Method and system for identifying spatiotemporal evolution of flood control risk of reservoir groups under runoff variation conditions
By identifying the risk space-time evolution of reservoir groups under runoff changes, using the improved SWAT model and risk space transfer matrix, the shortcomings of reservoir group risk analysis in the existing technology are solved, and comprehensive identification and decision-making guidance for reservoir group risks are achieved.
Patent Information
- Application Number
- CN202411171558.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-26
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-08-26
AI Technical Summary
The existing risk analysis method is difficult to identify the risk spatiotemporal evolution process of reservoir groups under non-consistent runoff changes, and its applicability to large-scale complex reservoir group systems is insufficient, and the impact of coupling between reservoir groups in large-scale development of the basin on the risk analytical means.
A method for identifying the spatial and temporal evolution of reservoir group flood control risks under runoff changes is proposed. By collecting historical runoff data, dividing the main flow and sub-basins, building an improved SWAT model, simulating the runoff scenario, calculating the reservoir flood control indicators and risks, and building a risk spatial transfer probability matrix , identify the spatiotemporal characteristics and development trends of risks.
Break through the limitations of existing methods, identify spatial evolutionary characteristics of complex reservoir groups caused by non-consistent changes in runoff, find spatial risk weaknesses, identify evolutionary characteristics and development trends of reservoir groups in time, and guide Reservoir group risk decision under runoff change conditions.
Smart Images

Figure CN118674281B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a method for identifying the spatiotemporal evolution of flood control risks of a reservoir group under the condition of runoff variation. Background Art
[0002] Under the dual influence of climate change and human activities, the runoff process in the basin has undergone significant changes. The basic assumption of the stationarity of the hydrological sequence will no longer apply. The inconsistent changes in runoff will bring new risks to the comprehensive development and utilization of water resources in the basin and the joint scheduling of water conservancy and hydropower systems.
[0003] Existing risk analysis methods are mainly used to evaluate the probability of occurrence of a certain risk event. They are still unable to identify the spatiotemporal evolution of reservoir group risk under inconsistent runoff changes. Therefore, it is difficult to guide the adaptive regulation of reservoir groups under runoff changes. Moreover, the system scale considered by existing methods is often small, and their applicability to large-scale complex reservoir group systems is insufficient. In particular, there is a lack of analytical methods for the impact of coupling effects between reservoir group systems on risks under large-scale development of river basins.
[0004] The present invention proposes a method and system for identifying the spatiotemporal evolution of flood control risks of a reservoir group under the condition of runoff changes, which solves the above-mentioned problems that exist currently. It can break through the limitations of existing methods, identify the spatial evolution characteristics of complex reservoir group risks caused by inconsistent runoff changes, find spatial risk weaknesses, and identify the temporal evolution characteristics and development trends of reservoir group risks. Summary of the invention
[0005] The purpose of the invention is to provide a method for identifying the spatiotemporal evolution of flood control risks of a reservoir group under runoff variation conditions to solve the above-mentioned problems existing in the prior art. On the other hand, a system for identifying the spatiotemporal evolution of flood control risks of a reservoir group under runoff variation conditions is provided.
[0006] The technical solution is a method for identifying the spatiotemporal evolution of flood control risks of reservoir groups under runoff variation conditions, including the following steps:
[0007] Step S1, collect historical runoff data of the study area, divide the study area into the main stream and several sub-basins, construct the historical runoff series of each sub-basin and divide it into a natural period and several change periods;
[0008] Step S2, constructing an improved SWAT model for the study area, calculating the posterior probability density function of the model parameters corresponding to each runoff sequence, randomly sampling to obtain a massive parameter set corresponding to each historical runoff sequence segment of each sub-basin, using the parameter value combination of the massive parameter set to set the improved SWAT model respectively, simulating the runoff to obtain the massive runoff scenario corresponding to each historical runoff sequence segment of each sub-basin;
[0009] Step S3, using the massive runoff scenarios of each historical runoff sequence segment of the sub-basin to drive the improved reservoir optimization scheduling model in SWAT, obtain the reservoir flood control index value and reservoir discharge flow of the sub-basin, calculate the water level and flow of the flood control section, and calculate the flood control risk corresponding to each reservoir in each historical runoff sequence segment of the sub-basin;
[0010] Step S4: Based on the historical runoff mutation time of each sub-basin, m historical runoff mutation time points of the mainstream are obtained, and the flood control risks corresponding to all reservoirs in the sub-basin where the runoff changes at each historical runoff mutation time point of the mainstream and the mainstream reservoirs affected by them are calculated in turn, and the flood control risk transfer probability from the sub-basin reservoir to the mainstream reservoir is calculated, and the risk space transfer probability matrix is constructed to obtain the risk transfer mechanism.
