
[0042] After linearizing the basis function group, the linear kernel-based LS-SVM algorithm is used to train the linear function using the training set to obtain the regression model. Solving the empirical coefficients of the basis function group according to the linearized relationship and the regression model, so as to solve the new meteorological element universal function, simplifying the process of solving the new meteorological element universal function, reducing the amount of calculation in the solution, and improving the accuracy of evaporation duct monitoring and forecasting.
[0043] S107: testing the regression model. Methods of testing regression model, comprising: Sil: building a test set. S112: testing the regression model or the new meteorological element universal function through the test set to obtain the test value. It should be pointed out that the regression model and the new meteorological element universal function have different forms and different parameters, and their test sets should be adjusted accordingly. S113: according to the test value and the corresponding real value in the test set, the root mean square error is calculated, and the root mean square error is required to be less than the first threshold value. In a specific embodiment, the first threshold value is set to 5 meters.
[0044] S108: optimizing the parameters of LS-SVM through the Mind Evolutionary
Algorithm (MEA) to improve the new meteorological element universal function. Among them, the parameters of LS-SVM include the penalty factor coefficient C, but it is not
limited to this, other parameters can also be optimized, such as the parameter y. MEA is a
new type of optimization algorithm by imitating the process of human thought evolution. The MEA inherits the "group" and "evolution" ideas of genetic algorithm. Aiming at the shortcomings of genetic algorithm (GA) such as slow convergence speed and lack of orientation in search, the algorithm proposes new operation operators-"convergence" and "alienation" to replace the crossover and mutation operators in genetic algorithm.
[0045] As shown in FIG. 2, methods to optimize the parameters of LS-SVM are as following:
[0046] S201: setting an initial value, an upper limit and a lower limit of the penalty factor coefficient. In a specific embodiment, the initial parameters of MEA are set as: 100 individuals, 5 subgroups with high scores, 5 subgroups with failure, 10 iterations, the upper limit of parameter C is 1000, and the lower limit is 2.
[0047] S202: according to the initial value, the upper limit and the lower limit, randomly generating a certain number of penalty factor coefficients.
[0048] S203: coding the penalty factor coefficients as individuals. Through a binary coding, the penalty factor coefficient is coded as an individual, and an individual particle value can be expressed as the following:
S, ={a,a 2 1 ,'' atr} (5)
[0049] comprising: ai(i=1 ,l) is a binary code string of the penalty factor coefficient, St is the value of a t-th generation particle, a length 1 of St depends on a parameter accuracy.
[0050] S204: taking the individual as the center, multiple new individuals are generated, and the individuals and the new individuals form subgroup, and multiple subgroup are obtained. Taking the individual as the center, generating new individuals:
X(1: num)= center(1: num)+0.5 x (rand(1,num)x 2-1) (7)
[0051] comprising: num is a length of a parameter encoding, X(1:num) is a new
individual, X(1: num) is a center individual,and rand(1,num) used to generate 1 xnum
matrix in a interval [0,1].
[0052] converting the new individual to the interval [0,1] by a formula (8):
S(x1 )= 1 1+ e-' (8)
[0053] comprising: x, is a decimal value of a i-th bit string in a new individual X,
S(x) is a converted value, and then the binary individual is generated by a formula (9):
B 1, random) S(x,) | 0, random>S(J (9)
[0054] comprising: Bi is a updated binary code value of Xi, a rand function generates
random numbers in the interval [0,1].
[0055] S205: building an individual's scoring function.
[0056] S206: iterating the subgroups according to the mind evolution algorithm to obtain the subgroup with the highest score and its score. Iteration includes convergence operations and alienation operations. Convergence enables each sub-group to compare individual scores, and the highest score is used as the sub-group score. At this time, the sub-group matures and is displayed on the local bulletin board. Alienation makes the subgroup with high scores replaced by the failed subgroup with higher scores, and the replaced subgroup is disbanded and a new failed subgroup is generated. After multiple iterations, get the best subgroup and its score, that is, get the highest molecular group.
[0057] FIG. 3 shows the changes in the scores of the initial high score subgroup with the number of iterations, and FIG. 4 shows the changes in the scores of the initial failed subgroup with the number of iterations, it can be seen: the scores of subgroups no longer change after several convergence operations. There are some sub-populations that do not perform convergence operations, such as 1, 3, and 5 in the high-scoring subgroup and 2 in the failing subgroup because there are no higher-scoring individuals around their centers. When the subgroup is mature, the 1, 3, and 5 of the failed subgroup have higher scores than the 3, 5 of the high score subgroup, so the next step is to perform the alienation operation.
