Inland natural river design water level calculation method and system

Abstract

The invention relates to an inland natural river design water level calculation method and system, and the calculation method comprises the following steps: S1, collecting multi-source hydrological data, analyzing the collected data, and generating a standardized hydrological database; s2, screening long-series continuous observation data of hydrometric stations, constructing a water level influence factor matrix, and generating water level data; s3, constructing a hydrological element comprehensive duration curve, identifying periodicity and abrupt change points, and calculating design water levels in different recurrence periods based on a probability distribution model and in combination with change feature correction; and S4, evaluating precision through cross validation, outputting a comprehensive duration curve, and designing a water level and an uncertainty interval. According to the method and system, multi-source data can be integrated, the hydrological element incidence relation can be comprehensively analyzed, and the design water level calculation method and system are dynamically corrected, so that the calculation precision and reliability of the design water level of the natural inland river are improved.

Description

Technical Field

[0001] This invention relates to the fields of port and waterway engineering, hydrological monitoring and hydrological data management, and more specifically, to a method and system for calculating the design water level of natural inland rivers. Background Technology

[0002] River water level changes directly affect flood control safety along the riverbanks, navigation scheduling, and water resource utilization efficiency. Design water level calculation is one of the core technologies in port and waterway engineering and water conservancy engineering design, and its accuracy is directly related to the construction standards of flood control projects and the rationality of port and water conservancy engineering design schemes.

[0003] Currently, the calculation of design water levels for inland natural rivers mainly relies on traditional hydrological models (such as the Muskingan model and the unit hydrograph method) or statistical analysis methods based on historical water levels (such as Pearson type III curve fitting). However, existing technologies have the following limitations:

[0004] Insufficient data integration: Traditional methods are mostly based on local data from a single hydrological station, failing to fully integrate multi-source heterogeneous data from meteorological (precipitation, evaporation), topographic (gradient, roughness), and human activities (reservoir scheduling, water intake) within the basin, resulting in an incomplete analysis of the driving factors of water level changes; Missing factors: The dynamic coupling relationship between hydrological elements such as flow, sediment concentration, and water temperature and water level is not systematically considered, making it difficult to reflect the comprehensive impact of complex hydrological processes on water level; Design frequency deviation: Design water levels based on simple frequency statistics do not fully consider the changes in water level frequency distribution caused by extreme climates (such as rainstorms and droughts) and human activities, leading to inaccurate calculations of design water levels. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method and system for calculating the design water level of natural inland rivers, which can solve the problems of insufficient data integration, missing factors and design frequency deviation in the existing technology, and realize the accurate calculation of the design water level of hydrological stations.

[0006] The technical solution adopted by this invention to solve its technical problem is: to construct a method for calculating the design water level of natural inland rivers, including the following steps:

[0007] S1. Collect multi-source hydrological data, analyze the collected data, and generate a standardized hydrological database;

[0008] S2. Screen long-series continuous observation data from hydrological stations, construct a matrix of factors influencing water level, and generate water level data;

[0009] S3. Construct a comprehensive time-lapse curve of hydrological elements, identify periodicity and abrupt change points, and calculate the design water level for different return periods based on the probability distribution model and combined with the change characteristics correction.

[0010] S4. Evaluate the accuracy through cross-validation and output the comprehensive duration curve, design water level, and uncertainty range.

[0011] According to the above scheme, step S1 includes the following steps:

[0012] S101. Obtain historical hydrological data from multiple hydrological stations along the main stream of the Yangtze River, as well as multi-source heterogeneous data such as meteorological station data, reservoir scheduling data, and land use change data within the basin;

[0013] S102. Parse the raw data from text, tables, and databases, remove outliers, fill in missing values, and unify the data time resolution;

[0014] S103. Convert the data of each element according to a unified standard to generate a standardized hydrological database.

[0015] According to the above scheme, step S2 specifically involves: based on a standardized historical database, screening long-series continuous observation data from various hydrological stations, and extracting daily or monthly average water levels as the basic water level sequence; combining meteorological elements and human activity influencing factors within the basin, constructing a water level influencing factor matrix, and screening key factors that significantly affect water levels using Pearson correlation coefficient and mutual information method.

[0016] According to the above scheme, step S3 includes the following steps:

[0017] S301. Using time series as the horizontal axis and key hydrological elements such as water level, flow rate, precipitation, and sediment concentration as the vertical axis, construct a comprehensive time-series curve of hydrological elements that change synchronously with multiple elements.

