CN103020423B - The method of output of wind electric field correlation properties is obtained based on copula function - Google Patents

Abstract

The invention discloses a kind of method obtaining output of wind electric field correlation properties based on copula function, the method obtains cumulative distribution function and the cumulative distribution function value of M output of wind electric field according to output of wind electric field sample; The parameter of copula function is obtained by maximum likelihood method; Each copula function is determined again according to parameter; Experience copula functional value is obtained according to cumulative distribution function and cumulative distribution function value; Copula functional value corresponding to cumulative distribution function value is calculated according to each copula function; According to copula functional value and described experience copula functional value acquisition Euclidean distance between the two; Determine that the minimum corresponding copula function of Euclidean distance is described output of wind electric field correlativity; According to the correlation properties obtained after 4 related coefficients between copula function calculating output of wind electric field between output of wind electric field.

Description

The method of output of wind electric field correlation properties is obtained based on copula function

Technical field

The invention belongs to wind-power electricity generation distribution technique field, more specifically, relate to a kind of method obtaining output of wind electric field correlation properties based on copula function.

Background technology

Along with the scale of China's wind energy develops, wind-powered electricity generation continues fast development, and installed capacity increases fast.According to statistics, China increase newly ratio that wind-powered electricity generation installation accounts for the newly-increased installation in the whole world from 2006 rise to less than 10% 2010 49%.By in June, 2012, national grid connected wind power capacity has had 5,258 ten thousand kilowatts.The operation characteristic of the uncertainty of wind-resources and Wind turbines itself makes the output power of wind energy turbine set have intermittence and undulatory property, the access of large-scale wind power certainly will bring difficulty to the safe and stable operation of electric system, the technical barrier of wind-electricity integration, operation difficulty, bottleneck progressively manifest, and wind-electricity integration difficulty becomes a focus.Rationally portray correlation properties between many output of wind electric field and random variation rule, to the precision of prediction improving output of wind electric field, and then improve operation of power networks level, thus reduce the consumption of non-regeneration energy, ensure power system safety and stability, improve power system economy, reduce greenhouse gas emission and be significant.At present, linearly dependent coefficient is only adopted to be described to the correlativity of output of wind electric field.But in many cases, the correlativity of output of wind electric field presents nonlinear relationship, be only difficult to the correlationship between accurate description output of wind electric field with linearly dependent coefficient.And there is no and find that exerting oneself of wind energy turbine set meets some specific distributions, the joint probability distribution function of multiple output of wind electric field is difficult to structure especially, and therefore, how the correlationship described exactly between multiple output of wind electric field is a difficulties.

Summary of the invention

For the defect of prior art, the object of the present invention is to provide a kind of method obtaining output of wind electric field correlation properties based on copula function, when being intended to solve multiple wind energy turbine set joint probability distribution the unknown, the problem of output of wind electric field correlativity is described more accurately from multi-angles such as Spearman rank correlation coefficient, Kendall rank correlation coefficient, upper tail dependence coefficient and lower tail related coefficients.

The invention provides a kind of method obtaining output of wind electric field correlation properties based on copula function, comprise the steps:

S1: the probability density function and the cumulative distribution function that calculate M output of wind electric field according to the historical data sample of output of wind electric field;

S2: the historical data sample of M output of wind electric field is substituted into described cumulative distribution function and obtain corresponding cumulative distribution function value;

S3: described cumulative distribution function value to be substituted into respectively in multiple copula function and to be obtained the parameter of copula function by maximum likelihood method; Each copula function is determined respectively again according to described parameter;

S4: obtain experience copula functional value according to cumulative distribution function and cumulative distribution function value;

S5: calculate copula functional value corresponding to cumulative distribution function value according to each the copula function in step S3;

S6: according to described copula functional value and described experience copula functional value acquisition Euclidean space distance between the two;

S7: using described Euclidean space apart from the copula function of minimum corresponding copula function as description output of wind electric field correlativity;

S8: according to the correlation properties obtained after Spearman rank correlation coefficient, Kendall rank correlation coefficient, upper tail dependence coefficient and the lower tail related coefficient between the copula function calculating output of wind electric field selected in step S7 between described wind energy turbine set.

