Multivariable flood design standard value calculation method and system based on combined relationship curve

Abstract

This invention discloses a method and system for calculating the design standard value of multivariate floods based on a combination relationship curve. First, it collects multi-year hydrological data upstream of the proposed reservoir and compiles data series of flood peaks and flood volumes for each year and time period. Then, it selects time period flood volumes to form a multivariate extreme value series. Next, it performs variability detection on the selected multivariate extreme value series and obtains the variability detection results. This invention achieves the function of diagnosing the variability of multivariate series and can construct a multivariate dynamic joint distribution model for multivariate series with variability. Then, it establishes a combination relationship curve between the primary and secondary variables using the conditional most probable combination method and obtains the design standard value of the corresponding secondary variable by querying the combination relationship curve based on the design value of the primary variable under inconsistency. Thus, it derives the multivariate flood design standard value for a given return period under inconsistency. This method not only has high calculation efficiency but also provides relatively accurate results, making it suitable for widespread promotion and use.

Description

Technical Field

[0001] The invention relates to the technical field of flood design standard value calculation, and in particular to a multivariable flood design standard value calculation method and system based on a combined relationship curve. Background Art

[0002] Design flood is an important basis for determining the scale of water conservancy project construction and for formulating management and operation strategies. The "Specifications for Calculation of Design Floods for Water Conservancy and Hydropower Projects" clearly states that the prerequisite for the application of the current hydrological frequency analysis and calculation method is that the flood extreme value series must meet consistency, that is, the hydrological extreme value series cannot have trend or jump variations. However, with global climate change and the increasing impact of human activities, especially the increasing frequency of large-scale human economic activities in recent years, the density of water conservancy projects, agriculture and urban infrastructure has continued to increase, changing the originally relatively stable underlying surface conditions of the basin, thereby changing the original runoff law and the spatial and temporal distribution of floods.

[0003] At present, the background of extreme hydrological events and historical hydrological extreme events are not the same, and the occurrence conditions and breeding environment of hydrological processes in the future period are also different. At the same time, the frequency and intensity characteristics of hydrological extreme events may also change accordingly. In recent years, many scientific research results at home and abroad have also shown that climate anomalies have occurred in many regions of the world, resulting in frequent flood and drought disasters and concurrent phenomena. This shows that the frequency and intensity of extreme climate have changed, which may further lead to the destruction of the consistency of the hydrological extreme value series, thus shaking the current traditional hydrological frequency analysis calculation that the flood extreme value series must meet the premise of consistency. For flood series that do not meet the consistency requirements, the traditional design flood calculation method is in principle no longer applicable. Since the hydrological event process usually contains multiple attribute characteristics, such as a flood process containing flood peaks and flood volume characteristics in different periods, it is usually difficult to describe the true characteristics of the hydrological event process using a single hydrological variable; therefore, it is necessary to design a multivariate flood design standard value calculation method and system based on a combined relationship curve. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art, in order to better and effectively solve the problem that under the set design standard recurrence period, the combined design value deduced based on the existing multivariate inconsistency frequency analysis method changes with time and cannot obtain a uniquely determined design value to guide the actual engineering, a multivariate flood design standard value calculation method and system based on a combined relationship curve are provided, which realizes the function of diagnosing the variability of a multivariate series, and can construct a multivariate dynamic joint distribution model for a multivariate series that has mutated, and then establishes a combined relationship curve between the primary and secondary variables through the conditional most likely combination method, and queries the combined relationship curve based on the design value of the primary variable under inconsistency to obtain the corresponding secondary variable design standard value, thereby deducing the multivariate flood design standard value for a given recurrence period under inconsistency, which not only has high calculation efficiency, but also has relatively accurate calculation results.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] The method and system for calculating multivariable flood design standard values ​​based on combined relationship curves include the following steps:

[0007] Step A, collect multi-year hydrological data of the upstream of the proposed reservoir design object, sort out the data series of flood peaks in each year and flood volumes in different periods, and then select flood volumes in different periods to form a multivariate extreme value series;

[0008] Step B, performing variability detection on the selected multivariate extreme value series to obtain variability detection results;

[0009] Step C, constructing a single variable variable parameter marginal distribution model under inconsistency based on the variability detection results, and then using the dynamic Copula function to construct a multivariate dynamic joint distribution model under inconsistency conditions;

[0010] Step D, establishing a combination relationship curve between primary and secondary variables using a conditional combination method based on a multivariate dynamic joint distribution model under inconsistent conditions;

[0011] Step E: Calculate the design value of the main variable under inconsistency using the equal reliability method, and then obtain the multivariate flood design standard value under inconsistency based on the combined relationship curve between the main and secondary variables.