[0018] S302. Based on a long series of water level sequences, select a probability distribution model suitable for extreme value data, and fit the model parameters through maximum likelihood estimation or Bayesian method; modify the probability distribution model to obtain the design water level probability density function that considers long-term changes; calculate the design maximum water level and design minimum water level corresponding to different return periods according to the modified probability distribution model.

[0019] According to the above scheme, in step S301, the sliding window method is used to perform segmented analysis on the comprehensive time-lapse curve, extract the extreme values ​​of water level, element extreme values ​​and corresponding occurrence times in each window, and identify the periodicity and abrupt change points of water level changes. Specifically, this includes the following steps:

[0020] S301a, Multidimensional Time Series Modeling:

[0021] Let the time series be T = {t1, t2, ..., t...} n}, corresponding to the water level sequence H={h1,h2,...,h nThe set of key influencing factors X = {X1, X2, ..., X} m}, where X j ={x j1 ,x j2 ,...,x jn The mathematical expression for constructing the comprehensive duration curve is as follows:

[0022] C(t i )=(h i ,x 1i ,x 2i ,...,x mi ), i = 1, 2, ..., n

[0023] S301b, Sliding window feature extraction:

[0024] Define a window function W(k,w,s), where w is the window width and s is the sliding step size. For each window k, calculate:

[0025] Extreme value characteristics:

[0026] H k ^{max}=max{h i |t i ∈W(k)}

[0027] H k ^{min}=min{h i |t i ∈W(k)}

[0028] Coupling characteristics:

[0029] ρ jk =corr(H k ,X jk (j=1,2,...,m)

[0030] Where corr(·) represents the Pearson correlation coefficient.

[0031] Mutation point detection:

[0032] Bayesian variable detection algorithm is used:

[0033] P(τ|H)∝P(H|τ)P(τ)

[0034] Where τ is the location of the mutation point, and it is determined to be a significant mutation point when the posterior probability P(τ|H)>0.95.

[0035] According to the above scheme, in step S302, the probability distribution model adopts the generalized extreme value distribution GEV for modeling:

[0036] F(h;μ,σ,ξ)=exp{-[1+ξ(h-μ) / σ]^{-1 / ξ}}

[0037] Where μ is the position parameter, σ is the scale parameter, and ξ is the shape parameter.

[0038] Parameter estimation is achieved through maximum likelihood estimation:

[0039] L(μ,σ,ξ|H)=∏f(h i ;μ,σ,ξ)

[0040] Where f(·) is the GEV probability density function.

[0041] According to the above scheme, in step S302, the probability distribution model is modified by combining the periodicity and abrupt change characteristics identified in step 301 to obtain the design water level probability density function considering long-term changes. Assuming a sudden change point τ is detected, the data is divided into:

[0042] D1={h i |t i <τ} and D2={h i |t i ≥τ},

[0043] The following results were obtained through fitting:

[0044] F1(h;μ1,σ1,ξ1) and F2(h;μ2,σ2,ξ2)

[0045] The corrected mixed distribution is:

[0046] F*(h)=αF1(h)+(1-α)F2(h)

[0047] Where α = n1 / (n1+n2), and n1 and n2 are the number of segments of data.

[0048] According to the above scheme, in step S302, the design maximum water level and design minimum water level corresponding to different return periods are calculated based on the modified probability distribution model, and the formula is as follows:

[0049] The design maximum water level corresponding to the return period T:

[0050] DHWMT=F -1 (1-1 / T)

[0051] Minimum design water level:

[0052] DHWMT=F -1 (1 / T)

[0053] Where F is the corrected probability distribution function and T is the return period.

[0054] According to the above scheme, in step S4, the accuracy of the calculation results is verified by leave-one-out cross-validation. Each time, the data of one hydrological station is removed, and the model is trained and the design water level of the station is predicted using the data of the remaining stations. The root mean square error and Nash efficiency coefficient are calculated.