Further, before step S3, also comprise copula function select step:

S30: select copula function according to cumulative distribution function value.

Further, described step S30 is specially:

S301: obtain frequency histogram according to any two wind energy turbine set cumulative distribution function values;

S302: according to shape and the related coefficient selection copula function of frequency histogram.

Further, step S302 is specially:

When frequency histogram has symmetrical afterbody and tail dependence coefficient is 0, select binary normal state Copula function and Frankcopula function;

When frequency histogram has symmetrical afterbody and tail dependence coefficient is not 0, select t-Copula function;

When the upper tail of frequency histogram is high and lower tail is low, select GumbelCopula function;

When the lower tail of frequency histogram is high and upper tail is low, select ClaytonCopula function.

Further, in step s 8, when for binary normal state copula function, described Kendall rank order correlation coefficient described Spearman rank order correlation coefficient described lower tail related coefficient is λ ^low=0; Described upper tail dependence coefficient is λ ^up=0, ρ is correlation parameter.

Further, in step s 8, when for t-copula function, described Kendall rank order correlation coefficient described Spearman rank order correlation coefficient described lower tail related coefficient is described upper tail dependence coefficient is ρ is correlation parameter, and k is degree of freedom parameter, t _k+1for the t distribution that degree of freedom is k+1.

Further, in step s 8, when for Gumbel-copula function, described Kendall rank order correlation coefficient described lower tail related coefficient is λ ^low=0; Described upper tail dependence coefficient is λ ^up=2-2 ^1/a; A is correlation parameter.

Further, in step s 8, when for Clayton-copula function, described Kendall rank order correlation coefficient described lower tail related coefficient is λ ^low=2 ^-1/a; Described upper tail dependence coefficient is λ ^up=0; A correlation parameter.

Further, in step s 8, when for Frank-copula function, described Kendall rank order correlation coefficient described Spearman rank order correlation coefficient described lower tail related coefficient is λ ^low=0; Described upper tail dependence coefficient is λ ^up=0; A correlation parameter, for debye function, m is the exponent number of debye function, D ₁and D ₂represent 1 rank and 2 rank debye functions respectively, x is argument of function, and y is integration variable.

The present invention adopts the correlativity between the adjacent output of wind electric field in Copula function representation space, the method edge distribution not restriction, can by the marginal distribution function of stochastic variable and the separately research of their dependency structure, avoid this difficult point of direct structure of multiple random variable joint distribution function, non-linear between variable, asymmetry and afterbody correlationship can be caught, thus provide more valuable wind power information for electric system.To the precision of prediction improving output of wind electric field, and then improve operation of power networks level, thus reduce the consumption of non-regeneration energy, ensure power system safety and stability, improve power system economy, reduce greenhouse gas emission and be significant.

Accompanying drawing explanation

Fig. 1 is a kind of method realization flow figure obtaining correlation properties between output of wind electric field based on copula function that the embodiment of the present invention provides;

Fig. 2 is the cumulative distribution function schematic diagram that wind energy turbine set WF1 exerts oneself;

Fig. 3 is the cumulative distribution function schematic diagram that wind energy turbine set WF2 exerts oneself;

Fig. 4 is the frequency histogram representative of the cumulative distribution function that two wind energy turbine set WF1 and WF2 exert oneself;

Binary normal state copula function schematic diagram when Fig. 5 is correlation parameter ρ=0.9697;

Fig. 6 be correlation parameter ρ=0.9697 and degree of freedom parameter k=3.9612 time t-copula function schematic diagram;

Fig. 7 is the experience copula function schematic diagram that two wind energy turbine set WF1 and WF2 exert oneself.

Embodiment

In order to make object of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.