[0012] The aforementioned multivariate flood design standard value calculation method based on the combined relationship curve, step A, collects multi-year hydrological data upstream of the proposed reservoir design object, and sorts out the data series of flood peaks in each year and flood volumes in different time periods, and then selects the time period flood volumes to form a multivariate extreme value series, wherein the multivariate extreme value series includes the flood peak-flood volume series and the flood volume-flood peak series.

[0013] The aforementioned multivariate flood design standard value calculation method based on the combined relationship curve, step B, performs variability detection on the selected multivariate extreme value series to obtain the variability detection result. The specific steps are as follows:

[0014] Step B1, use the multivariate Mann-Kendall type covariance to perform trend detection on the multivariate extreme value series, as shown in formula (1),

[0015] ,

[0016] ,

[0017] (1);

[0018] in, is the variance of the population statistic H, is the statistic of the u-th variable, is the covariance matrix, is the total number of variables, is the main variable number, is the secondary variable number, is the covariance estimator, is the value of the i-th point on the variable u; , is a symbolic function, and As shown in formula (2),

[0019] (2);

[0020] Step B2, using the Copula-based likelihood ratio test CLR to perform jump detection on multivariate series, and the test statistic Z n As shown in formula (3),

[0021] (3)

[0022] in, , and are the likelihood functions of the entire sequence, the sequence before the change point, and the sequence after the change point respectively; , and They are the maximum likelihood estimates of the parameters of the entire sequence, the sequence before the change point, and the sequence after the change point;

[0023] Step B3, calculate the likelihood ratio statistic The asymptotic distribution of , as shown in formula (4),

[0024] (4)

[0025] in, is the number of Copula parameters with change points, , O is a higher-order infinitesimal;

[0026] Step B4, calculate the efficiency coefficient for comprehensive detection of multivariate series variability , as shown in formula (5),

[0027] (5)

[0028] And in the trend variation model , while in the jump mutation model As shown in formula (6), (6)

[0029] in, is the efficiency coefficient, is the measured value, i is the time, j is the sequence dimension, is the measured sequence mean, is the simulated value of the mutation model.

[0030] The aforementioned multivariate flood design standard value calculation method based on the combined relationship curve, step C, constructs a single variable variable parameter marginal distribution model under inconsistency based on the variability detection results, and then uses the dynamic Copula function to construct a multivariate dynamic joint distribution model under inconsistency conditions. The specific steps are as follows:

[0031] Step C1, based on the P-III distribution and the GEV distribution, the parameters in the distribution function are set to change with covariates such as time, and a univariate variable parameter marginal distribution model is constructed;

[0032] Step C2, set the structural parameters in the Copula function to change with the covariate to describe the inconsistency characteristics of the correlation structure between variables, and then build a multivariate dynamic joint distribution model, as shown in formula (7),

[0033] (7)

[0034] in, and are the univariate variable parameter marginal distributions of flood peak and flood volume, is the Copula function structure parameter that changes with time and other covariates, is the probability density function of the dynamic Copula;

[0035] Step C3, select the GEV variable parameter marginal distribution as the marginal distribution, and then select the dynamic ClaytonCopula function as the structure function, as shown in formula (8),

[0036] ,

[0037] (8)

[0038] in, , and They are the location parameter, scale parameter, and shape parameter of the GEV distribution that vary with the covariate;

[0039] Step C4, use Adaptive Metropolis sampling method to estimate parameters. The specific steps are as follows:

[0040] Step C41, set the jth accepted sample among the extracted parameter samples as θ j , when sampling for the t+1th time, from the proposed distribution q(θ j , C0) randomly extracts a parameter sample, denoted as θ (t+1)^* ;

[0041] Step C42, randomly select a number from the (0, 1) uniform distribution, denoted as u (t+1) ;

