River channel drought alarm flow determination method considering multi-dimensional water demand and staged connection

Abstract

The invention provides a river channel drought alarm flow determination method considering multi-dimensional water demand and staged connection. The method comprises three links of water demand satisfaction degree analysis, staged division and threshold determination. Calculating ecological water demand by adopting a frequency value of 90% of a monthly minimum continuous five-day flow mean value, determining agricultural, living and industrial water demand of each month, introducing a water demand satisfaction degree, and determining a water demand satisfaction degree critical value of different drought grades; taking a month as a minimum staging unit, constructing a staging model of kernel principal component analysis and K-means clustering based on optimization parameters, and determining an intra-year optimal staging; and determining a drought alarm flow threshold value based on the multi-target water demand and the water demand satisfaction critical value, and realizing smooth transition of different staging thresholds based on a water consumption priority principle and a triangular membership function. According to the method, the nonlinear relation of hydrological characteristic data can be effectively processed, the threshold sudden change problem of traditional hard staging is solved, the drought alarm flow threshold is made to better conform to the gradual change characteristic of a hydrological system, and the practicability and adaptability of drought early warning are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of hydrology and water resources management, and particularly relates to a method for determining the dry - warning flow of a river channel considering multi - dimensional water demand and phased connection. Background Art

[0002] The determination of the dry - warning flow needs to comprehensively consider multi - dimensional water use demands such as ecological base flow, agricultural irrigation, industrial water intake, and domestic water use, and has significant spatial heterogeneity and time dynamics. Traditional methods for determining the dry - warning flow of river channels are mostly based on single water use demand or static analysis, and do not fully consider the priority of different water use demands under drought conditions, making it difficult to effectively adapt to the dynamic changes of multi - dimensional water use demands such as ecology, agriculture, industry, and life. In addition, the annual phased division in traditional methods is too simple, mostly using rigid phasing (fixed phasing), which easily leads to unreasonable threshold mutations at the phased connection, does not conform to the gradual response characteristics of the social - hydrological system, and affects the accuracy and practicality of drought warning. The existing technologies mainly have the following deficiencies: ① Do not fully integrate the dynamic response of multi - dimensional water use demands, resulting in the lack of comprehensiveness and adaptability of the dry - warning flow threshold; ② The phasing method does not fully consider the priority of different water use demands under limited water resources; ③ Traditional rigid phasing (fixed phasing) easily leads to threshold mutations at the phased connection, making it difficult to reflect the gradualness of drought warning.

[0003] Therefore, aiming at the deficiencies of the existing technologies, the present invention proposes a method for determining the dry - warning flow of a river channel considering multi - dimensional water demand and phased connection. Compared with the existing technologies, especially compared with the methods relying on ordered segmentation, the method of the present invention can more objectively and flexibly achieve the non - linear phasing of the dry - warning flow, and at the same time, based on the water - use priority principle and the triangular membership function, realize the smooth transition of different phased thresholds, solve the threshold mutation problem of traditional rigid phasing, make the dry - warning flow threshold more conform to the gradual characteristics of the hydrological system, improve the practicality and adaptability of drought warning, and has significant economic and social value. Summary of the Invention

[0004] The present invention proposes a method for determining the dry - warning flow of a river channel considering multi - dimensional water demand and phased connection. This method comprehensively considers the water demands of ecology, agriculture, industry, and life, quantifies the changes in monthly dynamic water demand by introducing the water - demand satisfaction degree; improves the accuracy of non - linear data processing and the objectivity of phasing results based on the KPCA - K - means method; and realizes the smooth transition of different phased thresholds based on the water - use priority principle and the triangular membership function, solving the threshold mutation problem of traditional rigid phasing.

[0005] To achieve the above object, the present invention adopts the following technical solutions: A method for determining the dry - warning flow of a river channel considering multi - dimensional water use demands and phased connection optimization, which includes three links: water - demand satisfaction analysis, phased division, and threshold determination. The main steps are as follows: Step S1, water demand satisfaction analysis: Taking a month as the minimum unit, calculate the ecological water demand for each month using the 90% frequency value of the average flow of the minimum consecutive 5-day flow in a month, and determine the agricultural, domestic, and industrial water demands for each month based on the measured data analysis; introduce the water demand satisfaction and set the critical values of water demand satisfaction for different drought levels. Step S2, stage division: Adopt the method of kernel principal component analysis and K-means clustering to conduct annual staging in units of months, and determine the optimal staging scheme by optimizing the kernel parameters. Step S3, threshold determination: Based on the multi-objective water demand and the critical value of water demand satisfaction, determine the drought warning flow threshold, and achieve the smooth transition of thresholds for different stages according to the water use priority principle and the triangular membership function, and finally obtain the dynamic drought warning flow threshold.

