Method for analyzing evolution characteristics of runoff based on coupling of stochastic hydrology and sequence statistics

Abstract

The application discloses a runoff evolution characteristic analysis method based on coupling of random hydrology and sequence statistics, and belongs to the technical field of water resources engineering. A hydrological time sequence of long-sequence natural runoff is acquired, the hydrological time sequence is preprocessed to generate an analyzable runoff sequence, and is decomposed into a deterministic sequence and a random sequence; sudden change diagnosis, trend analysis and cycle identification of the long-sequence natural runoff are carried out based on the deterministic sequence, and local evolution characteristics are extracted; the random sequence is subjected to stationarity test, probability distribution fitting and random fluctuation characteristic extraction, and random fluctuation rules are acquired; the local evolution characteristics and the random fluctuation rules are coupled, comprehensive quantitative analysis of runoff intra-annual distribution and rich-poor evolution characteristics is completed, and runoff evolution characteristic analysis results are output. The application significantly improves runoff evolution analysis precision and physical rationality, and provides scientific support for fine management of a basin water resource and hydrological risk prevention and control.

Description

Technical Field

[0001] This invention belongs to the technical field of water resources engineering, and in particular relates to a method for analyzing runoff evolution characteristics based on the coupling of stochastic hydrology and sequence statistics. Background Technology

[0002] Current analyses of runoff evolution characteristics primarily rely on single statistical methods or trend tests, which generally suffer from insufficient accuracy in identifying abrupt changes, inadequate trend and periodic analysis, and separation of deterministic patterns from random fluctuations. Existing technologies typically only perform simple trend or abrupt change tests on runoff sequences, failing to meet the comprehensive analytical needs of abrupt change diagnosis, trend assessment, and periodic identification. Furthermore, their ability to couple and analyze the intra-annual distribution and inter-annual abundance / dryness evolution of runoff is weak, especially in long-term runoff evolution analyses at multiple stations along the Yellow River main stream, such as Tangnaihai, Toudaoguai, and Lijin, where unified analysis across multiple stations and time scales is difficult to achieve.

[0003] Meanwhile, conventional methods are prone to misjudging and removing real abrupt change signals during outlier removal, leading to distortion of the regularity of runoff sequences during key abrupt change periods such as 1989-1990. Furthermore, there is a lack of standardized quantitative methods for characterizing random fluctuation components, making it impossible to distinguish the contribution differences between normal fluctuations and abrupt disturbances, and making it difficult to accurately identify the influence mechanisms of external factors such as meteorological disturbances and short-term changes in the watershed underlying surface on runoff fluctuations.

[0004] Furthermore, existing analytical models are mostly characterized by single indicators, making it difficult to achieve hierarchical and progressive analysis of annual distribution characteristics, wet-dry difference characteristics, and overall evolution characteristics. They cannot comprehensively characterize key features such as annual concentration, concentration period, and wet-dry level classification, and cannot provide comprehensive and reliable quantitative basis for runoff trend prediction, wet-dry risk warning, and water resource regulation. They are also unable to meet the actual needs of refined water resource management and hydrological risk prevention and control in the Yellow River Basin. Summary of the Invention

[0005] To address the technical problems existing in the background art, the present invention provides a method for analyzing runoff evolution characteristics based on the coupling of stochastic hydrology and sequence statistics.

[0006] This invention employs the following technical solution: a method for analyzing runoff evolution characteristics based on the coupling of stochastic hydrology and sequence statistics, comprising the following steps: Hydrological time series of long-sequence natural runoff is obtained based on the study area, and the hydrological time series includes continuous hydrological process sequences and discrete hydrological data sequences. The hydrological time series is preprocessed to generate an analyzable runoff series; then, a stochastic hydrological method is used to decompose the analyzable runoff series into deterministic and random sequences. Based on deterministic sequences, we conduct mutation diagnosis, trend analysis, and periodic identification of long-sequence natural runoff to extract local evolution features; for random sequences, we perform stationarity tests, probability distribution fitting, and random fluctuation feature extraction to obtain random fluctuation patterns. By coupling local evolution characteristics with random fluctuation patterns, a comprehensive quantitative analysis of the annual distribution and evolution characteristics of runoff is completed, and the analysis results of runoff evolution characteristics are output. Corresponding runoff trend prediction, early warning of abundant and scarce risks, and auxiliary decision-making for annual distribution regulation are then performed.

[0007] In a further embodiment, the process of generating the analyzable runoff sequence is as follows: Based on hydrological time series Construct a sequence progression difference index: , ; in, These are the runoff sequence values ​​for adjacent time periods. This represents the fluctuation range of runoff between adjacent time periods; This represents the cutoff point for calculating the sequence progressive difference index, indicating the current calculation up to the [number]th ... At a certain point in time, From the start of the sequence to the time point The cumulative fluctuation range ; The runoff fluctuation amplitude was calculated segment by segment for each adjacent time period. and the corresponding cumulative fluctuation range Compare this to the normal fluctuation threshold obtained based on local sequence statistics: When the fluctuation range of a certain period of time Exceeding the normal fluctuation range, and corresponding cumulative fluctuation amplitude When the upward slope of the value changes significantly, the location corresponding to that period is identified as a potential mutation point. This point is then marked separately and permanently retained, and is not included in the subsequent outlier removal criteria.