[0011] According to one aspect of the present application, step S1 further comprises:
[0012] Step S11, collect historical runoff data of the study area, identify the watersheds and rivers in the study area, determine the source of the rivers in the study area, obtain the main stream and n tributaries, take each tributary basin as a sub-basin, and obtain that the study area consists of n sub-basins and the main stream, and the main stream and each sub-basin each include a number of reservoirs, and n is a positive integer;
[0013] Step S12, extract the historical runoff of each sub-basin to construct a corresponding historical runoff sequence, the sequence length is Ti, Ti is a positive integer, i is the sub-basin number, and each historical runoff sequence is divided into a natural period and several change periods based on the historical runoff mutation point.
[0014] According to one aspect of the present application, the step S11 is further:
[0015] Step S11a, read the DEM data of the study area, traverse each grid, take the current grid center as the center, take the larger value between the resolution of the DEM data and the recognition accuracy requirement as the radius to draw a circle, include the grids completely covered by the circle into the inspection range, calculate the absolute value of the maximum elevation difference between the center grid and the grids within the detection range, and record the positive and negative signs of the maximum elevation difference, with the grid center being higher being positive and lower being negative, and record the vector direction at the same time;
[0016] Step S11b, identifying a grid whose maximum elevation difference absolute value is greater than the watershed threshold and whose maximum elevation difference sign is positive and whose adjacent grid maximum elevation difference sign is negative, as a watershed grid, drawing a watershed line based on all watershed grids, and dividing the sub-basins by the watershed line;
[0017] Step S11c, identifying a grid whose maximum elevation difference absolute value is greater than a river threshold and whose maximum elevation difference sign is negative and whose adjacent grid maximum elevation difference sign is positive, as a river grid, drawing rivers based on all river grids, and determining the flow direction based on the vector direction of the river grid;
[0018] Step S11d, moving upstream along the river from the exit of the study area, grid by grid, when the river intersection is identified, implement multi-index joint judgment of the main stream path, identify the main stream path, and continue to move upstream along the main stream, grid by grid, until the main stream end;
[0019] Step S11e, take the tributaries of the main source of the river as the main stream of the study area, and the other tributaries as the tributaries of the study area, to obtain a main stream and n tributaries, and take each tributary basin as a sub-basin, so that the study area consists of the main stream and n sub-basins, the main stream and each sub-basin each include a number of reservoirs, and n is a positive integer.
[0020] According to one aspect of the present application, step S12 further comprises:
[0021] Step S12a, for each sub-basin's historical runoff, calculate the annual mean, the annual daily maximum, daily minimum, monthly maximum, monthly minimum, the annual daily runoff variance, and the monthly runoff variance, and obtain a time series of seven indicators, where the sequence length is Ti, where Ti is a positive integer and i is the sub-basin number;
[0022] Step S12b, construct each time point in the research period of each sub-basin, that is, each year, as a point in a 7-dimensional space, each dimension corresponds to each indicator, and assign the calculated 7 indicator values to each time point;
[0023] Step S12c, for each sub-basin, construct the monitoring point of the time series of the study period from 2 to Ti-1, divide the study period into 2 sections with the monitoring point as the boundary, and calculate the centroid coordinates of the two sections of the sub-sequences before and after the monitoring point in the 7-dimensional space and the Euclidean distance between the two centroids;
[0024] Step S12d: set a threshold, and select monitoring points whose two centroid Euclidean distances are greater than the threshold as mutation points.
[0025] According to one aspect of the present application, step S2 further comprises:
[0026] Step S21, constructing an improved SWAT model of the study area;
[0027] Step S22, taking the hydrological model parameters that obey uniform distribution within their value range as the prior distribution of the first runoff sequence, using the DREAM algorithm with parameter adaptive optimization to perform Bayesian inference on the prior distribution of the first runoff sequence, and obtaining the posterior probability density function of the hydrological model as the model parameters of the first runoff sequence and the prior distribution of the next runoff sequence, and optimizing the DREAM algorithm parameters based on the posterior distribution of the first runoff sequence;
[0028] Step S23, using the posterior distribution of the previous runoff sequence as the prior distribution of the current runoff sequence for Bayesian inference, using the obtained posterior distribution as the prior distribution of the next runoff sequence, and sequentially calculating the posterior probability density function of the hydrological model corresponding to each runoff sequence;
[0029] Step S24, based on the posterior probability density function of the model parameters, randomly sample the massive parameters corresponding to each historical runoff sequence segment of each sub-basin, construct a massive parameter set for the study area, and use the parameter value combination of the massive parameter set to set the improved SWAT model of the study area respectively, simulate the runoff to obtain the massive runoff scenario and the reservoir flood control indicators and reservoir discharge flow of each reservoir under the massive runoff scenario.