In order to solve the technical matters that correlativity between output of wind electric field is difficult to portray, with reference to shown in Fig. 1, the invention provides a kind of method of portraying correlativity between many output of wind electric field based on copula function, specifically comprising following steps:

(1) determine the probability distribution function of each output of wind electric field, computing method are as follows:

If x _i(i=1,2 ..., N) and be the historical data sample of an output of wind electric field, probability density function f (x) and the cumulative distribution function F (x) at x place, arbitrfary point is calculated according to formula (1)-(4); $f\left(x\right)=\frac{1}{\mathrm{Nh}}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}K\left(\frac{x-x_i}{h}\right)---\left(1\right),$ $f\left(x\right)=\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}K\left(\frac{x-x_i}{h}\right)---\left(2\right),$ $K\left(\frac{x-x_i}{h}\right)=\frac{1}{\sqrt{2\mathrm{\π}}}\exp (-\frac{1}{2}\×{\left(\frac{x-x_i}{h}\right)}^2)---\left(3\right),$ H=1.06Sn ^-0.2(4), in formula, K is called kernel function, and h is called window width, and S is sample standard deviation, and n is total sample number.

Suppose there be M wind energy turbine set, the cumulative probability function of each output of wind electric field can be obtained by above-mentioned calculating, be designated as respectively: (F ₁(x), F ₂(x) ..., F _j(x) ..., F _m(x)).

(2) data sample (X that exerts oneself M wind energy turbine set observed ₁, X ₂..., X _j..., X _m), X _j=(x _j1..., x _jN) ^tsubstitute into wind energy turbine set cumulative distribution function (F respectively ₁(x), F ₂(x) ..., F _j(x) ..., F _m(x)), obtain corresponding cumulative distribution function value, be designated as (U ₁, U ₂..., U _j..., U _m), U _j=(U _j1, U _j2..., U _jn..., U _jN) ^t.

(3) copula function is selected according to cumulative distribution function value, such as: make any two wind energy turbine set cumulative distribution function value (U _i, U _j), U _i=(U _i1, U _i2..., U _in..., U _iN) ^t, U _j=(U _j1, U _j2..., U _jn..., U _jN) ^tfrequency histogram, choose according to the shape of dual histogram and select suitable copula function, specific as follows:

Rule 1: if frequency histogram has symmetrical afterbody and tail dependence coefficient is 0, then select binary normal state Copula function and Frankcopula function;

Rule 2: if frequency histogram has symmetrical afterbody and tail dependence coefficient is not 0, then select t-Copula function;

Rule 3: if frequency histogram resembles in " J " font, upper tail is high, lower tail is low, then select GumbelCopula function;

Rule 4: if frequency histogram resembles in " L " font, lower tail is high, and upper tail is low, then select ClaytonCopula function.

(4) the data obtained (U is utilized ₁, U ₂..., U _j..., U _m), the parameter in the copula function selected by being obtained by maximum likelihood method.Approximating method is as follows:

First the expression formula of binary GumbelCopula function, ClaytonCopula function, Frankcopula function, t-copula function and normal state copula function is provided:

(4.1) GumbelCopula function:

$C_{\mathrm{Gum}}(u,v;a)=\exp \{-{[{(-\log u)}^{\frac{1}{a}}+{(-\log v)}^{\frac{1}{a}}]}^a\},0<a\&le;1\&CenterDot;\&CenterDot;\&CenterDot;\&CenterDot;\&CenterDot;\&CenterDot;\left(5\right),$ A ∈ in formula (0,1] be correlation parameter, as a=1, stochastic variable u, v are independent, when a=0 stochastic variable u, v are tending towards being correlated with completely.

(4.2) ClaytonCopula function:

$C_{\mathrm{cl}}(u,v;a)=\max \{{(u^{-a}+v^{-a}-1)}^{\frac{1}{a}},0\},a>0\&CenterDot;\&CenterDot;\&CenterDot;\&CenterDot;\&CenterDot;\&CenterDot;\left(6\right),$ In formula, a ∈ (0, ∞) is correlation parameter, and when a → 0, stochastic variable u, v trend towards independence, when a → 1 stochastic variable u, v are tending towards being correlated with completely.