[0008] In a further embodiment, the deterministic sequence is a systematic evolutionary component containing mutation terms, trend terms, and periodic terms, used for runoff evolution law analysis; The random sequence is the stationary random component remaining after removing deterministic components, used to characterize random fluctuation features.

[0009] In a further embodiment, the extraction process of the local evolution features is as follows: Based on potential mutation sites, cross-validation of mutation sites is completed using multivariate statistical testing methods to screen out the true mutation sites of deterministic sequences; based on the spatiotemporal distribution characteristics of the true mutation sites, the dimensions, types, and core quantitative parameters of runoff mutations are defined to obtain mutation results; Using each real mutation point as the time series segmentation boundary, the deterministic sequence is divided into intervals, which are then decomposed into multiple continuous and independent subsequences. Quantitative trend analysis is carried out on each subsequence, and periodic characteristics are analyzed by combining harmonic analysis and wavelet analysis.

[0010] In a further embodiment, the process of obtaining the random fluctuation pattern is as follows: The random sequence is subjected to residual tracking and fluctuation retrieval methods to obtain the oscillation characteristic points; Based on the aforementioned characteristic points of oscillation changes, determine normal fluctuations and abnormal disturbances: If If the value is within the normal range, it is considered a normal fluctuation point; otherwise, it is considered an abnormal fluctuation point. These are the sequence values ​​corresponding to the oscillation characteristic points. The mean of the random sequence. The standard deviation of the random sequence. Preset fluctuation grading coefficients; Based on the normal fluctuation points, the frequency, amplitude range and distribution characteristics of time series fluctuations are statistically analyzed to analyze the inherent fluctuation evolution law of random sequences. For locations with abnormal disturbances, we will combine meteorological disturbances and short-term changes in the underlying surface of the watershed to trace and analyze the causes and mechanisms of short-term disturbances.

[0011] In a further embodiment, the comprehensive quantitative analysis method for the intra-annual distribution and evolution characteristics of runoff is as follows: Deterministic evolution components are obtained by quantification based on local evolution characteristics and random fluctuation patterns. and random fluctuation components ; A coupled analytical model of annual distribution, abundance / scarcity difference, and comprehensive evolution is constructed, which sequentially completes basic quantification, component correlation, and comprehensive characterization; wherein, the coupled analytical model is expressed in the following form: ; In the formula, The concentration of runoff distribution within the year is used to quantify the uniformity of runoff distribution throughout the year. The first in the year Time period runoff This represents the annual average runoff. This represents the total number of time periods divided within the year. This is the runoff fluctuation coefficient, used to quantify the magnitude of the difference between abundant and scarce runoff. This is the largest runoff volume of the year. This is the minimum runoff volume of the year; This is a comprehensive evolution quantification value used to characterize the overall evolution intensity of runoff. , These are the coupling weight coefficients.

[0012] In a further embodiment, the runoff evolution characteristic analysis results are output in the following form: Based on the concentration of allocation within the year Flood / dryness fluctuation coefficient and comprehensive evolution quantification value The calculation results, combined with preset grading thresholds, are used to output stratified conclusions on runoff evolution characteristics: Output the conclusion on the uniformity of the annual distribution, compare it with the concentration threshold to determine whether the annual distribution of runoff is concentrated or uniform, and output the degree of unevenness in the annual distribution; Output the conclusion on the magnitude of runoff abundance and scarcity, compare it with the threshold of runoff abundance and scarcity fluctuation to determine whether the difference between runoff abundance and scarcity is significant or gradual, and output the corresponding classification results of runoff abundance and scarcity level; The conclusions on the comprehensive evolution driving mechanism are output, and the dual component proportion thresholds are used to clarify whether the runoff evolution is dominated by deterministic evolution, random fluctuations, or both. The stability and alienation level of the overall runoff evolution are also determined.

[0013] In a further embodiment, the deterministic evolution component The quantification method is as follows: The mutation intensity obtained from mutation diagnosis, the trend change rate obtained from trend analysis, and the periodic fluctuation amplitude obtained from period identification are subjected to dimensionless standardization and weighted fusion to obtain deterministic evolution components, which are used to characterize the intensity of runoff being controlled by systematic evolution.

[0014] In a further embodiment, the random fluctuation component The quantification method is as follows: The discreteness of the normal fluctuation points, the frequency of occurrence of abnormal fluctuation points, and the amplitude of disturbances in the random sequence are dimensionlessly standardized and weighted to obtain the random fluctuation component, which is used to characterize the intensity of the influence of random fluctuations and abnormal disturbances on runoff.

[0015] In a further embodiment, the annual allocation and control auxiliary decision-making is to carry out proactive preventive management and control for key future time nodes, specifically including: Based on the identified local evolution characteristics and random fluctuation patterns, key time nodes are divided into flood season, transition period, and dry season; Based on the inherent runoff evolution characteristics of each period, advance control plans are formulated before the arrival of risk periods. By staggering peak water storage in advance, reserving ecological base flow in advance, and controlling water quotas in stages, the risks of concentrated runoff overload during the flood season and the hidden dangers of water shortage during the dry season can be avoided. Based on the source analysis of abnormal fluctuation points, predict the high-incidence periods of meteorological disturbances and short-term changes in the underlying surface of the watershed, and deploy a flexible water allocation mechanism to weaken the impact of random sudden disturbances on the stability of time-series water supply. Based on the conclusions of the comprehensive evolution-driven mechanism, and matching the medium- and long-term evolution trend of runoff, a phased rolling prevention and control strategy is formed.