[0006] Further, the water demand satisfaction analysis in step S1 is specifically as follows: Step S11, water demand calculation: The ecological water demand adopts the improved Q 10,5dmin method, that is, calculate using the 90% frequency value of the average flow of the minimum consecutive 5-day flow in a month, the agricultural water demand is determined based on the measured irrigation water use data, and the domestic and industrial water demands are determined based on the data of water plants and industrial water intake users respectively. Step S12, water demand satisfaction determination: The water demand satisfaction is defined as the ratio of the actual water supply to the target water demand, which is used to characterize the satisfaction degree of water use demand under different drought levels. The value range of water demand satisfaction is [0, 1], and the value closer to 1 indicates a higher degree of water demand satisfaction; according to industry standards and the water shortage ratio, set the critical values of water demand satisfaction corresponding to mild, moderate, severe, and extreme droughts.

[0007] Further, the stage division in step S2 is specifically as follows: Step S21, feature index selection: Select 6 indicators including the annual average flow of each month from January to December, the change in adjacent month flow, the annual average precipitation, evaporation, agricultural water demand, and ecological water demand as the staging feature indicators; the calculation of the change in adjacent month flow is as follows: ; Among them, represents the change in adjacent month flow of the th month; represents the annual average flow of the th month, i represents the month within the stage, the value range of is 1 to 12; represents the annual average flow of the Step S22, Construction of KPCA-K-means Model: Construct a staging model based on kernel principal component analysis and K-means clustering. In kernel principal component analysis, the Gaussian radial basis kernel function is used for data mapping, and its mathematical expression is: ; where and are feature index data points, represents the kernel function similarity between feature index data point and feature index data point y, is the square of the Euclidean distance between feature index data point and feature index data point y, is the width parameter of the Gaussian radial basis kernel function, which controls the similarity calculation range between feature index data points; Step S23, Determination of Optimal Staging: Input the staging feature indexes in Step S21 into the KPCA-K-means model constructed in Step S22, traverse and calculate within the effective range of the width parameter σ of the Gaussian radial basis kernel function, perform kernel principal component analysis using the Gaussian radial basis kernel function, map the high-dimensional feature data to a low-dimensional space, select the principal components that retain 90% of the information volume, and calculate the corresponding silhouette coefficient S for each combination of the width parameter σ of the Gaussian radial basis kernel function and the number of clusters K; select the combination of the width parameter σ of the Gaussian radial basis kernel function and the number of clusters K that maximizes the silhouette coefficient S to determine the optimal staging scheme; the calculation formula of the silhouette coefficient S is: ; where is the average distance between a sample point and other points in the same cluster, representing the compactness within the cluster; is the average distance between a sample point and all points in the nearest neighbor cluster, representing the separation between clusters.

[0008] where the sample point refers to the principal component feature data of each month obtained after dimensionality reduction by kernel principal component analysis. The silhouette coefficient S measures the compactness within the cluster and the separation between clusters. The closer the value is to 1, the better the clustering quality.

[0009] Furthermore, in Step S22, the specific optimization process of the KPCA-K-means model is as follows: KPCA-K-means Optimization Process: Adopt the KPCA-K-means optimization algorithm. The KPCA-K-means optimization algorithm realizes the optimal staging of nonlinear hydrological data by iteratively optimizing the width parameter σ of the Gaussian radial basis kernel function and the number of clusters K; The optimal staging of non-linear hydrological data mainly includes inputting staging characteristic indexes and the initial value of the width parameter σ of the Gaussian radial basis kernel function, dimensionality reduction by kernel principal component analysis, K-means clustering, and evaluation of the silhouette coefficient S. The optimal parameter combination is determined by selecting the maximum silhouette coefficient through evaluation.