[0055] RMSE=\sqrt{\frac{1}{N}\sum_{k=1}^{N}(h_k-\hat{h}_k) 2}

[0056] Where hk is the measured water level value of the k-th station, and N is the total number of hydrological stations. NSE = 1 - \frac{\sum_{k=1}^{N}(h_k-\hat{h}_k) 2}{\sum_{k=1}^{N}(h_k-\bar{h}) 2}

[0057] Where, \bar{h}=(1 / N)∑h k The average value is the result of the measured values. The output includes the comprehensive duration curves for each hydrological station, design water levels at different frequencies, and uncertainty intervals. The 98% confidence interval is calculated using the Bootstrap method.

[0058] CI = [F -1 (0.01),F -1 (0.99)].

[0059] The present invention also provides a system for calculating the design water level of natural inland rivers, including a data acquisition module, a data preprocessing module, a comprehensive duration curve construction module, and a result verification and output module;

[0060] The data acquisition module is used to access multi-source heterogeneous data from the Yangtze River main stream and its basin, and supports automatic data format recognition and parsing.

[0061] The data preprocessing module is used to complete data cleaning, missing value imputation, standardization transformation, and element matrix construction.

[0062] The comprehensive duration curve construction module is based on multi-factor synchronous data and is used to generate a comprehensive duration curve on a time series, and to identify periodicity and abrupt change points.

[0063] The high and low water level calculation module is used to integrate a probability distribution model and a frequency correction algorithm to calculate the design high and low water levels for different return periods.

[0064] The result verification and output module evaluates the model accuracy through cross-validation and outputs visualized results.

[0065] The method and system for calculating the design water level of inland natural rivers according to the present invention have the following beneficial effects:

[0066] 1. This invention integrates multi-source data, including hydrological, meteorological, and human activity data, to comprehensively reflect the driving factors of water level changes and overcome the limitations of a single data source.

[0067] 2. This invention quantifies the dynamic correlation between water level and factors such as flow rate and precipitation by integrating historical curves, overcoming the shortcomings of traditional methods that rely on only a single factor.

[0068] 3. This invention performs dynamic frequency correction and combines long-term variation characteristics, such as climate change and reservoir scheduling, to correct the probability distribution model and improve the accuracy of the design water level. It is especially suitable for high / low water level prediction under extreme climate conditions.

[0069] 4. This invention has a wide range of applications and can output design water levels and uncertainty ranges with different return periods, providing scientific support for flood control scheduling such as determining the threshold for joint scheduling of reservoir groups, water conservancy engineering design such as the design of levee height, and water resources planning such as the demonstration of the scale of inter-basin water transfer. Attached Figure Description

[0070] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings:

[0071] Figure 1 This is a schematic diagram of the architecture of the inland natural river design water level calculation system of the present invention;

[0072] Figure 2 This is a flowchart illustrating the construction process of the comprehensive duration curve of this invention;

[0073] Figure 3 This is the calculation logic diagram of the design water level at different frequencies according to the present invention;

[0074] Figure 4 This is a diagram showing the highest water level achieved in an embodiment of the present invention.

[0075] Figure 5 This is a frequency curve diagram in an embodiment of the present invention. Detailed Implementation

[0076] To provide a clearer understanding of the technical features, objectives, and effects of the present invention, specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0077] like Figure 1-5 As shown, the method for calculating the design water level of inland natural rivers according to the present invention includes the following steps:

[0078] S1. Collect multi-source hydrological data, analyze the collected data, and generate a standardized hydrological database, including the following steps:

[0079] S101. Obtain historical hydrological data from multiple hydrological stations along the main stream of the Yangtze River, including water level, flow rate, sediment concentration, water temperature, precipitation, evaporation, sediment particle size distribution, and multi-source heterogeneous data such as meteorological station data, reservoir scheduling data, and land use change data within the basin.

[0080] S102. Parse raw data in different formats, such as text, tables, and databases, remove outliers, fill in missing values, and use linear interpolation, time series interpolation, or machine learning interpolation methods to unify the data time resolution, such as daily or monthly scales.

[0081] S103. Convert the data of each element according to a unified standard, such as unit and reference surface, to generate a standardized hydrological database.

[0082] S2. Select long-term continuous observation data from hydrological stations, preferably no less than 20 years, to construct a water level influencing factor matrix and generate historical water level records. The specific method is as follows: Based on a standardized historical database, select long-term continuous observation data from each hydrological station and extract daily or monthly average water levels as the base water level sequence; combine this with meteorological elements within the basin, such as previous precipitation and concurrent evaporation, and human activity influencing factors, such as upstream reservoir storage and inter-regional water withdrawal, to construct a water level influencing factor matrix; and use Pearson correlation coefficient and mutual information methods to screen key factors that significantly affect water levels, such as five-day moving average precipitation and reservoir storage.