[0016] The beneficial effects of this invention are as follows: This invention achieves pre-screening of potential mutation points by constructing a sequence progressive difference index, and separately marks and protects the identified mutation points. The outlier removal process does not interfere with the real mutation signal. At the same time, it combines multiple statistical test methods for cross-verification, which greatly reduces the misjudgment rate of mutation points and ensures that the analysis of runoff evolution is not distorted.

[0017] The runoff sequence is decomposed into deterministic and random sequences, and mutation diagnosis, trend analysis, periodic identification, and extraction of normal fluctuations and abnormal disturbances are carried out respectively. This effectively separates the systematic evolution law from the random disturbance component, making the analysis of runoff evolution mechanism clearer and more in line with the actual hydrological and physical processes.

[0018] Dimensionless standardization and weighted fusion are performed on deterministic evolution components and random fluctuation components respectively to achieve unified quantification of multiple features such as mutation, trend, period, fluctuation and disturbance, solving the problems of inconsistent quantification standards and difficulty in horizontal comparison of results in traditional methods.

[0019] By employing three indicators—intra-annual distribution concentration, abundance-dampness fluctuation coefficient, and comprehensive evolution quantification value—a stratified analysis can be conducted, simultaneously characterizing intra-annual distribution, inter-annual abundance-dampness differences, and overall evolution intensity, thus overcoming the shortcomings of existing technologies in terms of incomplete single-indicator representation and insufficient coupled analysis.

[0020] Based on the analysis results, conclusions such as annual distribution characteristics, abundance / dryness level classification, and evolution driving mechanism can be directly output. Simultaneously, runoff trend prediction, abundance / dryness risk warning, and decision support for annual distribution regulation can be achieved. It has both theoretical integrity and engineering practicality, and can provide reliable support for watershed water resources management and hydrological risk prevention and control. Attached Figure Description

[0021] Figure 1 This is a flowchart of the runoff evolution characteristic analysis method based on the coupling of stochastic hydrology and sequence statistics in Example 1.

[0022] Figure 2 This is the Mann-Kendall mutation test curve of the annual runoff in Tang Naihai year in Example 1.

[0023] Figure 3 This is the Mann-Kendall mutation test curve of the annual runoff at Daoguai in Example 1.

[0024] Figure 4 This is the Mann-Kendall mutation test curve of the annual runoff in Lijin in Example 1.

[0025] Figure 5 This is the power spectrum of the runoff in the Tang Naihai year according to the harmonic analysis method in Example 1.

[0026] Figure 6 This is the wavelet variance plot of the runoff in the Tang Naihai year of Example 1. Detailed Implementation

[0027] The present invention will now be further described in conjunction with the embodiments and accompanying drawings.

[0028] Example 1 The runoff evolution characteristic analysis method based on the coupling of stochastic hydrology and sequence statistics described in this embodiment includes the following steps: Hydrological time series of long-sequence natural runoff is obtained based on the study area, and the hydrological time series includes continuous hydrological process sequences and discrete hydrological data sequences. The hydrological time series is preprocessed to generate an analyzable runoff series; then, a stochastic hydrological method is used to decompose the analyzable runoff series into deterministic and random sequences. Based on deterministic sequences, we conduct mutation diagnosis, trend analysis, and periodic identification of long-sequence natural runoff to extract local evolution features; for random sequences, we perform stationarity tests, probability distribution fitting, and random fluctuation feature extraction to obtain random fluctuation patterns. By coupling local evolution characteristics with random fluctuation patterns, a comprehensive quantitative analysis of the annual distribution and evolution characteristics of runoff is completed, and the analysis results of runoff evolution characteristics are output. Corresponding runoff trend prediction, early warning of abundant and scarce risks, and auxiliary decision-making for annual distribution regulation are then performed.

[0029] This embodiment takes the annual natural runoff sequence from 1956 to 2010 at the Tangnaihai Hydrological Station on the main stream of the Yellow River as an example, and the specific implementation is as follows: The Tangnaihai Hydrological Station on the upper reaches of the Yellow River was selected as the research object. Natural annual runoff data of the station from 1956 to 2010 were collected and organized to construct a long-sequence hydrological time series that includes continuous hydrological process sequences and discrete hydrological data sequences.

[0030] Data interpolation and outlier identification preprocessing were performed on the runoff time series from 1956 to 2010. The process for generating the analyzable runoff series is as follows: Based on hydrological time series Construct a sequence progression difference index: , ; in, They are time points and time The runoff sequence values, This represents the fluctuation range of runoff between adjacent time periods; This represents the cutoff point for calculating the sequence progressive difference index, indicating the current calculation up to the [number]th ... At a certain point in time, From the start of the sequence to the time point The cumulative fluctuation range ; The runoff fluctuation amplitude was calculated segment by segment for each adjacent time period. and the corresponding cumulative fluctuation range Compare this to the normal fluctuation threshold obtained based on local sequence statistics: When the fluctuation range of a certain period of time Exceeding the normal fluctuation range, and corresponding cumulative fluctuation amplitude When the upward slope of the value changes significantly, the location corresponding to that period is identified as a potential mutation point (e.g., 1989-1990 is identified as a potential mutation point). This point is then marked separately and permanently retained, and is not included in the subsequent outlier removal criteria.