[0010] Furthermore, in step S3, the determination of the threshold is specifically as follows: Step S31, determination of the basic threshold: Based on the ecological, agricultural, domestic, and industrial water demands of each month within different stages, combined with the critical values of water demand satisfaction corresponding to the corresponding drought levels, the drought warning flow threshold is determined. For the i-th month and drought level g, the upper limit of the drought warning flow threshold The calculation formula is: ; Among them, , , , are respectively the domestic, ecological, agricultural, and industrial water demands of the i-th month; , , are the critical values of water demand satisfaction for ecological, agricultural, and industrial water demands under drought level g; the critical value of domestic water demand satisfaction is fixed at 1.0; Step S32, determination of the threshold within the stage: The average value of the flow thresholds of the same drought level within the same stage is used as the corresponding drought warning flow threshold for the stage; the drought warning flow thresholds corresponding to different drought levels are calculated as follows: ; ; ; ; Among them, M is the number of months within the stage; is the upper limit value of the warning flow for mild drought; is the upper limit value of the warning flow for moderate drought and the lower limit value of the warning flow for mild drought; is the upper limit value of the warning flow for severe drought and the lower limit value of the warning flow for moderate drought; is the upper limit value of the warning flow for extreme drought and the lower limit value of the warning flow for severe drought; i represents the month within the stage; , , , respectively represent the critical values of ecological water demand satisfaction under mild, moderate, severe, and extreme drought levels; , , , respectively represent the critical values of agricultural water demand satisfaction under mild, moderate, severe, and extreme drought levels; , , , respectively represent the critical values of industrial water demand satisfaction under the mild, moderate, severe, and extremely severe drought levels; Step S33, fuzzy transition processing: Introduce a triangular membership function to implement the fuzzy processing of the transition period threshold. The transition period is centered at the boundary between periods and extends 15 days forward and backward, that is, it is defined as a 31-day buffer period between the two sub-periods; For the transition period from the t start th day to the t end th day, define the membership degrees of the adjacent sub-periods on the tth day: ; ; where, μ m (t) represents the membership degree of the tth day to the previous sub-period; μ n (t) represents the membership degree of the tth day to the subsequent sub-period. The sum of the membership degree of the tth day to the previous sub-period and the membership degree of the tth day to the subsequent sub-period is always 1, ensuring the continuity and normalization of the membership function; Step S34, determination of soft sub-period threshold: Expand the drought warning flow threshold within the original sub-period into a daily dynamic threshold within the transition period. The calculation formula is as follows: ; ; ; ; where: , , , are respectively the upper limits of the warning flow thresholds for mild, moderate, severe, and extremely severe droughts within the membership degree μ m of the previous sub-period; , , , are respectively the upper limits of the warning flow thresholds for mild, moderate, severe, and extremely severe droughts within the membership degree μ n of the subsequent sub-period; , , , are respectively the upper limits of the warning flow thresholds for mild, moderate, severe, and extremely severe droughts corresponding to the tth day within the transition period; The warning flow threshold intervals for mild, moderate, severe, and extremely severe droughts within each sub-period are respectively: ( MD , LD , ( SD , MD , ( ED , SD , (0, ED , during the transition period, the drought warning flow thresholds for daily mild, moderate, severe, and extreme droughts are respectively ( MD , LD , ( SD , MD , ( ED , SD , (0, ED .

[0011] Compared with the prior art, the present invention has the following beneficial effects: First, by introducing the water demand satisfaction index, the fine quantification of multi-dimensional water use demands for ecology, agriculture, life, and industry is realized, making the basis for determining the drought warning flow threshold more scientific and reasonable, and better reflecting the actual water use tension under different drought degrees.

[0012] Second, the KPCA-K-means method is used for annual staging. By systematically iteratively optimizing the kernel function width parameter and the number of clusters, and using the clustering mechanism of K-means based on feature similarity, compared with the traditional method and the KPCA-Fisher optimal segmentation method, it can more objectively and flexibly capture and divide the non-linear time-varying characteristics of the drought warning flow, improving the scientificity and accuracy of staging.

[0013] Third, the fuzzy membership function is innovatively introduced to achieve a smooth transition in staging connection, effectively solving the problem of threshold mutation caused by traditional hard staging, and making the drought warning flow threshold more in line with the gradual change characteristics of the hydrological system.

[0014] Fourth, the threshold is determined based on the priority principle of different water resource users, which is more in line with the actual water resource management needs, and significantly improves the operability and adaptability of drought warning.

[0015] In summary, the proposed method can effectively handle the non - linear relationships in hydrological feature data, achieve scientific annual hydrological staging, and resolve the problem of threshold mutation caused by traditional rigid staging through fuzzy transition. This makes the drought warning flow threshold more in line with the gradual change characteristics of the hydrological system and management practice, enhancing the practicality and adaptability of drought warning, and possessing significant economic and social value.