[0042] Step C43, using the Metropolis-Hasting algorithm to determine the criterion for random sample acceptance, if the random sample is accepted, let θ (j+1) =θ (t+1)^* ;

[0043] Step C44, when j> When , the sampling warm-up ends and the sample tends to be stable. At this time, the covariance matrix Perform adaptive update, as shown in formula (9),

[0044] (9)

[0045] in, is the scale factor, d is the variable dimension, is the covariance of the parameters, is a positive number, is a matrix, is the identity matrix;

[0046] Step C45, repeating steps C41-C44 to generate parameter samples that meet the requirements so as to perform statistical analysis of the parameters;

[0047] Step C5, using the deviation information criterion DIC to screen the optimal multivariate dynamic joint distribution model, where the model with the smallest deviation information criterion DIC is the optimal model. The specific steps are as follows:

[0048] Step C51, calculate , as shown in formula (10),

[0049] (10)

[0050] Where D is the deviation, is the mean deviation, is the model parameter, M is the number of iterations, B is the number of warm-up iterations, For samples;

[0051] Step C52, calculate the number of valid parameters of the model , as shown in formula (11),

[0052] (11)

[0053] in, is the model parameter The posterior mean of ;

[0054] Step C53, calculating model parameters The posterior mean of , as shown in formula (12),

[0055] (12);

[0056] Step C54, calculating the deviation information criterion DIC and the deviation As shown in formula (13) and formula (14),

[0057] (13)

[0058] (14)

[0059] in, For time point The observed value on .

[0060] The aforementioned multivariate flood design standard value calculation method based on the combination relationship curve, step D, according to the multivariate dynamic joint distribution model under the non-consistent condition and using the conditional combination method to establish the combination relationship curve between the primary and secondary variables, the specific steps are as follows:

[0061] Step D1: Build and select the optimal multivariate joint distribution model , as shown in formula (15),

[0062] , (15);

[0063] Step D2, based on the main variable inconsistency marginal distribution function Find the quantile of each year at the specified frequency p , as shown in formula (16),

[0064] (16);

[0065] Step D3, calculate the quantile using the conditional combination method The expected combined quantile of the corresponding secondary variable , as shown in formula (17),

[0066] (17);

[0067] Step D4, using the main variable X as the horizontal axis and the secondary variable Y as the vertical axis to draw the quantile combination data, and then determine the combination relationship curve by fitting .

[0068] The aforementioned method for calculating the multivariable flood design standard value based on the combined relationship curve, step E, uses the equal reliability method to calculate the design value of the main variable under inconsistency, and then obtains the multivariable flood design standard value under inconsistency according to the combined relationship curve between the main and secondary variables. The specific steps are as follows:

[0069] Step E1, according to the set standard recurrence period T and engineering life L, and based on the marginal distribution function of the main variable X, the equal reliability method is used to calculate the design value of the main variable under the inconsistent single variable situation , as shown in formula (18),

[0070] (18)

[0071] in, is the reliability of the hydrological design value within the service life of L years under the condition of inconsistency, is the extreme value distribution of the sample series in year t (t=1, 2, …, L), and is the change with time t, is the given standard design value, T is the design standard recurrence period;

[0072] Step E2: Set the main variable standard design value Substituting into the combined relationship curve equation obtained in step D, we have , and the combined value Design criteria values ​​for multivariate floods.

[0073] The multivariate flood design standard value calculation system based on the combined relationship curve comprises a data acquisition module, a variability detection module, a distribution model construction module, a combined relationship curve establishment module, and a multivariate flood design standard value acquisition module. The data acquisition module is used to collect the multi-year hydrological data of the upstream of the proposed reservoir design object and sort out the data series of the flood peaks of each year and the flood volume in different periods, and then select the flood volume in each period to form a multivariate extreme value series; the variability detection module is used to perform variability detection on the selected multivariate extreme value series and obtain the variability detection result; the distribution model construction module is used to construct a single variable variable parameter marginal distribution model under inconsistency based on the variability detection result, and then use the dynamic Copula function to construct a multivariate dynamic joint distribution model under inconsistency conditions; the combined relationship curve establishment module is used to establish a combined relationship curve between primary and secondary variables according to the multivariate dynamic joint distribution model under inconsistency conditions and by using the conditional combination method; the multivariate flood design standard value acquisition module is used to calculate the design value of the primary variable under inconsistency by using the equal reliability method, and then obtain the multivariate flood design standard value under inconsistency according to the combined relationship curve between primary and secondary variables.