[0031] By pre-identifying and marking potential mutation sites through the above technical solutions, the real mutation signals can be separately preserved and protected during the subsequent outlier removal process, avoiding the misjudgment and removal of real mutation points as outliers. This ensures that the real mutation characteristics in the runoff sequence are fully preserved, and that the results of subsequent mutation diagnosis, trend analysis and periodic identification are more accurate and reliable.

[0032] Meanwhile, by locking in the abrupt change period in advance, the calculation scope of subsequent multivariate statistical tests can be significantly reduced, the workload of redundant analysis can be reduced, and the overall analysis efficiency can be improved. It can also provide clear time series segmentation basis for subsequent segmented trend judgment and periodic analysis, making the analysis of the evolution law of hydrological time series more in line with the actual physical process, and the analysis results more stable and reliable.

[0033] Therefore, this embodiment also includes: outlier identification and removal. Specifically, local deviation discrimination coefficients are constructed simultaneously. : ; In the formula, For data points The mean of runoff data within the local sliding window. For data points Standard deviation of runoff data within the local sliding window The length parameter of the local sliding window is used to define the local extent (e.g., using...). Centered on the front, take from both the front and back (A window is formed by several data points).

[0034] The above calculation formula is used to calculate the degree of deviation of a single data point from its local window. The greater the deviation, the lower the deviation. The higher the value, the more... Values ​​exceeding a preset threshold (e.g., 3) are considered outliers. It should be noted that subsequent discrimination based on local deviation coefficients... When identifying outliers, the sequence is segmented and locally statistically analyzed using the marked potential mutation points as boundaries. Local means and standard deviations are calculated only for runoff data at non-mutation points, and based on... Outlier identification and removal are performed to ensure that identified mutation points are not misjudged as outliers and removed, thereby accurately preserving the real mutation signals in the runoff sequence while completely eliminating random interference noise.

[0035] In this embodiment, the deterministic sequence is a systematic evolution component containing mutation terms, trend terms, and periodic terms, which is used to reflect the regular evolution of runoff over a long time scale, and can be further used for mutation diagnosis, trend analysis, and periodic identification.

[0036] The random sequence is the stationary random component remaining after removing deterministic components. It does not contain obvious trends or abrupt changes, but only retains random fluctuations and noise components. It is used to carry out stationarity tests, probability distribution fitting, and feature extraction and pattern characterization of normal fluctuations and abnormal disturbances.

[0037] Taking the annual runoff sequence from 1956 to 2010 at the Tangnaihai Hydrological Station on the main stream of the Yellow River as an example, after decomposition, the deterministic sequence shows obvious abrupt changes around 1990, a phased upward trend, and periodic characteristics at multiple time scales of 7 years, 11.8 years, and 24 years, which can be used to reveal the overall evolution law of runoff. The random sequence shows a stationary random process, which can objectively reflect the influence of random factors such as meteorological disturbances and short-term changes in the underlying surface on runoff, making it easier to distinguish between normal fluctuations and abnormal disturbances, and improving the accuracy and physical significance of runoff evolution analysis.

[0038] In a further embodiment, the extraction process of the local evolution features is as follows: Based on potential mutation sites, a multivariate statistical test is used to cross-validate these mutation sites and screen for true mutation points in the deterministic sequence. Based on the spatiotemporal distribution characteristics of these true mutation points, the runoff mutation dimension, mutation type, and core quantitative parameters are defined to obtain the mutation results. The multivariate statistical test methods described in this embodiment are the Mann-Kendall test, ordered cluster analysis, and sliding t-test. Potential mutation sites that pass verification by at least two methods are identified as true mutation points in the runoff sequence.

[0039] Furthermore, the Mann-Kendall test process involves constructing a positive statistic for the time corresponding to the potential mutation point. with inverse statistics And plot the time series statistics curve to determine the time at that moment. and Whether the intersection occurs within the preset significance level threshold; if the intersection occurs, the potential mutation point is determined to be a mutation candidate point that passes the Mann-Kendall test. Furthermore, the process of ordered cluster analysis involves calculating the total sum of squared deviations point-by-point for the deterministic sequence. To obtain the total sum of squared deviations The minimum optimal segmentation time; if the optimal segmentation time coincides with the potential mutation point, it is determined as a mutation candidate point that has passed the test.

[0040] The total sum of squares described in this embodiment The calculation formula is as follows: ;in, The total length of the deterministic sequence, i.e., the total number of data points; For the time index of the sequence, 1 ; For the splitting time (the variable to be optimized), 1 , For deterministic sequences, The first half of the sequence (time 1 to...) The mean of ) For the second half of the sequence (times) to The mean of ).

[0041] Furthermore, the process of the moving t-test is as follows: for the time corresponding to each potential mutation point, construct a statistic. ,like If the significance level threshold is exceeded, the potential mutation site is considered a candidate mutation site that passes the test. The statistic described in this embodiment... The construction formula is: ;in, , These are the means of the two subsequences, respectively. , These represent the number of samples in the two subsequences, respectively. This represents the combined standard deviation of the sequences.

[0042] Using each real mutation point as the temporal segmentation boundary, the deterministic sequence is divided into intervals, forming multiple continuous and independent subsequences. For each subsequence Quantitative trend analysis was conducted one by one, and the periodic characteristics were analyzed by combining harmonic analysis and wavelet analysis.