[0016] It is worth noting that when determining the drought warning flow using the method of the present invention, it mainly focuses on the four key water demand categories of ecology, agriculture, domestic use, and industry. In specific applications, other specific in - channel or out - of - channel water demands may need to be further considered according to the functional positioning (such as shipping, power generation, etc.) and management objectives of a specific river, and their impacts on the drought warning flow threshold should be evaluated. In addition, the reliability of the calculation results of this method depends on the accuracy and representativeness of the input historical hydrometeorological data and various water demand data, and the quality of the basic data should be ensured during application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is the flowchart of the method of the present invention.

[0018] Figure 2 It is the flowchart of the KPCA - K - means optimization algorithm of the present invention.

[0019] Figure 3 It is the graph of the daily drought warning flow threshold (soft staging threshold) in the embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0020] The following describes the technical solutions in the embodiments of the present invention clearly and completely with reference to specific embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the protection scope of the present invention.

[0021] As Figure 1 shown, a method for determining the drought warning flow of a river channel considering multi - dimensional water demands and staging connection provided by the present invention includes three links: water demand satisfaction analysis, staging division, and threshold determination. The main steps are as follows: Step S1, water demand satisfaction analysis: Taking a month as the smallest unit, calculate the ecological water demand for each month using the 90% frequency value of the average flow of the minimum continuous 5 - day flow in a month, and determine the agricultural, domestic, and industrial water demands for each month based on the analysis of measured data; introduce the water demand satisfaction degree and set the critical values of water demand satisfaction degrees for different drought levels; Step S2, staging division: Adopt the method of kernel principal component analysis and K - means clustering to conduct annual staging on a monthly basis, and determine the optimal staging plan by optimizing the kernel parameter; Step S3, Threshold determination: Based on the multi-objective water demand and the critical value of water demand satisfaction, determine the drought warning flow threshold, and achieve the smooth transition of different stage thresholds according to the water use priority principle and the triangular membership function, and finally obtain the dynamic threshold of the drought warning flow.

[0022] Furthermore, the analysis of water demand satisfaction in step S1 is specifically as follows: Step S11, Water demand calculation: The ecological water demand adopts the improved Q 10,5dmin method, that is, it is calculated by the 90% frequency value of the average flow of the minimum continuous 5 days per month, the agricultural water demand is determined based on the measured irrigation water use data, and the domestic water demand and industrial water demand are determined based on the data of water plants and industrial water users respectively; Step S12, Water demand satisfaction determination: The water demand satisfaction is defined as the ratio of the actual water supply to the target water demand, which is used to characterize the satisfaction degree of water use demand under different drought levels. The value range of the water demand satisfaction is [0, 1], and the closer the value is to 1, the higher the water demand satisfaction degree; According to the industry standard and the water shortage ratio, set the critical values of water demand satisfaction corresponding to mild, moderate, severe and extreme droughts.

[0023] As Figure 2 shown, furthermore, the stage division in step S2 is specifically as follows: Step S21, Feature index selection: Select six indexes including the average flow of each month from January to December, the change in adjacent month flow, the average annual precipitation, evaporation, agricultural water demand, and ecological water demand as the stage feature indexes; The calculation formula of the change in adjacent month flow is as follows: ; Among them, represents the change in adjacent month flow of the th month; represents the average annual flow of the th month, i represents the month within the stage, and takes values in the range of 1 to 12; represents the average annual flow of the th month; In particular, the change in adjacent month flow in January is equal to the difference between the average annual flows of January and December; Step S22, KPCA-K-means model construction: Construct a stage model based on kernel principal component analysis and K-means clustering, where the kernel principal component analysis uses the Gaussian radial basis kernel function for data mapping, and the mathematical expression is: ; Among them, and are the data points of the feature indexes, represents the kernel function similarity between the feature index data point and the feature index data point y, is the characteristic index data point and the square of the Euclidean distance between the characteristic index data point y, is the width parameter of the Gaussian radial basis kernel function, which controls the similarity calculation range between characteristic index data points.