[0030] According to one aspect of the present application, step S21 is further:
[0031] Step S21a, constructing the sub-basin SWAT hydrological model in parallel, and optimizing the model according to the actual conditions of the sub-basin boundaries, rivers, reservoirs, etc.;
[0032] Step S21b, construct an optimization dispatching model for each reservoir respectively, replace the reservoir calculation module of the original SWAT model, use the AHPSO algorithm to solve the model, and set flood control parameters;
[0033] Step S21c: Use a variety of sensitivity analysis methods to perform sensitivity analysis on the model parameters of each historical runoff segment of each sub-basin, screen out the sensitive parameters of the SWAT model, and use them together with the parameters of the reservoir optimization scheduling model as key parameters to calibrate the key parameters.
[0034] Step S21d: Based on the source-sink relationship of each sub-basin, the calibrated SWAT models of each sub-basin are coupled to form an improved SWAT model of the entire study area.
[0035] According to one aspect of the present application, step S22 is further:
[0036] Step S22a, using DREAM algorithm parameters and hyperparameters as variables, their initial value ranges as constraints, and the probability distribution of the first p candidate points obtained by the DREAM algorithm being closest to the posterior distribution of the first segment of parameters as the objective function, constructing a DREAM algorithm parameter optimization model, where p is a positive integer greater than 100;
[0037] Step S22b, using a variety of multi-objective optimization algorithms to solve the DREAM algorithm parameter optimization model, to obtain a plurality of non-inferior solution sets, and taking the union of the plurality of non-inferior solution sets as a comprehensive non-inferior solution set of the model parameters;
[0038] Step S22c, construct a multi-attribute decision-making model using multiple methods to obtain the optimal solution of the comprehensive non-inferior solution set of model parameters. When the results obtained by multiple methods are inconsistent and y optimal solutions are obtained, the centroid of the y optimal solutions is obtained as the optimal solution.
[0039] According to one aspect of the present application, step S3 further comprises:
[0040] Step S31, calculating the water level and flow of the flood control section based on the reservoir discharge;
[0041] Step S32: Calculate the flood control risk corresponding to each historical runoff sequence segment of each reservoir in each sub-basin in turn.
[0042] According to one aspect of the present application, step S32 is further:
[0043] Step S31a, constructing a machine learning model for simulating flow and water level of the downstream flood control section;
[0044] Step S31b, extract the historical upstream reservoir discharge, interval inflow and the distance from the upstream reservoir to the downstream prevention and control section and input them into the machine learning model, calculate the simulated flow and water level of the section, compare the actual flow of the section with the simulated flow and water level, and optimize the model parameters;
[0045] Step S31c, extracting the reservoir discharge flow and the distance from the upstream reservoir to the downstream prevention and control section and inputting them into the optimized machine learning model to calculate the flow and water level of the flood control section;
[0046] Step S31d, the water level is calculated. The flow and water level of the flood control section are the flood control safety indicators of the control section.
[0047] According to one aspect of the present application, step S4 is further:
[0048] Step S41, extracting the historical runoff mutation time points of n sub-basins and taking their union as the historical runoff mutation time points of the mainstream, to obtain m historical runoff mutation time points of the mainstream, where m is a positive integer;
[0049] Step S42, sequentially extract all reservoirs in the sub-basin where runoff changes occurred at the time point of historical runoff mutation of each mainstream and the mainstream reservoirs affected by them, and calculate the corresponding flood control risks respectively;
[0050] Step S43: based on the flood control risks of all reservoirs in the sub-basin where the runoff changes occurred at a certain mainstream historical runoff mutation time point and the mainstream reservoirs affected by the sub-basin, the flood control risk transfer probability from the sub-basin reservoir to the mainstream reservoir is calculated, and the flood control risk transfer probability from the sub-basin reservoir to the mainstream reservoir corresponding to m mainstream historical runoff mutation time points is calculated in turn;
[0051] Step S44: construct a risk space transfer probability matrix based on the flood control risk transfer probability from the sub-basin reservoir to the mainstream reservoir corresponding to the m mainstream historical runoff mutation time points to obtain the risk transfer mechanism.