[0043] Furthermore, the quantitative trend analysis process described in this embodiment is as follows: for each subsequence... The standardized trend statistic is constructed using the following formula: ;in, For the first Standardized trend statistics of segment sequences This is the sequence rank statistic for the corresponding Mann-Kendall test. The variance is the rank statistic.

[0044] Set a uniform significance level threshold Quantitative trend determination is achieved through comparison of segmented statistical values: like The segmented runoff sequence was determined to show a significant upward trend; if The segmented runoff sequence was determined to show a significant downward trend; if It was determined that the segmented sequence did not show a significant temporal evolution trend.

[0045] By combining the positive and negative values ​​and magnitudes of the statistical measures for each segment, the strength and magnitude of the evolution of different segments before and after the abrupt change are compared horizontally to quantify the differences in runoff evolution at different stages.

[0046] Furthermore, power spectral analysis and wavelet analysis were used to identify the periodicity of each segment of the deterministic sequence. Both methods are existing and mature periodic analysis techniques. By calculating the significant periods of the sequence at different time scales, 1 to 3 of the most significant periods were selected as the principal periods to clarify the periodic fluctuation patterns of the runoff sequence in each segment.

[0047] Using the above technical solution, based on potential mutation points, the Mann-Kendall test was first used to analyze the annual runoff at Tangnaihai Station, Toudaoguai Station, and Lijin Station, respectively, and UF-UB statistical curves were plotted to preliminarily identify potential mutation years. For example... Figures 2 to 4It can be seen that the annual runoff at Tangnaihai Hydrological Station may have changed abruptly in 1993, 2008, and 2009; the annual runoff at Toudaoguai Hydrological Station may have changed abruptly in 1990, 1991, and 1992; and the annual runoff at Lijin Hydrological Station may have changed abruptly in 1990, 1991, and 1992.

[0048] Further combining ordered cluster analysis and sliding t-test for multi-method cross-verification and comprehensive judgment, the year 1990 was finally determined to be the common true mutation point of the three hydrological stations.

[0049] Using the actual mutation point in 1990 as the time series segmentation boundary, the deterministic sequence was divided into two independent subsequences: 1956–1989 and 1990–2010. Quantitative trend analysis and periodic feature analysis were carried out for each subsequence.

[0050] Combination Figure 5 and Figure 6 The periodicity identification results of the annual runoff at Tangnaihai Station are as follows: According to Figure 5 The display shows that at frequency Significant amplitude peaks appear at approximately 0.22 and 0.88, corresponding to periods of approximately 28.5 years and 7.1 years, respectively; Figure 6 Significant peaks appeared in the runoff at Tangnaihai station around 11, 24, and 30 years, corresponding to periods of approximately 11.8, 24, and 28.5 years. Cross-validation using two methods ultimately identified three significant main periods—7, 11.8, and 24 years—reflecting the periodic fluctuations in runoff at different time scales. Trend analysis showed that runoff at Tangnaihai station exhibited a significant upward trend both before and after abrupt changes.

[0051] In another embodiment, the process of obtaining the random fluctuation pattern is as follows: The random sequence is subjected to residual tracking and fluctuation retrieval methods to obtain the oscillation characteristic points; Taking the random sequence of the Tangnaihai hydrological station as an example, based on the preset fluctuation classification coefficient... Based on the aforementioned oscillation characteristic points, normal fluctuations and abnormal disturbances are determined: if If the value is within the normal range, it is considered a normal fluctuation point; otherwise, it is considered an abnormal fluctuation point. These are the sequence values ​​corresponding to the oscillation characteristic points. The mean of the random sequence. The standard deviation of the random sequence. Preset fluctuation grading coefficients; Based on the normal fluctuation points, the frequency, amplitude range and distribution characteristics of time series fluctuations are statistically analyzed to analyze the inherent fluctuation evolution law of random sequences. For locations with abnormal disturbances, we will combine meteorological disturbances and short-term changes in the underlying surface of the watershed to trace and analyze the causes and mechanisms of short-term disturbances.

[0052] In this embodiment, the normal fluctuation points of the random sequence at Tangnaihai Station account for about 95%, reflecting the normal fluctuations of meteorological and hydrological conditions; the remaining 5% of abnormal disturbance points are mostly related to factors such as extreme precipitation and short-term disturbances of the watershed underlying surface, and their impact on runoff fluctuations can be further clarified through source tracing analysis.

[0053] Existing runoff evolution analyses often employ single indicators or independent analysis methods, making it difficult to achieve unified quantification and coupled evaluation of intra-annual distribution characteristics, inter-annual abundance / dryness differences, and overall evolution patterns. This results in an insufficiently systematic analysis of runoff evolution mechanisms and makes it difficult to support refined water resource management decisions.

[0054] To address the aforementioned issues, this embodiment presents a comprehensive quantitative analysis method for the intra-annual distribution and evolution characteristics of runoff, as follows: Deterministic evolution components are obtained by quantification based on local evolution characteristics and random fluctuation patterns. and random fluctuation components Furthermore, the deterministic evolutionary component The quantification method is as follows: based on the mutation intensity obtained from mutation diagnosis, the trend change rate obtained from trend analysis, and the periodic fluctuation amplitude obtained from period identification, dimensionless standardization is performed, and weighted fusion is used to obtain deterministic evolution components, which are used to characterize the intensity of runoff being controlled by systematic evolution.