[0074] The beneficial effects of the present invention are as follows: the multivariate flood design standard value calculation method and system based on the combined relationship curve of the present invention first collects the multi-year hydrological data upstream of the proposed reservoir design object and sorts out the data series of the flood peaks of each year and the flood volumes in different time periods, then selects the flood volumes in each time period to form a multivariate extreme value series, then performs variability detection on the selected multivariate extreme value series and obtains the variability detection results, then constructs a single variable variable parameter marginal distribution model under inconsistency based on the variability detection results, then uses the dynamic Copula function to construct a multivariate dynamic joint distribution model under inconsistency conditions, then uses the multivariate dynamic joint distribution model under inconsistency conditions and adopts the conditional combination method to establish a combined relationship curve between the primary and secondary variables, finally uses the equal reliability method to calculate the design value of the primary variable under inconsistency, and then uses the The multivariate flood design standard value under inconsistency is obtained according to the combined relationship curve between the primary and secondary variables; the multivariate flood design standard value calculation method and system are effectively realized to have the function of diagnosing the variability of the multivariate series, and a multivariate dynamic joint distribution model can be constructed for the multivariate series that has mutated, and then the combined relationship curve between the primary and secondary variables is established through the conditional most likely combination method, and the combined relationship curve is queried based on the design value of the primary variable under inconsistency to obtain the corresponding secondary variable design standard value, thereby deducing the multivariate flood design standard value for a given recurrence period under inconsistency. Not only is the calculation efficiency high, but the calculation result is also relatively accurate, which ensures the safety of water conservancy projects. The design standard value obtained by deduction can change over time and obtain a uniquely determined design value, thereby accurately guiding the actual operation of the project. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1It is a flow chart of the method for calculating multivariable flood design standard value based on the combined relationship curve of the present invention;

[0076] Figure 2 It is a schematic diagram of the combination relationship curve of the present invention. DETAILED DESCRIPTION

[0077] The present invention will be further described below in conjunction with the accompanying drawings.

[0078] like Figure 1 As shown, the multivariable flood design standard value calculation method and system based on the combined relationship curve of the present invention includes the following steps:

[0079] Step A, collect multi-year hydrological data upstream of the proposed reservoir design object, and sort out the data series of flood peaks in each year and flood volumes in different periods, and then select the flood volumes in different periods to form a multivariate extreme value series, where the multivariate extreme value series includes the flood peak-flood volume series and the flood volume-flood peak series.

[0080] Step B, perform variability detection on the selected multivariate extreme value series to obtain the variability detection results. The specific steps are as follows:

[0081] Step B1, use the multivariate Mann-Kendall type covariance to perform trend detection on the multivariate extreme value series, as shown in formula (1),

[0082] ,

[0083] ,

[0084] (1);

[0085] in, is the variance of the population statistic H, is the statistic of the u-th variable, is the covariance matrix, is the total number of variables, is the main variable number, is the secondary variable number, is the covariance estimator, is the value of the i-th point on the variable u; , is a symbolic function, and As shown in formula (2),

[0086] (2);

[0087] Step B2, using the Copula-based likelihood ratio test CLR to perform jump detection on multivariate series, and the test statistic Zn As shown in formula (3),

[0088] (3)

[0089] in, , and are the likelihood functions of the entire sequence, the sequence before the change point, and the sequence after the change point respectively; , and They are the maximum likelihood estimates of the parameters of the entire sequence, the sequence before the change point, and the sequence after the change point;

[0090] Step B3, calculate the likelihood ratio statistic The asymptotic distribution of , as shown in formula (4),

[0091] (4)

[0092] in, is the number of Copula parameters with change points, , O is a higher-order infinitesimal;

[0093] Step B4, calculate the efficiency coefficient for comprehensive detection of multivariate series variability , as shown in formula (5),

[0094] (5)

[0095] And in the trend variation model , while in the jump mutation model As shown in formula (6), (6)

[0096] in, is the efficiency coefficient, is the measured value, i is the time, j is the sequence dimension, is the measured sequence mean, is the simulated value of the mutation model.