[0083] S3. Construct a comprehensive time-series curve of hydrological elements, identify periodicity and abrupt change points, and calculate the design high and low water levels for different return periods based on a probability distribution model and corrections for variation characteristics; including the following steps:

[0084] S301. Using time series as the horizontal axis and key hydrological elements such as water level, flow rate, precipitation, and sediment concentration as the vertical axis, construct a comprehensive time-series curve of hydrological elements that change synchronously with multiple elements.

[0085] The sliding window method is used to segment the comprehensive historical curve, such as a 30-year window with a step size of 5 years. Extreme water level values ​​(highest and lowest values), element extremes, and their corresponding occurrence times are extracted within each window. The periodicity of water level changes, such as interannual and interdecadal fluctuations, and abrupt changes, such as water level trends following reservoir construction or climate change, are identified. The specific steps include:

[0086] S301a, Multidimensional Time Series Modeling:

[0087] Let the time series be T = {t1, t2, ..., t...} n}, corresponding to the water level sequence H={h1,h2,...,h n The set of key influencing factors X = {X1, X2, ..., X} m}, where Xj ={x j1 ,x j2 ,...,x jn The mathematical expression for constructing the comprehensive duration curve is as follows:

[0088] C(t i )=(h i ,x 1i ,x 2i ,...,x mi ), i = 1, 2, ..., n

[0089] S301b, Sliding window feature extraction:

[0090] Define a window function W(k,w,s), where w is the window width (30 years recommended) and s is the sliding step size (5 years recommended). For each window k, calculate:

[0091] Extreme value characteristics:

[0092] H k ^{max}=max{h i |t i ∈W(k)}

[0093] H k ^{min}=min{h i |t i ∈W(k)}

[0094] Coupling characteristics:

[0095] ρ jk =corr(H k ,X jk (j=1,2,...,m)

[0096] Where corr(·) represents the Pearson correlation coefficient.

[0097] Mutation point detection:

[0098] Bayesian variable detection algorithm is used:

[0099] P(τ|H)∝P(H|τ)P(τ)

[0100] Where τ is the location of the mutation point, and it is determined to be a significant mutation point when the posterior probability P(τ|H)>0.95.

[0101] S302. Based on a long series of water level sequences, select a probability distribution model suitable for extreme value data, such as the generalized extreme value distribution GEV or the generalized Pareto distribution GP, ​​and fit the model parameters through maximum likelihood estimation or Bayesian methods; modify the probability distribution model to obtain the design water level probability density function that considers long-term changes; and calculate the design maximum and design minimum water levels corresponding to different return periods based on the modified probability distribution model.

[0102] Among them, the generalized extreme value distribution GEV modeling:

[0103] F(h;μ,σ,ξ)=exp{-[1+ξ(h-μ) / σ]^{-1 / ξ}}

[0104] Where μ is the position parameter, σ is the scale parameter, and ξ is the shape parameter.

[0105] Parameter estimation is achieved through maximum likelihood estimation:

[0106] L(μ,σ,ξ|H)=∏f(h i ;μ,σ,ξ)

[0107] Where f(·) is the GEV probability density function.

[0108] In step S302, the probability distribution model is modified based on the periodicity and abrupt change characteristics identified in step 301. For example, the distribution parameters are adjusted after a climate abrupt change point to obtain the design water level probability density function considering long-term changes. Let the abrupt change point τ be detected, and the data be divided into:

[0109] D1={h i |t i <τ} and D2={h i |t i ≥τ},

[0110] The following results were obtained through fitting:

[0111] F1(h;μ1,σ1,ξ1) and F2(h;μ2,σ2,ξ2)

[0112] The corrected mixed distribution is:

[0113] F*(h)=αF1(h)+(1-α)F2(h)

[0114] Where α = n1 / (n1+n2), and n1 and n2 are the number of segments of data.

[0115] In step S302, the design maximum water level (DHWM) and design minimum water level (DLWM) corresponding to different return periods are calculated according to the modified probability distribution model, such as 5 years, 10 years, 100 years, and 1000 years, using the following formulas:

[0116] The design maximum water level corresponding to the return period T:

[0117] DHWMT=F -1 (1-1 / T)

[0118] Minimum design water level:

[0119] DHWMT=F -1 (1 / T)

[0120] Where F is the corrected probability distribution function and T is the return period (in years).