[0055] Random fluctuation components The quantification method is as follows: based on the dispersion of the normal fluctuation points of the random sequence, the frequency of occurrence of abnormal fluctuation points and the amplitude of disturbance, dimensionless standardization is performed, and weighted fusion is used to obtain the random fluctuation component, which is used to characterize the intensity of the influence of random fluctuations and abnormal disturbances on runoff.

[0056] A coupled analytical model of annual distribution, abundance / scarcity difference, and comprehensive evolution is constructed, which sequentially completes basic quantification, component correlation, and comprehensive characterization; wherein, the coupled analytical model is expressed in the following form: ; In the formula, The concentration of runoff distribution within the year is used to quantify the uniformity of runoff distribution throughout the year. The first in the year Time period runoff This represents the annual average runoff. This represents the total number of time periods divided within the year. This is the runoff fluctuation coefficient, used to quantify the magnitude of the difference between abundant and scarce runoff. This is the largest runoff volume of the year. This is the minimum runoff volume of the year; This is a comprehensive evolution quantification value used to characterize the overall evolution intensity of runoff. , The coupling weighting coefficient is determined based on the hydrological characteristics of the watershed.

[0057] Taking the Tangnaihai Hydrological Station on the main stream of the Yellow River as an example, a statistical method of dividing time periods by month is adopted, and the total number of time periods is set. Quantitative calculations were performed by combining long-term measured runoff data. Calculations were performed using measured data from typical representative years, yielding the following results: Station monthly runoff mean Concentration of allocation within the year This indicates that the annual distribution of runoff is concentrated and uneven.

[0058] Furthermore, the maximum runoff within the year Minimum runoff per year The fluctuation coefficient of abundance and scarcity was calculated. This indicates that the stations show significant differences between wet and dry seasons throughout the year, and that hydrological fluctuations are substantial.

[0059] Simultaneously, by combining the results of previous sequence decomposition, the deterministic evolution components were determined. Random fluctuation components Based on the hydrological characteristics of the Yellow River Basin, weighting coefficients were set. , .

[0060] After completing the multi-indicator coupled quantitative calculation, the runoff evolution characteristic analysis results are output in a standardized manner according to hierarchical classification and threshold comparison. Based on the measured calculation results of annual distribution concentration, abundance-dryness fluctuation coefficient and comprehensive evolution quantitative value, combined with the preset watershed adaptation judgment threshold, the runoff evolution characteristic conclusions are output hierarchically and step by step, as follows: This embodiment combines the hydrological characteristics of the Yellow River Basin with industry standards and pre-sets clear judgment thresholds: the concentration threshold is based on... It is of uniform type. It is a centralized type; the threshold for fluctuations in abundance and scarcity is based on It is a gentle type. The threshold for the driving mechanism is: if the proportion of the deterministic evolution component is ≥0.6, it is determined to be deterministically dominant; if the proportion of the random fluctuation component is ≥0.6, it is determined to be stochastically dominant; otherwise, it is jointly dominated by both.

[0061] First, output the conclusion on the uniformity of the annual distribution, compare it with the concentration threshold, clarify whether the annual distribution of runoff is concentrated or uniform, and quantify the degree of uneven distribution within the year; Second, output the conclusion of the variation range of abundant and dry periods, compare it with the threshold of the abundant and dry period fluctuation coefficient, clarify whether the difference between abundant and dry periods is significant or gradual, and combine it with the general standards of the watershed to complete the five levels of abundant and dry periods: exceptionally abundant, moderately abundant, normal, moderately dry, and exceptionally dry. Third, output the conclusion of the comprehensive evolution driving mechanism, compare the proportion threshold of the two components, clarify whether the runoff evolution is dominated by deterministic evolution, random fluctuation, or both, and combine the long-sequence fluctuation amplitude to determine the stability and alienation level of the overall runoff evolution.

[0062] Taking the calculation results of the Tangnaihai Hydrological Station in this embodiment as an example: the annual distribution concentration was calculated. If the percentage is greater than the critical threshold of 30%, the annual runoff distribution is determined to be concentrated, with uneven annual water distribution and prominent runoff aggregation characteristics during the flood season.

[0063] Abundance and drought fluctuation coefficient If the value is greater than the critical threshold of 1.0, the difference between annual and interannual abundance and scarcity is determined to be significant, the alternation of abundance and scarcity in the sequence is obvious, and the annual abundance and scarcity level can be determined based on the runoff anomaly interval. Calculated comprehensive evolution quantification value Based on the comprehensive analysis of various indicators, it can be seen that the runoff evolution at Tangnaihai Station is dominated by deterministic trends, abrupt changes, and cyclical patterns, while the impact of random fluctuations caused by meteorological disturbances and short-term underlying surface changes is relatively limited, thus achieving an integrated quantitative analysis of the multi-dimensional characteristics of runoff.

[0064] The regulation and control auxiliary decision-making described in this embodiment includes flood season regulation and storage optimization, wet and dry season level regulation, medium and long-term runoff forecasting, and water resource supply and demand balance scheduling, providing support for watershed water resource management.

[0065] After completing a comprehensive quantitative analysis of the annual runoff distribution and the characteristics of its abundance and scarcity, based on the evolutionary characteristics conclusions output in a stratified manner, proactive preventive management and control are carried out for key future time nodes, forming an auxiliary decision-making mechanism for annual distribution regulation adapted to the scenario of the upper Yellow River basin.