[0058] S207: In the molecular group with the highest score, the individual with the highest score is decoded to obtain the optimal penalty factor coefficient, which can be decoded according to formula 6.
[0059] In S205, the method of constructing an individual's score function includes:
[0060] S210: decoding the individual to obtain the penalty factor coefficient C.
[0061] The formula for decoding is as follows: C-C. C4 dec (S) = C. +Cma 'i x2' (6) -1
[0062] Comprising, Cj is the decimal value of C, that is, the decoded value of the
penalty factor coefficient C, Cmax and Cminare the upper and lower limits of C,
respectively, and dec(S,) is the decimal value corresponding to St.
[0063] S211: Substituting the decoded penalty factor coefficient into the LS-SVM algorithm, and use the training set to train the linear function group to obtain the regression model.
[0064] S212: using the regression model to predict the test set to obtain a predicted value.
[0065] S213: calculating the mean square error according to the predicted value and the true value of the test set. The smaller the mean square error, the higher the accuracy of the predicted value.
[0066] S214: using the reciprocal of the mean square error as the value of the score function, that is, the score. The larger the score, the smaller the mean square error and the higher the accuracy of the predicted value.
[0067] A construction system of the method for constructing the new meteorological element universal function is provided. As shown in FIG. 5, it includes a filtering module 1, a linearization module 2, a training set selection module 3, a training module 4, and an expansion module 5.
[0068] The filtering module 1 is used for screening the basis function group; the linearization module 2 is used to linearize the basis function group to obtain the linear function group; the training set selection module 3 is used to establish a training set of a evaporation waveguide region; the training module 4 is used to train the linear function group based on the LS-SVM algorithm using the training set to obtain a regression model; and the expansion module 5 is used to solve the empirical coefficients of the basis function group according to the regression model to obtain the new meteorological element universal function.
[0069] The construction system of the method for constructing the new meteorological element universal function may also include a test module 6 and a parameter optimization module 7, the test module is used to test the regression model, and the parameter optimization module is used to optimize the parameters of the LS-SVM through the Mind
11)
Evolutionary Algorithm (MEA).
[0070] Take the WS2000 function scheme as the basis function group for description, and the WS2000 function scheme is:
4 ()=+p
+~ >0 (Ph + h'
[0071] comprising, ., Yh Pm and h re empirical coefficients, P isawind
speed universal function, is a Mourning-Obukhov parameters, m is a wind speed, and h is a temperature.
[0072] the linear function group is as following:
| 1yh r h. ( )2 1 = h. -n ())2< 1(( Yhl = Ph =1VhX(2)
Y.n2 =,n(P. ->n 0 Yh 2 =(3)
[0073] comprising, x and Yn1 are linear functions of Pm(), Yhl is the
linearized function of (h
[0074] The method for solving the WS2000 function scheme, comprising:
[0075] S301: collecting data through a meteorological gradiometer. Specifically, collecting meteorological data in the marine evaporation duct area.
[0076] S302: normalizing collected data, and building the training set and the test set. In a specific embodiment, 20% of the randomly selected data is used to establish the test set, and 80% is used to establish the training set. The normalized mapping calculation method is shown in the following formula:
f:x ->== X"" Xm - Xmin (4)
[0077] Comprising, x is the data value, xmi is the minimum value, and xmax is the maximum value, and f : x -> y is the result of the normalized mapping.
1 1
[0078] S303: based on the LS-SVM algorithm with the linear kernel selected and the penalty factor coefficient preset, the linear function group is trained using the training set to obtain the regression model, that is, formula 2 and formula 3 are solved.
[0079] The penalty factor coefficient can be set to a fixed value, for example, C=500, or the optimal value of the penalty factor coefficient can be calculated by the LS-SVM parameter optimization method of the Mind Evolution Algorithm of the present invention.
[0080] S304: predicting the test set through the regression model, and calculating the root mean square error between the predicted value and the true value. The test set contains real values.
[0081] S305: evaluating the regression model through the root mean square error. In a specific embodiment, the root mean square error of control is not higher than 5m.
[0082] S306: according to the evaluated regression model, the empirical coefficient of the WS2000 function scheme is solved to obtain the new meteorological element universal function. In a specific embodiment, the empirical coefficient y, =-0.5081, yh 1 1.1437,
#, =14.6342, #h =24.545.