(4.3) FrankCopula function:

$C_F(u,v;a)=-\frac{1}{a}\log (1+\frac{(e^{-\mathrm{au}}-1)(e^{-\mathrm{av}}-1)}{e^{-a}-1}),a>0\&CenterDot;\&CenterDot;\&CenterDot;\&CenterDot;\&CenterDot;\&CenterDot;\left(7\right),$ In formula, a is correlation parameter, a ≠ 0, a>0 stochastic variable u, v positive correlation, when a → 0 stochastic variable u, v are tending towards independent, and a<0 stochastic variable u, v negative correlation.

(4.4) t-Copula function:

$C_t(u,v;\mathrm{\&rho;},\mathrm{\&tau;})={\&Integral;}_{-\&infin;}^{T_{\mathrm{\&tau;}}^{-1}\left(u\right)}{\&Integral;}_{-\&infin;}^{T_{\mathrm{\&tau;}}^{-1}\left(v\right)}\frac{1}{2\mathrm{\&pi;}\sqrt{1-{\mathrm{\&rho;}}^2}}{(1+\frac{s^2-2\mathrm{\&rho;st}+t^2}{\mathrm{\&tau;}(1-{\mathrm{\&rho;}}^2)})}^{\frac{-(\mathrm{\&tau;}+2)}{2}}\mathrm{dsdt}\&CenterDot;\&CenterDot;\&CenterDot;\&CenterDot;\&CenterDot;\&CenterDot;\left(8\right),$ Wherein-1≤ρ≤1 is interdependent parameter, and τ is degree of freedom parameter, T _zwith be respectively student t-distribution and inverse function thereof that degree of freedom is τ.

(4.5) normal state Copula function (Gauss Copula function):

wherein-1≤ρ≤1 is interdependent parameter, Φ and Φ ^-1be respectively standardized normal distribution and inverse function thereof.

Wherein (5), (6), (7) are called Archimedes's contiguous function, (8), (9) are called elliptical-type contiguous function.

Suppose that the probability density function of the Copula function treating matching is expressed as c (u ₁, u ₂..., u _n| θ), wherein θ is the parameter needing to estimate: structure likelihood function get the logarithm of likelihood function $\ln L\left(\mathrm{\&theta;}\right)=\underset{i=1}{\overset{n}{\mathrm{\&Pi;}}}\ln f(u_{1,i},u_{2,i},\&CenterDot;\&CenterDot;\&CenterDot;,u_{n,i}|\mathrm{\&theta;});$ Order $\frac{\&PartialD;\ln L}{\&PartialD;\mathrm{\&theta;}}=0;$ Separate the maximum likelihood estimator that likelihood equation obtains parameter θ

(5) experience copula functional value is obtained according to cumulative distribution function and cumulative distribution function value;

Cumulative distribution function is utilized to be (F ₁(x), F ₂(x) ..., F _j(x) ..., F _m(x)) and (U ₁, U ₂..., U _j..., U _m), calculate experience copula functional value, computing method are as follows:

$C_n(u_1,u_2,\&CenterDot;\&CenterDot;\&CenterDot;,u_M)=\frac{1}{N}\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{I}_{F_1(x_{1,i}\&le;u_1)}\&CenterDot;\&CenterDot;\&CenterDot;I_{F_M(x_{M,i}\&le;u_M)}$

Wherein, I _[]for indicative function, work as F _n(x _m,i)≤u _mtime, otherwise

(6) by (U ₁, U ₂..., U _j..., U _m) substitute into the copula function that obtains of matching and experience copula function, the copula function of digital simulation and the Euclidean space distance of experience copula function, select apart from minimum copula function as the copula function describing many output of wind electric field correlativity.Computing method are as follows:

(6.1) the Euclidean distance computing formula of Gauss copula function and experience copula function: $d_{\mathrm{Ga}}^2=\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}{|C_{\mathrm{Ga}}(u_1,u_2,\&CenterDot;\&CenterDot;\&CenterDot;,u_M)-C_n(u_1,u_2,\&CenterDot;\&CenterDot;\&CenterDot;,u_M)|}^2$

(6.2) the Euclidean distance computing formula of t-copula function and experience copula function: $d_t^2=\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}{|C_t(u_1,u_2,\&CenterDot;\&CenterDot;\&CenterDot;,u_M)-C_n(u_1,u_2,\&CenterDot;\&CenterDot;\&CenterDot;,u_M)|}^2$

(6.3) the Euclidean distance computing formula of GumbelCopula function and experience copula function: $d_{\mathrm{Gum}}^2=\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}{|C_{\mathrm{Gum}}(u_1,u_2,\&CenterDot;\&CenterDot;\&CenterDot;,u_M)-C_n(u_1,u_2,\&CenterDot;\&CenterDot;\&CenterDot;,u_M)|}^2$

(6.4) the Euclidean distance computing formula of ClaytonCopula function and experience copula function: $d_{\mathrm{Cla}}^2=\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}{|C_{\mathrm{Cla}}(u_1,u_2,\&CenterDot;\&CenterDot;\&CenterDot;,u_M)-C_n(u_1,u_2,\&CenterDot;\&CenterDot;\&CenterDot;,u_M)|}^2$

(6.5) the Euclidean distance computing formula of Frankcopula function and experience copula function: $d_F^2=\underset{i=1}{\overset{M}{\mathrm{\&Sigma;}}}{|C_F(u_1,u_2,\&CenterDot;\&CenterDot;\&CenterDot;,u_M)-C_n(u_1,u_2,\&CenterDot;\&CenterDot;\&CenterDot;,u_M)|}^2$

(7) utilize the copula function calculated, calculate the Spearman rank correlation coefficient of many output of wind electric field, Kendall rank correlation coefficient, upper afterbody facies relationship coefficient and lower tail related coefficient.

Various different copula function computing method are as follows:

(7.1) normal state copula function: Kendall rank correlation coefficient spearman rank correlation coefficient lower tail related coefficient λ ^low=0; Upper tail dependence coefficient λ ^up=0; In formula, ρ is correlation parameter.

(7.2) t-copula function: Kendall rank correlation coefficient spearman rank correlation coefficient ${\mathrm{\&rho;}}_s=\frac{6\arcsin \frac{\mathrm{\&rho;}}{2}}{\mathrm{\&pi;}};$ Lower tail related coefficient ${\mathrm{\&lambda;}}^{\mathrm{low}}=2-2t_{k+1}\left(\frac{\sqrt{k+1}\sqrt{1-\mathrm{\&rho;}}}{\sqrt{1+\mathrm{\&rho;}}}\right);$ Upper tail dependence coefficient in formula, ρ is correlation parameter, t _k+1for the t distribution that degree of freedom is k+1.

(7.3) Gumbel-copula function: Kendall rank correlation coefficient lower tail related coefficient λ ^low=0; Upper tail dependence coefficient λ ^up=2-2 ^1/a, a correlation parameter.

(7.4) Clayton-copula function: Kendall rank correlation coefficient lower tail related coefficient λ ^low=2 ^-1/a; Upper tail dependence coefficient λ ^up=0, a correlation parameter.

(7.5) Frank-copula function: Kendall rank correlation coefficient spearman rank correlation coefficient lower tail related coefficient λ ^low=0; Upper tail dependence coefficient λ ^up=0, a correlation parameter, for debye function, m is the exponent number of debye function, D ₁and D ₂represent 1 rank and 2 rank debye functions respectively, x is argument of function, and y is integration variable.