[0066] Based on the local evolution characteristics and random fluctuation patterns obtained from the long-sequence runoff identification in the upper reaches of the Yellow River in this embodiment, and combined with the natural water flow rhythm of the study area, key time nodes of the flood season, transition period, and dry season are divided, and the seasonal variation characteristics of natural runoff at the Tangnaihai section are matched.

[0067] Based on the inherent runoff evolution characteristics of each period, and considering the current situation of the watershed in the study area, which features highly concentrated runoff during the flood season, uneven distribution throughout the year, and significant interannual alternation of wet and dry seasons, a pre-emptive control plan is formulated before the arrival of each year's risk period: staggered water storage and optimized reservoir capacity scheduling before the arrival of the flood season, pre-reservation of ecological base flow in the river channel during the transition period, and phased water quota control during the dry season. This effectively avoids the risk of flood overload caused by concentrated water inflow in the upper reaches of the Yellow River during the flood season, as well as the potential regional water supply shortage caused by the decline of natural water inflow during the winter and spring dry season.

[0068] Based on the source analysis of the abnormal fluctuations in the random sequence and relying on the long-term observation data of the basin, the high-incidence periods of monsoon fluctuations, periodic meteorological disturbances and short-term changes in the local underlying surface in the upper reaches of the Yellow River are predicted. The flexible water allocation mechanism is deployed in advance to reduce the interference of random sudden disturbances such as abnormal precipitation and local environmental changes on the stability of the basin's temporal water supply and alleviate the hydrological imbalance caused by short-period random fluctuations.

[0069] Based on the conclusions of the comprehensive evolution-driven mechanism and combined with the objective laws governing the runoff at the Tangnaihai section, which are dominated by deterministic trends, abrupt changes, and cyclical evolution, a phased rolling prevention and control strategy is formed by matching the medium- and long-term evolution trend of the basin runoff. Utilizing the cyclical rhythms and evolution trends derived from long-sequence statistical extrapolation, the time-segmented control plan is continuously optimized. Relying on historical evolution patterns, the plan enables early prediction and prevention of hydrological risks, providing technical support for the long-term allocation and seasonal scheduling management of water resources in the upper reaches of the Yellow River.

Claims

1. A method for analyzing runoff evolution characteristics based on the coupling of stochastic hydrology and serial statistics, characterized in that, Includes the following steps: Hydrological time series of long-sequence natural runoff is obtained based on the study area, and the hydrological time series includes continuous hydrological process sequences and discrete hydrological data sequences. The hydrological time series is preprocessed to generate an analyzable runoff series; then, a stochastic hydrological method is used to decompose the analyzable runoff series into deterministic and random sequences. Based on deterministic sequences, we conduct mutation diagnosis, trend analysis, and periodic identification of long-sequence natural runoff to extract local evolution features; for random sequences, we perform stationarity tests, probability distribution fitting, and random fluctuation feature extraction to obtain random fluctuation patterns. By coupling local evolution characteristics with random fluctuation patterns, a comprehensive quantitative analysis of the annual distribution and evolution characteristics of runoff is completed, and the analysis results of runoff evolution characteristics are output. Corresponding runoff trend prediction, early warning of abundant and scarce risks, and auxiliary decision-making for annual distribution regulation are then performed.

2. The method for analyzing runoff evolution characteristics based on the coupling of stochastic hydrology and sequence statistics as described in claim 1, characterized in that, The process of generating the analyzable runoff sequence is as follows: Based on hydrological time series Construct a sequence progression difference index: ， ； in, They are time points and time The runoff sequence values, This represents the fluctuation range of runoff between adjacent time periods; This represents the cutoff point for calculating the sequence progressive difference index, indicating the current calculation up to the [number]th ... At a certain point in time, From the start of the sequence to the time point The cumulative fluctuation range ; The runoff fluctuation amplitude was calculated segment by segment for each adjacent time period. and the corresponding cumulative fluctuation range Compare this to the normal fluctuation threshold obtained based on local sequence statistics: When the fluctuation range of a certain period of time Exceeding the normal fluctuation range, and corresponding cumulative fluctuation amplitude When the upward slope of the value changes significantly, the location corresponding to that period is identified as a potential mutation point. This point is then marked separately and permanently retained, and is not included in the subsequent outlier removal criteria.

3. The method for analyzing runoff evolution characteristics based on the coupling of stochastic hydrology and sequence statistics according to claim 1, characterized in that, The deterministic sequence is a systematic evolutionary component containing mutation terms, trend terms, and periodic terms, used for runoff evolution law analysis; The random sequence is the stationary random component remaining after removing deterministic components, used to characterize random fluctuation features.

4. The method for analyzing runoff evolution characteristics based on the coupling of stochastic hydrology and sequence statistics according to claim 2, characterized in that, The extraction process of the local evolution features is as follows: Based on potential mutation sites, cross-validation of mutation sites is completed using multivariate statistical testing methods to screen out the true mutation sites of deterministic sequences; based on the spatiotemporal distribution characteristics of the true mutation sites, the dimensions, types, and core quantitative parameters of runoff mutations are defined to obtain mutation results; Using each real mutation point as the time series segmentation boundary, the deterministic sequence is divided into intervals, which are then decomposed into multiple continuous and independent subsequences. Quantitative trend analysis is carried out on each subsequence, and periodic characteristics are analyzed by combining harmonic analysis and wavelet analysis.