[0121] S4. Evaluate the accuracy through cross-validation and output the comprehensive duration curve, design high and low water levels, and uncertainty range.

[0122] The accuracy of the calculation results was verified by leave-one-out cross-validation (LOOCV). Each time, the data of one hydrological station was removed, and the model was trained and the design water level of the station was predicted using the data of the remaining stations. The root mean square error (RMSE) and Nash efficiency coefficient (NSE) were calculated.

[0123] RMSE=\sqrt{\frac{1}{N}\sum_{k=1}^{N}(h_k-\hat{h}_k) 2}

[0124] Where hk is the measured water level value of the k-th station, and N is the total number of hydrological stations. NSE = 1 - \frac{\sum_{k=1}^{N}(h_k-\hat{h}_k) 2}{\sum_{k=1}^{N}(h_k-\bar{h}) 2}

[0125] Where, \bar{h}=(1 / N)∑h k The output is the average of the measured values, the comprehensive duration curve of each hydrological station, the design water level at different frequencies, and the uncertainty interval, such as the 98% confidence interval. The 98% confidence interval is calculated using the Bootstrap method.

[0126] CI = [F -1 (0.01),F -1 (0.99)].

[0127] The present invention also provides a system for calculating the design water level of natural inland rivers, including a data acquisition module, a data preprocessing module, a comprehensive duration curve construction module, and a result verification and output module.

[0128] The system comprises the following modules: Data Acquisition Module: Accessing multi-source heterogeneous data from the Yangtze River main stream and its basin, supporting automatic data format recognition and parsing; Data Preprocessing Module: Performing data cleaning, missing value imputation, standardization transformation, and element matrix construction; Comprehensive Duration Curve Construction Module: Generating comprehensive duration curves over time series based on multi-element synchronous data, and identifying periodicity and abrupt change points; Design Water Level Calculation Module: Integrating probability distribution models and frequency correction algorithms to calculate design high and low water levels for different return periods; Result Validation and Output Module: Evaluating model accuracy through cross-validation and outputting visualized results.

[0129] Example

[0130] The calculation of high and low water levels at a hydrological station in a certain city includes the following steps:

[0131] S1. Data Acquisition and Preprocessing

[0132] Daily water level data from 1950 to 2020 were collected from a station in a certain city. The data came from the Hydrology Bureau of the Yangtze River Water Resources Commission. Simultaneously, the scheduling data (water storage and discharge) of the Three Gorges Reservoir (operated in 1994) and Gezhouba Reservoir (operated in 1988) within the basin were collected, as well as daily precipitation and evaporation data from 1950 to 2020 for the basin above the station in the city. The data came from the China Meteorological Administration.

[0133] Missing water level data (such as the 5% missing data due to instrument maintenance) were filled using time series interpolation. Spatial interpolation (Kriging method) was used to obtain the average precipitation of the basin above a certain station in a certain city. The time resolution of the data was unified to the daily scale, and the data was standardized (water level was based on the Wusong datum, and precipitation was in mm).

[0134] S2, Generation of Historical Water Level Records

[0135] The daily average water level sequence of a certain station in a certain city from 1950 to 2020 was extracted, and 71 years of continuous and complete data were selected as the basic water level sequence. Combined with the average precipitation of the basin, the water storage of the Three Gorges Reservoir (after 1994), and the water storage of the Gezhouba Reservoir (after 1988), an influencing factor matrix was constructed. The three factors with the highest correlation with water level were selected by mutual information method: the average precipitation of the previous three days (correlation coefficient 0.72), the daily discharge of the Three Gorges Reservoir (0.68), and the water surface evaporation during the same period (0.55).

[0136] S3, Implementation of the high and low water level generation algorithm, specifically as follows:

[0137] S301, Construction of Comprehensive Duration Curve

[0138] Using 1950-2020 as the timeline, plot the synchronous variation curves of water level, precipitation in the previous 3 days, and Three Gorges Dam discharge (see...). Figure 2Using a 30-year sliding window (1950-1979, 1955-1984…2000-2029) to analyze the curves, it was identified that after 1994 (when the Three Gorges Reservoir began impounding water), the average water level rose by about 0.8m, and the interannual fluctuation range increased (the standard deviation increased from 0.5m to 0.7m), which was determined to be a point of abrupt change.