[0024] Furthermore, the specific process of KPCA-K-means optimization is as follows: KPCA-K-means optimization process: The KPCA-K-means optimization algorithm is adopted. The KPCA-K-means optimization algorithm realizes the optimal staging of nonlinear hydrological data by iteratively optimizing the width parameter σ of the Gaussian radial basis kernel function and the number of clusters K; The optimal staging of nonlinear hydrological data mainly includes inputting the initial values of the staging characteristic index and the width parameter σ of the Gaussian radial basis kernel function, kernel principal component analysis for dimensionality reduction, K-means clustering, and evaluation of the silhouette coefficient S. The optimal parameter combination is determined by evaluating and selecting the maximum silhouette coefficient; Step S23, determination of the optimal staging: Input the staging characteristic index of step S21 into the KPCA-K-means model constructed in step S22, traverse and calculate within the effective range of the width parameter σ of the Gaussian radial basis kernel function, perform kernel principal component analysis using the Gaussian radial basis kernel function, map the high-dimensional feature data to a low-dimensional space, select the principal components that retain 90% of the information volume, and calculate the corresponding silhouette coefficient S for each combination of the width parameter σ of the Gaussian radial basis kernel function and the number of clusters K; Select the combination of the width parameter σ of the Gaussian radial basis kernel function and the number of clusters K that makes the silhouette coefficient S the largest to determine the optimal staging scheme; The calculation formula of the silhouette coefficient S is: ; where, is the average distance between the sample point and other points within the same cluster, representing the compactness within the cluster; is the average distance between the sample point and all points within the nearest neighbor cluster, representing the separation between clusters.

[0025] The above sample points refer to the principal component feature data of each month obtained through kernel principal component analysis for dimensionality reduction. The silhouette coefficient S measures the compactness within the cluster and the separation between clusters. The closer the value is to 1, the better the clustering quality.

[0026] Furthermore, the determination of the threshold in step S3 is specifically as follows: Step S31, determination of the basic threshold: Based on the ecological, agricultural, domestic, and industrial water demands of each month within different stages, combined with the critical value of the water demand satisfaction degree corresponding to the drought level, determine the drought warning flow threshold; For the i-th month and the drought level g, the upper limit of the drought warning flow threshold The calculation formula is: ; Among them, , , , are the domestic, ecological, agricultural, and industrial water demands in the i-th month respectively; , , are the critical satisfaction values of ecological, agricultural, and industrial water demands under drought level g; the critical satisfaction value of domestic water demand is fixed at 1.0.

[0027] Step S32, determination of the threshold within the sub-period: To reduce the interference of extreme values in a single month and ensure the consistency of the threshold within the sub-period, the average value of the flow thresholds of the same drought level within the same sub-period is used as the drought warning flow threshold for the corresponding sub-period; the drought warning flow thresholds corresponding to different drought levels are calculated as follows: ; ; ; ; where M is the number of months within the sub-period; is the upper limit value of the warning flow for mild drought; is the upper limit value of the warning flow for moderate drought and the lower limit value of the warning flow for mild drought; is the upper limit value of the warning flow for severe drought and the lower limit value of the warning flow for moderate drought; is the upper limit value of the warning flow for extreme drought and the lower limit value of the warning flow for severe drought; i represents the month within the sub-period; , , , represent the critical satisfaction values of ecological water demand under mild, moderate, severe, and extreme drought levels respectively; , , , represent the critical satisfaction values of agricultural water demand under mild, moderate, severe, and extreme drought levels respectively; , , , represent the critical satisfaction values of industrial water demand under mild, moderate, severe, and extreme drought levels respectively.

[0028] Step S33, fuzzy transition processing: Introduce a triangular membership function to achieve fuzzy processing of the threshold during the transition period. The transition period is centered at the boundary between sub-periods and extends 15 days forward and backward for each month, that is, it is defined as a 31-day buffer period between the two adjacent sub-periods; For the transition period from the t start -th day to the t end -th day, define the membership degree of the t-th day in the adjacent sub-periods: ; ; wherein, μ m (t) represents the membership degree of the t-th day to the previous sub-period; μ n (t) represents the membership degree of the t-th day to the subsequent sub-period, and the sum of the membership degree of the t-th day to the previous sub-period and the membership degree of the t-th day to the subsequent sub-period is always 1, ensuring the continuity and normalization of the membership function; Step S34, determination of the soft staging threshold: Expand the drought warning flow threshold within the original stage into a daily dynamic threshold during the transition period. The calculation formula is as follows: ; ; ; ; where: , , , are respectively the upper limits of the warning flow thresholds for mild, moderate, severe, and extreme droughts within the membership degree μ m of the previous sub-period; , , , are respectively the upper limits of the warning flow thresholds for mild, moderate, severe, and extreme droughts within the membership degree μ n of the subsequent sub-period; , , , are respectively the upper limits of the warning flow thresholds for mild, moderate, severe, and extreme droughts corresponding to the t-th day during the transition period; The intervals of the drought warning flow thresholds for mild, moderate, severe, and extreme droughts within each stage are respectively: ( MD , LD , ( SD , MD , ( ED , SD , (0, ED , and the daily drought warning flow thresholds for mild, moderate, severe, and extreme droughts during the transition period are respectively ( MD , LD , ( SD , MD , ( ED , SD , (0, ED .