[0097] Step C: Based on the variability detection results, a single variable variable parameter marginal distribution model under inconsistency is constructed, and then a multivariate dynamic joint distribution model under inconsistency is constructed using the dynamic Copula function. The specific steps are as follows:

[0098] Step C1, based on the P-III distribution and the GEV distribution, the parameters in the distribution function are set to change with covariates such as time, and a univariate variable parameter marginal distribution model is constructed;

[0099] Step C2, set the structural parameters in the Copula function to change with the covariate to describe the inconsistency characteristics of the correlation structure between variables, and then build a multivariate dynamic joint distribution model, as shown in formula (7),

[0100] (7)

[0101] in, and are the univariate variable parameter marginal distributions of flood peak and flood volume, is the Copula function structure parameter that changes with time and other covariates, is the probability density function of the dynamic Copula;

[0102] Step C3, select the GEV variable parameter marginal distribution as the marginal distribution, and then select the dynamic ClaytonCopula function as the structure function, as shown in formula (8),

[0103] ,

[0104] (8)

[0105] in, , and They are the location parameter, scale parameter, and shape parameter of the GEV distribution that vary with the covariate;

[0106] Step C4, use Adaptive Metropolis sampling method to estimate parameters. The specific steps are as follows:

[0107] Step C41, set the jth accepted sample among the extracted parameter samples as θ j , when sampling for the t+1th time, from the proposed distribution q(θ j , C0) randomly extracts a parameter sample, denoted as θ (t+1)^* ;

[0108] Step C42, randomly select a number from the (0, 1) uniform distribution, denoted as u (t+1) ;

[0109] Step C43, using the Metropolis-Hasting algorithm to determine the criterion for random sample acceptance, if the random sample is accepted, let θ (j+1) =θ (t+1)^* ;

[0110] Step C44, when j> When , the sampling warm-up ends and the sample tends to be stable. At this time, the covariance matrix Perform adaptive update, as shown in formula (9),

[0111] (9)

[0112] in, is the scale factor, d is the variable dimension, is the covariance of the parameters, is a positive number, is a matrix, is the identity matrix;

[0113] Step C45, repeating steps C41-C44 to generate parameter samples that meet the requirements so as to perform statistical analysis of the parameters;

[0114] Step C5, using the deviation information criterion DIC to screen the optimal multivariate dynamic joint distribution model, where the model with the smallest deviation information criterion DIC is the optimal model. The specific steps are as follows:

[0115] Step C51, calculate , as shown in formula (10),

[0116] (10)

[0117] Where D is the deviation, is the mean deviation, is the model parameter, M is the number of iterations, B is the number of warm-up iterations, For samples;

[0118] Step C52, calculate the number of valid parameters of the model , as shown in formula (11),

[0119] (11)

[0120] in, is the model parameter The posterior mean of ;

[0121] Step C53, calculating model parameters The posterior mean of , as shown in formula (12),

[0122] (12);

[0123] Step C54, calculating the deviation information criterion DIC and the deviation As shown in formula (13) and formula (14),

[0124] (13)

[0125] (14)

[0126] in, For time point The observed value on .

[0127] like Figure 2 As shown, step D, based on the multivariate dynamic joint distribution model under inconsistent conditions and using the conditional combination method to establish the combined relationship curve between the primary and secondary variables, the specific steps are as follows:

[0128] Step D1: Build and select the optimal multivariate joint distribution model , as shown in formula (15),

[0129] , (15);

[0130] Step D2, based on the main variable inconsistency marginal distribution function Find the quantile of each year at the specified frequency p , as shown in formula (16),

[0131] (16);

[0132] Step D3, calculate the quantile using the conditional combination method The expected combined quantile of the corresponding secondary variable , as shown in formula (17),

[0133] (17);

[0134] Step D4, using the main variable X as the horizontal axis and the secondary variable Y as the vertical axis to draw the quantile combination data, and then determine the combination relationship curve by fitting .