5. The method for analyzing runoff evolution characteristics based on the coupling of stochastic hydrology and sequence statistics according to claim 2, characterized in that, The process of obtaining the random fluctuation pattern is as follows: The random sequence is subjected to residual tracking and fluctuation retrieval methods to obtain the oscillation characteristic points; Based on the aforementioned characteristic points of oscillation changes, determine normal fluctuations and abnormal disturbances: if If the value is within the normal range, it is considered a normal fluctuation point; otherwise, it is considered an abnormal fluctuation point. These are the sequence values ​​corresponding to the oscillation characteristic points. The mean of the random sequence. The standard deviation of the random sequence. Preset fluctuation grading coefficients; Based on the normal fluctuation points, the frequency, amplitude range and distribution characteristics of time series fluctuations are statistically analyzed to analyze the inherent fluctuation evolution law of random sequences. For locations with abnormal disturbances, we will combine meteorological disturbances and short-term changes in the underlying surface of the watershed to trace and analyze the causes and mechanisms of short-term disturbances.

6. The method for analyzing runoff evolution characteristics based on the coupling of stochastic hydrology and sequence statistics according to claim 1, characterized in that, The comprehensive quantitative analysis method for the intra-annual distribution and evolution characteristics of runoff is as follows: Deterministic evolution components are obtained by quantification based on local evolution characteristics and random fluctuation patterns. and random fluctuation components ; A coupled analytical model of annual distribution, abundance / scarcity difference, and comprehensive evolution is constructed, which sequentially completes basic quantification, component correlation, and comprehensive characterization; wherein, the coupled analytical model is expressed in the following form: ； In the formula, The concentration of runoff distribution within the year is used to quantify the uniformity of runoff distribution throughout the year. The first in the year Time period runoff This represents the annual average runoff. This represents the total number of time periods divided within the year. This is the runoff fluctuation coefficient, used to quantify the magnitude of the difference between abundant and scarce runoff. This is the largest runoff volume of the year. This is the minimum runoff volume of the year; This is a comprehensive evolution quantification value used to characterize the overall evolution intensity of runoff. , These are the coupling weight coefficients.

7. The method for analyzing runoff evolution characteristics based on the coupling of stochastic hydrology and sequence statistics according to claim 6, characterized in that, The analysis results of the runoff evolution characteristics are output in the following format: Based on the concentration of allocation within the year Flood / dryness fluctuation coefficient and comprehensive evolution quantification value The calculation results, combined with preset grading thresholds, are used to output stratified conclusions on runoff evolution characteristics: Output the conclusion on the uniformity of the annual distribution, compare it with the concentration threshold to determine whether the annual distribution of runoff is concentrated or uniform, and output the degree of unevenness in the annual distribution; Output the conclusion on the magnitude of runoff abundance and scarcity, compare it with the threshold of runoff abundance and scarcity fluctuation to determine whether the difference between runoff abundance and scarcity is significant or gradual, and output the corresponding classification results of runoff abundance and scarcity level; The conclusions on the comprehensive evolution driving mechanism are output, and the dual component proportion thresholds are used to clarify whether the runoff evolution is dominated by deterministic evolution, random fluctuations, or both. The stability and alienation level of the overall runoff evolution are also determined.

8. The method for analyzing runoff evolution characteristics based on the coupling of stochastic hydrology and sequence statistics according to claim 6, characterized in that, The deterministic evolution component The quantification method is as follows: The mutation intensity obtained from mutation diagnosis, the trend change rate obtained from trend analysis, and the periodic fluctuation amplitude obtained from period identification are subjected to dimensionless standardization and weighted fusion to obtain deterministic evolution components, which are used to characterize the intensity of runoff being controlled by systematic evolution.

9. The method for analyzing runoff evolution characteristics based on the coupling of stochastic hydrology and sequence statistics according to claim 6, characterized in that, The random fluctuation component The quantification method is as follows: The discreteness of the normal fluctuation points, the frequency of occurrence of abnormal fluctuation points, and the amplitude of disturbances in the random sequence are dimensionlessly standardized and weighted to obtain the random fluctuation component, which is used to characterize the intensity of the influence of random fluctuations and abnormal disturbances on runoff.

10. The method for analyzing runoff evolution characteristics based on the coupling of stochastic hydrology and sequence statistics according to claim 1, characterized in that, The aforementioned annual allocation and control auxiliary decision-making is a proactive preventative management approach for key future time nodes, specifically including: Based on the identified local evolution characteristics and random fluctuation patterns, key time nodes are divided into flood season, transition period, and dry season; Based on the inherent runoff evolution characteristics of each period, advance control plans are formulated before the arrival of risk periods. By staggering peak water storage in advance, reserving ecological base flow in advance, and controlling water quotas in stages, the risks of concentrated runoff overload during the flood season and the hidden dangers of water shortage during the dry season can be avoided. Based on the source analysis of abnormal fluctuation points, predict the high-incidence periods of meteorological disturbances and short-term changes in the underlying surface of the watershed, and deploy a flexible water allocation mechanism to weaken the impact of random sudden disturbances on the stability of time-series water supply. Based on the conclusions of the comprehensive evolution-driven mechanism, and matching the medium- and long-term evolution trend of runoff, a phased rolling prevention and control strategy is formed.

Images