[0029] A hydrological station is located in the river basin of Jiangxi hilly mountainous area. The river channel section where the hydrological station is located has the water demand for ecological, agricultural, industrial and domestic uses, and has no water demand for shipping, power generation, etc. Taking the river channel section where the hydrological station is located as an example, the implementation steps of using the above method to determine the drought warning flow threshold of the section where the hydrological station is located are introduced. Collect the daily runoff from January 1, 1990 to December 31, 2022, the monthly precipitation from 1990 to 2022, and the monthly evaporation from 1990 to 2022 at this station. According to the on-site investigation and statistics, collect the data of industrial water use, domestic water use and agricultural irrigation water use from January to December from 2018 to 2022 at this station. These data comprehensively reflect the runoff situation, climate impact and the time series change characteristics of the main water demand of the research section.

[0030] Step S1, water demand satisfaction analysis: Adopt the improved Q 10,5dmin method, that is, calculate the ecological water demand of each month from January to December respectively with the 90% frequency value of the average of the minimum continuous 5-day flow in a month, and determine the agricultural water demand, industrial water demand and domestic water demand of each month respectively with the average value of the data of industrial water use, domestic water use and agricultural irrigation water use from January to December from 2018 to 2022. See the following table for details.

[0031] Statistical table of water demand of different types in each month

[0032] The above unit: cubic meters per second. According to "Regional Drought Level: GB / T 32135-2015" and the water shortage ratio, set the critical values of water demand satisfaction corresponding to mild, moderate, severe and extreme droughts. Among them, the critical values of ecological water demand satisfaction corresponding to mild, moderate, severe and extreme droughts are taken as 0.8, 0.6, 0.4 and 0.2 respectively; the critical values of agricultural water demand satisfaction corresponding to mild, moderate, severe and extreme droughts are taken as 0.95, 0.8, 0.65 and 0.5 respectively; the critical values of domestic and industrial water demand satisfaction corresponding to mild, moderate, severe and extreme droughts are taken as 0.95, 0.9, 0.8 and 0.7 respectively. See the following table for the range of water demand satisfaction under different drought levels.

[0033] Range of water demand satisfaction under different drought levels

[0034] Step S2, stage division: Select six indicators, namely the multi-year average flow of each month from January to December, the change in flow between adjacent months, the multi-year average precipitation, evaporation, agricultural water demand, and ecological water demand, as the stage characteristic indicators. Among them, the agricultural water demand and ecological water demand of each month are obtained from Step S1. The stage characteristic indicators are shown in the following table: Stage index values of drought warning flow

[0035] Input the above stage characteristic values into the constructed KPCA-K-means model. The value range of the width parameter σ of the Gaussian radial basis kernel function is set to 0.01 - 10, the step size of traversing the width parameter σ of the Gaussian radial basis kernel function is set to 0.01, and a total of 1000 traversal calculations are performed. Through the optimization of the width parameter σ of the Gaussian radial basis kernel function, when σ = 5.84 finally, the silhouette coefficient S of the Kmeans clustering result (K = 3) reaches the maximum value of 0.56, and the contribution rates of the corresponding six principal components are 0.516, 0.384, 0.04, 0.025, 0.017, and 0.018 in sequence. Among them, the cumulative contribution rate of the first two items reaches 90%. That is, extract the characteristic data of two principal components and input them into K-means clustering analysis. At this time, the corresponding optimal stages are divided into three categories: April - June, July - October, and November - March.

[0036] Step S3, threshold determination: Fix the critical value of domestic water demand satisfaction at 1.0, and use the average value of the flow thresholds of the same drought level within the same stage as the drought warning flow threshold of this stage and this level. The drought warning flow thresholds corresponding to different drought levels are calculated as follows: ; ; ; ; Among them, M is the number of months within the stage; is the upper limit value of the warning flow for mild drought; is the upper limit value of the warning flow for moderate drought and the lower limit value of the warning flow for mild drought; is the upper limit value of the warning flow for severe drought and the lower limit value of the warning flow for moderate drought; is the upper limit value of the warning flow for extreme drought and the lower limit value of the warning flow for severe drought; i represents the month within the stage.