[0139] S302, Design Water Level Calculation

[0140] Probability distribution model construction: Select the water level series from 1950 to 2020, fit the generalized extreme value distribution (GEV), and estimate the location parameter μ = 14.2m, the scale parameter σ = 0.6m, and the shape parameter ξ = -0.1 (indicating that the water level is weakly bounded by an upper limit).

[0141] Frequency correction: After considering the abrupt change point in 1994, the data is divided into two segments: 1950-1993 (early period) and 1994-2020 (late period). GEV models are fitted to each segment. The location parameter of the later model is corrected to μ = 15.0m (reflecting the water level rise caused by reservoir impoundment). Finally, the weighted average method (the weight is the proportion of each segment length) is used to obtain the corrected distribution function.

[0142] Quantile calculation: Based on the modified GEV model, the design maximum water level for a 100-year return period is calculated to be 18.5m (17.8m using the traditional method), and the design maximum water level for a 1000-year return period is calculated to be 19.2m (18.1m using the traditional method); the design minimum water level (based on fitting of negative precipitation extreme values) is 2.1m (2.5m using the traditional method).

[0143] S4. Verify the results.

[0144] Using leave-one-out cross-validation, data from 2000-2020 were removed, and the model was trained using data from 1950-1999 to predict the design water level from 2000-2020. The calculated RMSE was 0.3m (RMSE = 0.6m using the traditional method) and NSE was 0.89 (NSE = 0.72 using the traditional method), verifying the accuracy of the method of the present invention.

[0145] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for calculating the design water level of a natural inland river, characterized in that, Includes the following steps: S1. Collect multi-source hydrological data, analyze the collected data, and generate a standardized hydrological database; S2. Screen long-series continuous observation data from hydrological stations, construct a matrix of factors influencing water level, and generate water level data; S3. Construct a comprehensive time-lapse curve of hydrological elements, identify periodicity and abrupt change points, and calculate the design water level for different return periods based on the probability distribution model and combined with the change characteristics correction. S4. Evaluate the accuracy through cross-validation and output the comprehensive duration curve, design water level, and uncertainty range.

2. The method for calculating the design water level of inland natural rivers according to claim 1, characterized in that, Step S1 includes the following steps: S101. Obtain historical hydrological data from multiple hydrological stations along inland natural rivers, as well as multi-source heterogeneous data such as meteorological station and reservoir scheduling data within the basin; S102. Parse the raw data from text, tables, and databases, remove outliers, fill in missing values, and unify the data time resolution; S103. Convert the data of each element according to a unified standard to generate a standardized hydrological database.

3. The method for calculating the design water level of inland natural rivers according to claim 1, characterized in that, Step S2 specifically involves: based on a standardized database, screening long-series continuous observation data from various hydrological stations, and extracting daily or monthly average water levels as the basic water level sequence; combining meteorological elements and human activity influencing factors within the basin to construct a water level influencing factor matrix, and screening key factors that significantly affect water levels using Pearson correlation coefficient and mutual information method.

4. The method for calculating the design water level of inland natural rivers according to claim 1, characterized in that, Step S3 includes the following steps: S301. Using time series as the horizontal axis and key hydrological elements such as water level, flow rate, precipitation, and sediment concentration as the vertical axis, construct a comprehensive time-series curve of hydrological elements that change synchronously with multiple elements. S302. Based on a long series of water level sequences, select a probability distribution model suitable for extreme value data, and fit the model parameters through maximum likelihood estimation or Bayesian method; modify the probability distribution model to obtain the design water level probability density function that considers long-term changes; calculate the design water level corresponding to different return periods based on the modified probability distribution model.