[0135] Step E, using the equal reliability method to calculate the design value of the main variable under inconsistency, and then obtaining the multivariate flood design standard value under inconsistency based on the combined relationship curve between the main and secondary variables. The specific steps are as follows:

[0136] Step E1, according to the set standard recurrence period T and engineering life L, and based on the marginal distribution function of the main variable X, the equal reliability method is used to calculate the design value of the main variable under the inconsistent single variable situation , as shown in formula (18),

[0137] (18)

[0138] in, is the reliability of the hydrological design value within the service life of L years under the condition of inconsistency, is the extreme value distribution of the sample series in year t (t=1, 2, …, L), and is the change with time t, is the given standard design value, T is the design standard recurrence period;

[0139] Step E2: Set the main variable standard design value Substituting into the combined relationship curve equation obtained in step D, we have , and the combined value Design criteria values ​​for multivariate floods.

[0140] The multivariate flood design standard value calculation system based on the combined relationship curve comprises a data acquisition module, a variability detection module, a distribution model construction module, a combined relationship curve establishment module, and a multivariate flood design standard value acquisition module. The data acquisition module is used to collect the multi-year hydrological data of the upstream of the proposed reservoir design object and sort out the data series of the flood peaks of each year and the flood volume in different periods, and then select the flood volume in each period to form a multivariate extreme value series; the variability detection module is used to perform variability detection on the selected multivariate extreme value series and obtain the variability detection result; the distribution model construction module is used to construct a single variable variable parameter marginal distribution model under inconsistency based on the variability detection result, and then use the dynamic Copula function to construct a multivariate dynamic joint distribution model under inconsistency conditions; the combined relationship curve establishment module is used to establish a combined relationship curve between primary and secondary variables according to the multivariate dynamic joint distribution model under inconsistency conditions and by using the conditional combination method; the multivariate flood design standard value acquisition module is used to calculate the design value of the primary variable under inconsistency by using the equal reliability method, and then obtain the multivariate flood design standard value under inconsistency according to the combined relationship curve between primary and secondary variables.

[0141] In summary, the multivariate flood design standard value calculation method and system based on the combined relationship curve of the present invention first collects the multi-year hydrological data upstream of the proposed reservoir design object and sorts out the data series of the flood peaks of each year and the flood volumes in different periods, then selects the flood volumes in each period to form a multivariate extreme value series, then performs variability detection on the selected multivariate extreme value series and obtains the variability detection results, then constructs a single variable variable parameter marginal distribution model under inconsistency based on the variability detection results, then uses the dynamic Copula function to construct a multivariate dynamic joint distribution model under inconsistency conditions, then uses the multivariate dynamic joint distribution model under inconsistency conditions and adopts the conditional combination method to establish the combined relationship curve between the primary and secondary variables, finally uses the equal reliability method to calculate the design value of the primary variable under inconsistency, and then uses the primary and secondary variables according to the conditional combination method. The combined relationship curve between secondary variables is used to obtain the multivariate flood design standard value under inconsistency; the multivariate flood design standard value calculation method and system are effectively realized to have the function of diagnosing the variability of multivariate series, and a multivariate dynamic joint distribution model can be constructed for the multivariate series that has mutated, and then the combined relationship curve between the primary and secondary variables is established through the conditional most likely combination method, and the combined relationship curve is queried based on the design value of the primary variable under inconsistency to obtain the corresponding secondary variable design standard value, thereby deducing the multivariate flood design standard value for a given recurrence period under inconsistency. Not only is the calculation efficiency high, but the calculation result is also relatively accurate, which ensures the safety of water conservancy projects. The design standard value obtained by deduction can change over time and obtain a uniquely determined design value, thereby accurately guiding the actual operation of the project.