[0083] The above embodiments are only preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
11)
Claims (5)
- What is claimed is: 1. A method for constructing a meteorological element universal function, wherein, the method for constructing the new meteorological element universal function includes: filtering a basis function group; linearizing the basis function group to obtain a linear function group; building a training set for an evaporation duct area; using the training set to train the linear function group to obtain a regression model according to a LS-SVM algorithm; and according to the regression model, solving an empirical coefficient of the basis function group to obtain the new meteorological element universal function.
- 2. The method for constructing the new meteorological element universal function according to claim 1, wherein, comprising a method for testing the regression model: building a test set; testing the regression model or the new meteorological element universal function through the test set to obtain a test value; and calculating a root mean square error according to the test value and a corresponding real value in the test set, and the root mean square error is required to be less than a first threshold value.
- 3. The method for constructing the new meteorological element universal function according to claim 1, wherein, filtering the basis function group from following function schemes: BD74, BH91, WS2000, CG05 and AT2005, and the basis function group includes a WS2000 function scheme:(h h )= 7" 2/3)( ({) =1+p {(h ()=1 qh (1)comprising, 'i' , P. and are empirical coefficients, a 'n is a windspeed universal function, is a Mourning-Obukhov parameters, m is a wind speed, and h is a temperature; the linear function group is as following:{< 0 Yh 1 (hhX (2)1 2Y.2=() (. -1 = ; E>0 Yh 2 =(3)comprising, = 2 3 and are linear functions of "( ,Yhl is a linearizedfunction of (h ; and a method for solving the WS2000 function scheme, comprising: collecting data through a meteorological gradiometer; normalizing collected data, and building the training set and the test set; training the linear function group through the training set and obtaining the regression model according to a LS-SVM algorithm with a linear kernel and a preset penalty factor coefficient; predicting the test set by the regression model and calculating the root mean square error between a predicted value and a true value; evaluating the regression model by the root mean square error; according to an evaluated regression model, solving the empirical coefficient of the WS2000 function scheme, and obtaining the new meteorological element universal function; and the method for constructing the new meteorological element universal function also includes a method of improving the new meteorological element universal function through a mind evolution algorithm: optimizing a parameter of LS-SVM through the mind evolution algorithm, and the parameter include a penalty factor coefficient.
- 4. The method for constructing the new meteorological element universal function according to claim 3, wherein, a method for optimizing the parameter of LS-SVM through the mind evolution algorithm, comprising: setting an initial value, an upper limit and a lower limit of the penalty factor coefficient; according to the initial value, the upper limit and the lower limit, randomly generating a certain number of penalty factor coefficients; coding the penalty factor coefficients as individuals; taking the individuals as the center, generating multiple new individuals, and letting the individuals and new individuals form a subgroup; building a score function for the individuals; iterating the subgroup according to the mind evolution algorithm, and obtaining a1 / subgroup with the highest score and a subgroup score; decoding a individual with the highest score in the subgroup with the highest score to obtain an optimal penalty factor coefficient; and a method for building an individual's score function, comprising: decoding the individual and obtaining the penalty factor coefficient; substituting a decoded penalty factor coefficient into the LS-SVM algorithm, and using the training set to train the linear function group to obtain the regression model; using the regression model to predict the test set and obtaining the predicted value; calculating a mean square error based on the predicted value and the true value of the test set; and using a reciprocal of the mean square error as a value of a score function.