The present invention adopts Copula function (contiguous function) to describe correlativity between the adjacent wind energy turbine set in space, the method edge distribution not restriction, can by the marginal distribution function of stochastic variable and the separately research of their dependency structure, avoid this difficult point of direct structure of multiple random variable joint distribution function, non-linear between variable, asymmetry and afterbody correlationship can be caught, thus provide more valuable wind power information for electric system.To the precision of prediction improving output of wind electric field, and then improve operation of power networks level, thus reduce the consumption of non-regeneration energy, ensure power system safety and stability, improve power system economy, reduce greenhouse gas emission and be significant.

In order to a kind of method obtaining output of wind electric field correlation properties based on copula function further illustrating that the embodiment of the present invention provides, with reference to the accompanying drawings and details are as follows in conjunction with instantiation:

The principle of the invention and step for convenience of explanation, embodiment is studied for Texas adjacent two wind energy turbine set history of 1 year observed value (representing with W1 and W2) of exerting oneself, the basic geography information of wind energy turbine set WF1 and WF2 as shown in Table 1:

Wind field

Position

Height above sea level (m)

Wind density

Wind speed

Capacity

			（W/m2）	(m/s)	(MW)
						WF1	（31.19N，102.24W）	850	401.3	7.6	30
WF2	（31.19N，102.21W）	850	419	7.8	31.3
						WF3	（31.23N，102.21W）	850	427.8	8.1	33.6

Table one

(1) distribution function of each output of wind electric field is determined.

Adopt method of the present invention, obtain two output of wind electric field cumulative distribution function F as shown in accompanying drawing 2 and accompanying drawing 3 ₁(x) and F ₂(x).From accompanying drawing, though the distribution function calculated and empirical distribution function incomplete same, but both difference are very small.

(2) the cumulative distribution function U of two output of wind electric field is drawn ₁=F ₁(x) and U ₂=F ₂the frequency histogram of (x), as shown in Figure 4, as can be seen from accompanying drawing 4, two output of wind electric field have symmetrical afterbody, therefore can select the dependency structure that binary normal state Copula function or binary t-Copula function representation wind energy turbine set W1 and wind energy turbine set W2 exert oneself.

(3) method of the present invention is utilized, the parameter of matching binary normal state copula and binary t-Copula function, as shown in Table 2, binary normal state copula function, binary t-Copula function and the experience distribution copula function that matching obtains is as shown in accompanying drawing 5, Fig. 6 and Fig. 7 for result:

Table two

(4) according to the method that the present invention sets forth, calculate the copula function of gained and the theorem in Euclid space distance of experience copula function, thus select best copula function.More known by Euclidean distance duplicate ratio, the Euclidean distance square of binary normal state Copula and experience Copula be 0.2074, binary t-Copula and the Euclidean distance square of experience Copula be 0.2314.Therefore, under Euclidean distance square index instructs, can think binary normal state Copula and experience Copula model can the W1 and W2 of matching wind energy turbine set better exert oneself between correlativity.

Correlation matrix ρ=0.9697 is substituted into normal state copula function obtain describing wind energy turbine set W1 and W2 to exert oneself the function of correlativity:

(5) utilize copula function, try to achieve the joint probability distribution function of many output of wind electric field:

(6) the copula function of gained is utilized, the Spearman rank correlation coefficient of the many output of wind electric field in zoning, Kendall rank correlation coefficient, upper afterbody facies relationship coefficient and lower tail related coefficient; Wherein, Kendall rank correlation coefficient spearman rank correlation coefficient lower tail related coefficient λ ^low=0; Upper tail dependence coefficient λ ^up=0.

Upper tail dependence coefficient and lower tail related coefficient are 0, illustrate that two output of wind electric field have symmetrical afterbody.

Those skilled in the art will readily understand; the foregoing is only preferred embodiment of the present invention; not in order to limit the present invention, all any amendments done within the spirit and principles in the present invention, equivalent replacement and improvement etc., all should be included within protection scope of the present invention.