[0142] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A multivariable flood design standard value calculation method based on a combined relationship curve, characterized by: The following steps are included: Step A, collecting multi-year hydrological data of the upstream of the proposed reservoir design object, and sorting out the data series of flood peaks in each year and flood volumes in different periods, and then selecting flood volumes in different periods to form a multivariate extreme value series, wherein the multivariate extreme value series includes a flood peak-flood volume series and a flood volume-flood peak series; Step B, perform variability detection on the selected multivariate extreme value series to obtain the variability detection results. The specific steps are as follows: Step B1, use the multivariate Mann-Kendall type covariance to perform trend detection on the multivariate extreme value series, as shown in formula (1), Among them, var(H) is the variance of the population statistic H, M u is the statistic of the u-th variable, C u,v is the covariance matrix, d is the total number of variables, u is the main variable number, v is the secondary variable number, is the covariance estimator, is the value of the ith point on the variable u; u≠v, sgn(·) is the sign function, and sgn(·) is as shown in formula (2), Step B2, using the Copula-based likelihood ratio test CLR to perform jump detection on multivariate series, and the test statistic Z n As shown in formula (3), Among them, L n (·),L k (·)and are the likelihood functions of the entire sequence, the sequence before the change point, and the sequence after the change point respectively; and They are the maximum likelihood estimates of the parameters of the entire sequence, the sequence before the change point, and the sequence after the change point; Step B3, calculate the likelihood ratio statistic The asymptotic distribution P of is shown in formula (4). Where q is the number of Copula parameters with change points, h = l = [lg (n) 3 / 2 / n], O is a high-order infinitesimal; Step B4, calculate the efficiency coefficient N for comprehensive detection of multivariate series variability se , as shown in formula (5), And in the trend variation model, S i,j =f(i,j), and in the jump mutation model, S i As shown in formula (6), Among them, N se is the efficiency coefficient, O ij is the measured value, i is the time, j is the sequence dimension, is the measured sequence mean, S ij is the simulated value of the mutation model; Step C, constructing a single variable variable parameter marginal distribution model under inconsistency based on the variability detection results, and then using the dynamic Copula function to construct a multivariate dynamic joint distribution model under inconsistency conditions; Step D, establishing a combination relationship curve between primary and secondary variables using a conditional combination method based on a multivariate dynamic joint distribution model under inconsistent conditions; Step E: Calculate the design value of the main variable under inconsistency using the equal reliability method, and then obtain the multivariate flood design standard value under inconsistency based on the combined relationship curve between the main and secondary variables.

2. The method for calculating multivariable flood design standard values ​​based on combined relationship curves according to claim 1 is characterized in that: Step C: Based on the variability detection results, a single variable variable parameter marginal distribution model under inconsistency is constructed, and then a multivariate dynamic joint distribution model under inconsistency is constructed using the dynamic Copula function. The specific steps are as follows: Step C1, based on the P-III distribution and the GEV distribution, the parameters in the distribution function are set to change with the time covariate, and a univariate variable parameter marginal distribution model is constructed; Step C2, set the structural parameters in the Copula function to change with the covariate to describe the inconsistency characteristics of the correlation structure between variables, and then build a multivariate dynamic joint distribution model, as shown in formula (7), F t (x,y)=C(F X (x),F Y (y),θ(·))=C(u,v,θ(t)) (7) Among them, F X (x) and F Y (y) are the univariate variable parameter marginal distributions of flood peak and flood volume, θ(·) is the structural parameter of the Copula function that changes with time covariates, and c(u,v,θ(t)) is the probability density function of the dynamic Copula; Step C3, select the GEV variable parameter marginal distribution as the marginal distribution, and then select the dynamic Clayton Copula function as the structure function, as shown in formula (8), f ClaytonS (x,y)=(1+θ(·))(uv) -θ(·)-1 (u -θ(·) +v -θ(·) -1) -2-1 / θ(·) f X (x)f Y (y), Among them, ξ(t), μ(t) and K(t) are the location parameter, scale parameter and shape parameter of the GEV distribution that vary with the covariate; Step C4, use Adaptive Metropolis sampling method to estimate parameters. The specific steps are as follows: Step C41, set the jth accepted sample among the extracted parameter samples as θ j , when sampling for the t+1th time, from the proposed distribution q(θ j , C0) randomly extracts a parameter sample, denoted as θ (t+1)^* ; Step C42, randomly select a number from the (0, 1) uniform distribution, denoted as u (t+1) ; Step C43, using the Metropolis-Hasting algorithm to determine the criterion for random sample acceptance, if the random sample is accepted, let θ (j+1) =θ (t+1)^* ; Step C44, when j>N0, the sampling preheating ends and the sample tends to be stable. At this time, the covariance matrix C0 is adaptively updated, as shown in formula (9), Among them, s is the scale factor, d is the variable dimension, Cov() is the covariance of the parameter, ε is a positive number, C i is a matrix, I d is the identity matrix; Step C45, repeating steps C41-C44 to generate parameter samples that meet the requirements so as to perform statistical analysis of the parameters; Step C5, using the deviation information criterion DIC to screen the optimal multivariate dynamic joint distribution model, where the model with the smallest deviation information criterion DIC is the optimal model. The specific steps are as follows: Step C51, calculate As shown in formula (10), Where D is the deviation, is the deviation from the mean, β s is the model parameter, M is the number of iterations, B is the number of warm-up iterations, and m is the sample; Step C52, calculate the number of valid parameters n of the model D , as shown in formula (11), in, is the model parameter β s The posterior mean of ; Step C53, calculating the model parameter β s The posterior mean of As shown in formula (12), Step C54, calculate the deviation information criterion DIC and the deviation D(β s ) are shown in formula (13) and formula (14) respectively, in, is time point t i The observed value on .