[0037] The drought warning flow thresholds (hard stage thresholds) of each stage are obtained from the above formulas, as shown in the following table specifically.

[0038] Dry warning flow thresholds for different periods (hard staging thresholds)

[0039] Note: The above table shows the hard staging thresholds of dry warning flow corresponding to different drought levels, and the above unit is cubic meters per second.

[0040] As Figure 3 shown, based on the above hard staging thresholds, a triangular membership function is introduced to realize the fuzzy processing of the transition period thresholds. According to the staging results determined in step S2 and the above hard staging thresholds, there are sudden changes in the dry warning flow thresholds at the junctions of March / April, June / July, and October / November in this embodiment. The transition periods extend 15 days forward and backward from the cross-period boundaries as the center, that is, the three transition periods are from March 17th to April 16th, from June 16th to July 16th, and from October 17th to November 16th respectively. The daily dry warning flow thresholds for the transition periods are obtained by using the triangular membership function respectively, and combined with the above hard staging thresholds, the annual daily dry warning flow thresholds (soft staging thresholds) of this hydrological station are obtained.

Claims

1. A method for determining river drought warning flow considering multi-dimensional water demand and phased connection, characterized by: The following steps are involved: Step S1, water demand satisfaction analysis: taking the month as the smallest unit, using the 90% frequency value of the monthly minimum continuous 5-day flow average to calculate the ecological water demand of each month, and determining the agricultural, domestic and industrial water demand of each month based on the measured data analysis; introducing water demand satisfaction, and setting the critical value of water demand satisfaction for different drought levels; Step S2, staging: kernel principal component analysis and K-means clustering are used to perform annual staging in units of months, and the optimal staging scheme is determined by optimizing kernel parameters; Step S3, threshold determination: Based on the multi-objective water demand and water demand satisfaction critical values, the drought warning flow threshold is determined, and the smooth transition of different stage thresholds is achieved according to the water use priority principle and the triangular membership function, and finally the drought warning flow dynamic threshold is obtained.

2. A method for determining river drought warning flow considering multi-dimensional water demand and phased connection according to claim 1, characterized in that: The water demand satisfaction analysis in step S1 is specifically as follows: Step S11, water demand calculation: ecological water demand adopts improved Q 10,5dmin The method is to calculate the water demand based on the 90% frequency value of the minimum five-day average flow rate in a month. The agricultural water demand is determined based on the measured irrigation water use data, and the domestic water demand and industrial water demand are determined based on the data of water plants and industrial water users respectively. Step S12, water demand satisfaction is determined: the water demand satisfaction is defined as the ratio of actual water supply to target water demand, which is used to characterize the satisfaction degree of water demand under different drought levels. The value range of water demand satisfaction is [0,1]. The closer the value is to 1, the higher the water demand satisfaction degree is. According to industry standards and water shortage ratios, the critical values ​​of water demand satisfaction corresponding to mild, moderate, severe and extreme droughts are set.

3. A method for determining river drought warning flow considering multi-dimensional water demand and phased connection according to claim 2, characterized in that: The phase division in step S2 is specifically as follows: Step S21, characteristic index selection: select six indicators from January to December, including the multi-year average flow, the change in flow in the adjacent month, the multi-year average precipitation, evaporation, agricultural water demand, and ecological water demand as phase characteristic indicators; the change in flow in the adjacent month is calculated as follows: ； in, Indicates The change in flow between the month and the adjacent month; Indicates The multi-year average flow of the month, i represents the month in the period, The value range is 1~12; Indicates Monthly multi-year average flow; Step S22, KPCA-K-means model construction: construct a staging model based on kernel principal component analysis and K-means clustering, wherein the kernel principal component analysis uses Gaussian radial basis kernel function for data mapping, and the mathematical expression is: ； in, and is the characteristic indicator data point, Represents feature indicator data points and the kernel function similarity of the feature index data point y, is the characteristic indicator data point The square of the Euclidean distance of the feature index data point y, is the width parameter of the Gaussian radial basis kernel function, which controls the similarity calculation range between feature index data points; Step S23, optimal staging determination: input the staging characteristic index of step S21 into the KPCA-K-means model constructed in step S22, traverse and calculate within the effective range of the Gaussian radial basis kernel function width parameter σ, use the Gaussian radial basis kernel function to perform kernel principal component analysis, map the high-dimensional feature data to the low-dimensional space, select the principal component that retains 90% of the information, and calculate the corresponding silhouette coefficient S for each combination of the Gaussian radial basis kernel function width parameter σ and the number of clusters K; select the combination of the Gaussian radial basis kernel function width parameter σ and the number of clusters K that maximizes the silhouette coefficient S to determine the optimal staging plan; the calculation formula of the silhouette coefficient S is: ； in, is the average distance between the sample point and other points in the same cluster, indicating the compactness within the cluster; It is the average distance between the sample point and all points in the nearest neighbor cluster, indicating the degree of separation between clusters.