- 5. The method for constructing the new meteorological element universal function according to claim 1, wherein, through a binary coding, the penalty factor coefficient is coded as an individual, and an individual particle value can be expressed as the following:S, ={aaa 2,,. .. ,a ,} (5)comprising: at(i=1,2,- )is a binary code string of the penalty factor coefficient, St is the value of a t-th generation particle, a length 1 of St depends on a parameter accuracy; taking the individual as the center, generating new individuals:X(1: num)= center(1: num)+0.5x (rand(1,num)x 2 -1) (7)comprising: num is a length of a parameter encoding, X(1: num) is a new individual,X(1: num) is a center individual,and rand(1,num)used to generate 1 xnum matrix in ainterval [0,1]; converting the new individual to the interval [0,1] by a formula (8):_ S(xi)= I+ (8) comprising: xi is a decimal value of a i-th bit string in a new individual X, S(x)is a converted value, and then the binary individual is generated by a formula (9):B 1, random S(x,) j 0, random>S(x (9)comprising: Bi is a updated binary code value of , a randO function generates random numbers in the interval [0,1];and a construction system of the method for constructing the new meteorological element universal function according to claim 1, wherein, the new meteorological element universal function includes a filtering module, a linearization module, a training set selection module, a training module and an expansion module, the screening module is used for screening the basis function group; the linearization module is used to linearize the basis function group to obtain the linear function group; the training set selection module is used to establish a training set of a evaporation waveguide region; the training module is used to train the linear function group based on the LS-SVM algorithm using the training set to obtain a regression model; and the expansion module is used to solve the empirical coefficients of the basis function group according to the regression model to obtain the new meteorological element universal function.1/I
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| AU2021106032A AU2021106032A4 (en) | 2021-08-19 | 2021-08-19 | Method for constructing meteorological element universal function |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| AU2021106032A AU2021106032A4 (en) | 2021-08-19 | 2021-08-19 | Method for constructing meteorological element universal function |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| AU2021106032A4 true AU2021106032A4 (en) | 2021-10-28 |
Family
ID=78179633
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| AU2021106032A Active AU2021106032A4 (en) | 2021-08-19 | 2021-08-19 | Method for constructing meteorological element universal function |
Country Status (1)
| Country | Link |
|---|---|
| AU (1) | AU2021106032A4 (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117421601A (en) * | 2023-12-19 | 2024-01-19 | 山东省科学院海洋仪器仪表研究所 | Sea surface evaporation waveguide near-future rapid forecasting method |
-
2021
- 2021-08-19 AU AU2021106032A patent/AU2021106032A4/en active Active
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117421601A (en) * | 2023-12-19 | 2024-01-19 | 山东省科学院海洋仪器仪表研究所 | Sea surface evaporation waveguide near-future rapid forecasting method |
| CN117421601B (en) * | 2023-12-19 | 2024-03-01 | 山东省科学院海洋仪器仪表研究所 | Sea surface evaporation waveguide near-future rapid forecasting method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109617065B (en) | Power transmission network line planning method of power system considering mass operation scenes | |
| CN106503839A (en) | A kind of marine wind electric field annular current collection network hierarchy planing method | |
| Tao et al. | Nonuniform wind farm layout optimization: A state-of-the-art review | |
| Ye et al. | Combined approach for short-term wind power forecasting based on wave division and Seq2Seq model using deep learning | |
| CN110263998B (en) | Double-layer correction method for multisource numerical weather forecast set | |
| CN105512766A (en) | Wind power plant power predication method | |
| Liang et al. | A combined model based on CEEMDAN, permutation entropy, gated recurrent unit network, and an improved bat algorithm for wind speed forecasting | |
| AU2021106032A4 (en) | Method for constructing meteorological element universal function | |
| Jadidoleslam et al. | Application of shuffled frog leaping algorithm to long term generation expansion planning | |
| CN104036356A (en) | Method for predicting future operating state of power grid by using fractal algorithm | |
| CN114021483A (en) | Ultra-short-term wind power prediction method based on time domain characteristics and XGboost | |
| Chen et al. | Research on wind power prediction method based on convolutional neural network and genetic algorithm | |
| Lin et al. | Scenario generation and reduction methods for power flow examination of transmission expansion planning | |
| Xia et al. | A new multi‐objective hybrid optimization algorithm for wind‐thermal dynamic economic emission power dispatch | |
| Ruan et al. | On vulnerability of renewable energy forecasting: Adversarial learning attacks | |
| CN116415708B (en) | Power grid robust planning method considering confidence level | |
| CN116128211A (en) | Wind-light-water combined short-term optimization scheduling method based on wind-light uncertainty prediction scene | |
| CN115965134A (en) | Regional power grid wind power generation power prediction optimization method | |
| CN115526671A (en) | New energy power station site selection method based on improved analytic hierarchy process | |
| Sun et al. | Optimization Design of Wind Turbine Blade Based on an Improved Particle Swarm Optimization Algorithm Combined with Non‐Gaussian Distribution | |
| Kou et al. | Generation prediction of ultra-short-term wind farm based on quantum genetic algorithm and fuzzy neural network | |
| Xue et al. | Typical transmission section searching method considering geographical attributes for large power grids | |
| Shen | Short-term traffic flow prediction based on harmony search algorithm optimized wavelet neural network | |
| CN104991885B (en) | A kind of method for generating multiple output of wind electric field scenes | |
| Ma et al. | Research on time series data of renewable energy output based on ga-lstm model of multi-source data fusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| FGI | Letters patent sealed or granted (innovation patent) | ||
| PC | Assignment registered |
Owner name: PLA NAVAL ENGINEERING UNIVERSITY Free format text: FORMER OWNER(S): TIAN, BIN; CHA, HAO; LIANG, XIANG; DENG, DONGLI; BO, WENBO |