3. The method for calculating multivariable flood design standard values ​​based on combined relationship curves according to claim 2 is characterized in that: Step D, based on the multivariate dynamic joint distribution model under inconsistent conditions and using the conditional combination method to establish the combined relationship curve between the primary and secondary variables, the specific steps are as follows: Step D1: Build and select the optimal multivariate joint distribution model F t (x,y), as shown in formula (15), F t (x,y)=C(F X (x),F Y (y),θ(·))=C(u,v,θ(t)),t=1,2,…,n+L (15); Step D2, based on the main variable inconsistency marginal distribution function F X (x) Find the quantile x of each year at the specified frequency p (t) , as shown in formula (16), x (t) ={x (1) ,x (2) ,…,x (n+L) } (16); Step D3, calculate the quantile x using the conditional combination method (t) The expected combined quantile y of the corresponding secondary variable (t) , as shown in formula (17), and (t) ={and (1) ,and (2) ,…,and (n+L) } (17); Step D4, using the main variable X as the horizontal axis and the secondary variable Y as the vertical axis to draw the quantile combination point data, and then determine the combination relationship curve y=f(x) by fitting.

4. The method for calculating multivariable flood design standard values ​​based on combined relationship curves according to claim 3 is characterized in that: Step E, using the equal reliability method to calculate the design value of the main variable under inconsistency, and then obtain the multivariate flood design standard value under inconsistency according to the combined relationship curve between the main and secondary variables. The specific steps are as follows: Step E1, according to the set standard recurrence period T and engineering life L, and based on the marginal distribution function of the main variable X, the equal reliability method is used to calculate the main variable design value x under the inconsistent single variable situation T , as shown in formula (18), in, is the reliability of the hydrological design value within the service life of L years under the condition of inconsistency, F t () is the extreme value distribution of the sample series in year t, and F t () is the change with time t, X T is the given standard design value, T is the design standard recurrence period; Step E2: Set the main variable standard design value x T Substitute the combined relationship curve equation obtained in step D into the equation, then we have y T =f(x T ), and the combined value (x T ,y T ) is the multivariate flood design standard value.

5. A multivariable flood design standard value calculation system based on a combined relationship curve, wherein the calculation process of the calculation system is based on the calculation method according to any one of claims 1 to 4, characterized in that: It includes a data acquisition module, a variability detection module, a distribution model building module, a combination relationship curve building module, and a multivariate flood design standard value acquisition module. The data acquisition module is used to collect multi-year hydrological data upstream of the proposed reservoir design object and sort out the data series of flood peaks in each year and flood volumes in different periods, and then select the flood volumes in different periods to form a multivariate extreme value series; The variability detection module is used to perform variability detection on the selected multivariate extreme value series and obtain variability detection results; The distribution model building module is used to build a single variable variable parameter marginal distribution model under inconsistency based on the variability detection results, and then use the dynamic Copula function to build a multivariate dynamic joint distribution model under inconsistency conditions; The combination relationship curve establishment module is used to establish a combination relationship curve between primary and secondary variables based on a multivariate dynamic joint distribution model under inconsistent conditions and using a conditional combination method; The multivariable flood design standard value acquisition module is used to calculate the main variable design value under inconsistency using the equal reliability method, and then obtain the multivariable flood design standard value under inconsistency according to the combination relationship curve between the main and secondary variables.

Images