4. A method for determining river drought warning flow considering multi-dimensional water demand and phased connection according to claim 3, characterized in that: In step S22, the KPCA-K-means model optimization process is specifically as follows: KPCA-K-means optimization process: The KPCA-K-means optimization algorithm is used. The KPCA-K-means optimization algorithm achieves the optimal periodization of nonlinear hydrological data by iteratively optimizing the Gaussian radial basis kernel function width parameter σ and the number of clusters K; The optimal staging of nonlinear hydrological data mainly includes input staging characteristic indicators and the initial value of the Gaussian radial basis kernel function width parameter σ, kernel principal component analysis dimensionality reduction, K-means clustering and silhouette coefficient S evaluation. The optimal parameter combination is determined by evaluating and selecting the largest silhouette coefficient.

5. A method for determining river drought warning flow considering multi-dimensional water demand and phased connection according to claim 4, characterized in that: The threshold value is determined in step S3, specifically: Step S31, basic threshold determination: based on the ecological, agricultural, domestic and industrial water demand in each month in different periods, combined with the critical value of water demand satisfaction corresponding to the drought level, determine the drought warning flow threshold. For the i-th month and drought level g, the upper limit of the drought warning flow threshold is The calculation formula is: ； in, , , , are the water requirements for life, ecology, agriculture and industry in the ith month respectively; , , is the critical value of ecological, agricultural and industrial water demand satisfaction under drought level g; the critical value of domestic water demand satisfaction is fixed at 1.0; Step S32, determining the threshold within each period: using the average flow threshold of the same drought level within the same period as the drought warning flow threshold of the corresponding period; the drought warning flow threshold corresponding to different drought levels is calculated as follows: ； ； ； ； Where M is the number of months in the installment; It is the upper limit of the flow rate for mild drought warning; The upper limit of moderate drought warning flow and the lower limit of mild drought warning flow; The upper limit of the severe drought warning flow and the lower limit of the moderate drought warning flow; is the upper limit of the flow warning for extremely severe drought and the lower limit of the flow warning for severe drought; i represents the month within the period; , , , They represent the critical values ​​of ecological water demand satisfaction under mild, moderate, severe and extremely severe drought levels respectively; , , , They represent the critical values ​​of agricultural water demand satisfaction under mild, moderate, severe and extremely severe drought levels respectively; , , , They represent the critical values ​​of industrial water demand satisfaction under mild, moderate, severe and extreme drought levels respectively; Step S33, fuzzy transition processing: introducing a triangular membership function to implement fuzzy processing of the transition period threshold, the transition period is centered at the inter-period boundary, and extends 15 days forward and backward each month, that is, it is defined as a 31-day buffer period between the two periods; For the transition period from t start Day to day end Days, define the membership of the adjacent periods on day t: ； ； Among them, µ m (t) represents the membership degree of the tth day to the previous period; µ n (t) represents the membership of the t-th day to the next period. The sum of the membership of the t-th day to the previous period and the membership of the t-th day to the next period is always 1, ensuring the continuity and normalization of the membership function. Step S34, soft stage threshold determination: the drought warning flow threshold in the original stage is expanded to a daily dynamic threshold in the transition period, and the calculation formula is as follows: ； ； ； ； in: , , , are the membership degree µ of the previous period respectively. m The upper limits of warning flow thresholds for mild, moderate, severe and extreme droughts; , , , are the membership degree µ of the next period respectively. n The upper limits of warning flow thresholds for mild, moderate, severe and extreme droughts; , , , are the upper limits of the mild, moderate, severe and extremely severe drought warning flow thresholds corresponding to day t during the transition period; The drought warning flow threshold intervals for mild, moderate, severe and extremely severe drought in each stage are: ( MD , LD ],( SD , MD ],( ED , SD ],(0, ED ], the daily drought warning flow thresholds for mild, moderate, severe and extreme drought during the transition period are ( MD , LD ],( SD , MD ],( ED , SD ],(0, ED ].

Images