Claims (6)

1. obtain a method for output of wind electric field correlation properties based on copula function, it is characterized in that, comprise the steps:

S1: the probability density function and the cumulative distribution function that calculate M output of wind electric field according to the historical data sample of output of wind electric field;

S2: the historical data sample of M output of wind electric field is substituted into described cumulative distribution function and obtain corresponding cumulative distribution function value;

S3: described cumulative distribution function value to be substituted into respectively in multiple copula function and to be obtained the parameter of copula function by maximum likelihood method; Each copula function is determined respectively again according to described parameter;

S4: obtain experience copula functional value according to cumulative distribution function and cumulative distribution function value;

S5: calculate copula functional value corresponding to cumulative distribution function value according to each the copula function in step S3;

S6: according to described copula functional value and described experience copula functional value acquisition Euclidean space distance between the two;

S7: using described Euclidean space apart from the copula function of minimum corresponding copula function as description output of wind electric field correlativity;

S8: according to the correlation properties obtained after Spearman rank correlation coefficient, Kendall rank correlation coefficient, upper tail dependence coefficient and the lower tail related coefficient between the copula function calculating output of wind electric field selected in step S7 between described wind energy turbine set;

Before step S3, also comprise copula function select step:

S30: select copula function according to cumulative distribution function value;

Described step S30 is specially:

S301: obtain frequency histogram according to any two wind energy turbine set cumulative distribution function values;

S302: according to shape and the related coefficient selection copula function of frequency histogram;

Step S302 is specially:

When frequency histogram has symmetrical afterbody and tail dependence coefficient is 0, select binary normal state copula function and Frank-copula function;

When frequency histogram has symmetrical afterbody and tail dependence coefficient is not 0, select t-copula function;

When the upper tail of frequency histogram is high and lower tail is low, select Gumbel-copula function;

When the lower tail of frequency histogram is high and upper tail is low, select Clayton-copula function.

2. the method for claim 1, is characterized in that, in step s 8, when for binary normal state copula function, and described Kendall rank order correlation coefficient described Spearman rank order correlation coefficient described lower tail related coefficient is λ ^low=0; Described upper tail dependence coefficient is λ ^up=0, ρ is correlation parameter.

3. the method for claim 1, is characterized in that, in step s 8, when for t-copula function, and described Kendall rank order correlation coefficient described Spearman rank order correlation coefficient ${\mathrm{\&rho;}}_s=\frac{6arcsin\frac{\mathrm{\&rho;}}{2}}{\mathrm{\&pi;}};$ Described lower tail related coefficient is ${\mathrm{\&lambda;}}^{low}=2-2t_{k+1}\left(\frac{\sqrt{k+1}\sqrt{1-\mathrm{\&rho;}}}{\sqrt{1+\mathrm{\&rho;}}}\right);$ Described upper tail dependence coefficient is ρ is correlation parameter, and k is degree of freedom parameter, t _k+1for the t distribution that degree of freedom is k+1.

4. the method for claim 1, is characterized in that, in step s 8, when for Gumbel-copula function, and described Kendall rank order correlation coefficient described lower tail related coefficient is λ ^low=0; Described upper tail dependence coefficient is λ ^up=2-2 ^1/a; A is correlation parameter.

5. the method for claim 1, is characterized in that, in step s 8, when for Clayton-copula function, and described Kendall rank order correlation coefficient described lower tail related coefficient is λ ^low=2 ^-1/a; Described upper tail dependence coefficient is λ ^up=0; A is correlation parameter.

6. the method for claim 1, is characterized in that, in step s 8, when for Frank-copula function, and described Kendall rank order correlation coefficient described Spearman rank order correlation coefficient described lower tail related coefficient is λ ^low=0; Described upper tail dependence coefficient is λ ^up=0; A is correlation parameter, for debye function, m is the exponent number of debye function, D ₁and D ₂represent 1 rank and 2 rank debye functions respectively, x is argument of function, and y is integration variable.