5. The method for calculating the design water level of inland natural rivers according to claim 4, characterized in that, In step S301, the sliding window method is used to perform segmented analysis on the comprehensive time-lapse curve, extracting the extreme values ​​of water level, element extreme values, and corresponding occurrence times within each window, and identifying the periodicity and abrupt change points of water level changes. Specifically, this includes the following steps: S301a, Multidimensional Time Series Modeling: Let the time series be T = {t1, t2, ..., t...} n }, corresponding to the water level sequence H={h1,h2,...,h n The set of key influencing factors X = {X1, X2, ..., X} m }, where X j ={x j1 ,x j2 ,...,x jn The mathematical expression for constructing the comprehensive duration curve is as follows: C(t i )=(h i ,x 1i ,x 2i ,...,x mi ),i=1,2,...,n S301b, Sliding window feature extraction: Define a window function W(k,w,s), where w is the window width and s is the sliding step size. For each window k, calculate: Extreme value characteristics: H k ^{max}=max{h i |t i ∈W(k)} H k ^{min}=min{h i |t i ∈W(k)} Coupling characteristics: ρ jk =corr(H k ,X jk )(j=1,2,...,m) Where corr(·) represents the Pearson correlation coefficient. Mutation point detection: Bayesian variable detection algorithm is used: P(τ|H)∝P(H|τ)P(τ) Where τ is the location of the mutation point, and it is determined to be a significant mutation point when the posterior probability P(τ|H)>0.

95.

6. The method for calculating the design water level of inland natural rivers according to claim 4, characterized in that, In step S302, the probability distribution model is modeled using the generalized extreme value distribution GEV: F(h;μ,σ,ξ)=exp{-[1+ξ(h-μ) / σ]^{-1 / ξ}} Where μ is the position parameter, σ is the scale parameter, and ξ is the shape parameter. Parameter estimation is achieved through maximum likelihood estimation: L(μ,σ,ξ|H)=∏f(h i (m,s,x) Where f(·) is the GEV probability density function.

7. The method for calculating the design water level of inland natural rivers according to claim 4, characterized in that, In step S302, the probability distribution model is modified based on the periodicity and abrupt change characteristics identified in step 301 to obtain the design water level probability density function considering long-term changes. Assuming a sudden change point τ is detected, the data is divided into: D1 = {h i |t i < τ} and D2 = {h i |t i ≥ τ}, The following results were obtained through fitting: F1(h;μ1,σ1,ξ1) and F2(h;μ2,σ2,ξ2) The corrected mixed distribution is: F*(h)=αF1(h)+(1-α)F2(h) Where α = n1 / (n1+n2), and n1 and n2 are the number of segments of data.

8. The method for calculating the design water level of inland natural rivers according to claim 4, characterized in that, In step S302, the design maximum water level and design minimum water level corresponding to different return periods are calculated according to the modified probability distribution model, as shown in the following formula: The design maximum water level corresponding to the return period T: DHWMT=F -1 (1-1 / T) Minimum design water level: DHWMT=F -1 (1 / T) Where F is the corrected probability distribution function and T is the return period.

9. The method for calculating the design water level of inland natural rivers according to claim 1, characterized in that, In step S4, the accuracy of the calculation results is verified by leave-one-out cross-validation. Each time, the data of one hydrological station is removed, and the model is trained and the design water level of the station is predicted using the data of the remaining stations. The root mean square error and Nash efficiency coefficient are calculated. RMSE=\sqrt{\frac{1}{N}\sum_{k=1}^{N}(h_k-\hat{h}_k) 2 } Where hk is the measured water level value of the k-th station, and N is the total number of hydrological stations. NSE = 1 - \frac{\sum_{k=1}^{N}(h_k-\hat{h}_k) 2 }{\sum_{k=1}^{N}(h_k-\bar{h}) 2 } Where, \bar{h}=(1 / N)∑h k The average value is the result of the measured values. The output includes the comprehensive duration curves for each hydrological station, design high and low water levels at different frequencies, and uncertainty intervals. The Bootstrap method is used to calculate the 98% confidence interval. CI=[F -1 (0.01),F -1 (0.99)]。 10. A system for calculating the design water level of an inland natural river, characterized in that, It includes a data acquisition module, a data preprocessing module, a comprehensive duration curve construction module, and a result verification and output module; The data acquisition module is used to access multi-source heterogeneous data from the Yangtze River main stream and its basin, and supports automatic data format recognition and parsing. The data preprocessing module is used to complete data cleaning, missing value imputation, standardization transformation, and element matrix construction. The comprehensive duration curve construction module is based on multi-factor synchronous data and is used to generate a comprehensive duration curve on a time series, and to identify periodicity and abrupt change points. The high and low water level calculation module is used to integrate a probability distribution model and a frequency correction algorithm to calculate the design high and low water levels for different return periods. The result verification and output module evaluates the model accuracy through cross-validation and outputs visualized results.

Images