[0052] According to another aspect of the present application, a system for identifying the spatiotemporal evolution of flood control risks of a reservoir group under runoff variation conditions is provided, comprising:
[0053] at least one processor; and
[0054] a memory communicatively connected to at least one of the processors; wherein,
[0055] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the method for identifying the spatiotemporal evolution of flood control risks of a reservoir group under the condition of runoff changes described in any of the above technical solutions.
[0056] Beneficial effects: The spatiotemporal evolution identification method of flood control risks of reservoir groups under runoff changes is adopted to break through the limitations of existing methods, identify the spatial evolution characteristics of complex reservoir group risks caused by inconsistent runoff changes, find spatial risk weaknesses, identify the temporal evolution characteristics and development trends of reservoir group risks, and guide risk decision-making of reservoir groups under current and future runoff changes. BRIEF DESCRIPTION OF THE DRAWINGS
[0057] Figure 1 It is a flow chart of the present invention.
[0058] Figure 2 It is a flow chart of step S1 of the present invention.
[0059] Figure 3 It is a flow chart of step S2 of the present invention.
[0060] Figure 4 It is a flow chart of step S3 of the present invention.
[0061] Figure 5 It is a flow chart of step S4 of the present invention. DETAILED DESCRIPTION
[0062] like Figure 1 According to one aspect of the present application, a method for identifying the spatiotemporal evolution of flood control risks of a reservoir group under runoff variation conditions is provided, characterized in that it includes the following steps:
[0063] Step S1, collect historical runoff data of the study area, divide the study area into the main stream and several sub-basins, construct the historical runoff series of each sub-basin and divide it into a natural period and several change periods;
[0064] Step S2, constructing an improved SWAT model for the study area, calculating the posterior probability density function of the model parameters corresponding to each runoff sequence, randomly sampling to obtain a massive parameter set corresponding to each historical runoff sequence segment of each sub-basin, using the parameter value combination of the massive parameter set to set the improved SWAT model respectively, simulating the runoff to obtain the massive runoff scenario corresponding to each historical runoff sequence segment of each sub-basin;
[0065] Step S3, using the massive runoff scenarios of each historical runoff sequence segment of the sub-basin to drive the improved reservoir optimization scheduling model in SWAT, obtain the reservoir flood control index value and reservoir discharge flow of the sub-basin, calculate the water level and flow of the flood control section, and calculate the flood control risk corresponding to each reservoir in each historical runoff sequence segment of the sub-basin;
[0066] Step S4: Based on the historical runoff mutation time of each sub-basin, m historical runoff mutation time points of the mainstream are obtained, and the flood control risks corresponding to all reservoirs in the sub-basin where the runoff changes at each historical runoff mutation time point of the mainstream and the mainstream reservoirs affected by them are calculated in turn, and the flood control risk transfer probability from the sub-basin reservoir to the mainstream reservoir is calculated, and the risk space transfer probability matrix is constructed to obtain the risk transfer mechanism.
[0067] The present invention first identifies the watersheds and rivers in the study area, obtains the main stream and tributaries through river source identification, divides the study area, constructs the historical runoff sequence of each sub-basin, segments the historical runoff sequence by identifying the mutation points, and uses the union of the runoff mutation time points of each sub-basin as the time series of the spatiotemporal evolution of the flood control risk of the reservoir group;
[0068] Secondly, the sub-basin SWAT hydrological model is constructed in parallel, and the model is optimized according to the actual situation of the sub-basin boundary, river, reservoir and other basins. The optimal operation model is established for each reservoir, and the reservoir calculation module of the original SWAT model is replaced. The key parameters are calibrated, and the calibrated SWAT models of each sub-basin are coupled based on the source-sink relationship of each sub-basin to form an improved SWAT model for the entire study area. The improved SWAT model of the study area is set using the constructed massive parameter set, and the massive runoff scenarios are simulated for the subsequent calculation of the probability of risk occurrence.
[0069] By calculating the risk probability of each sub-basin reservoir and mainstream reservoir at different runoff mutation time points, the evolution of flood control risk of the reservoir group under the condition of runoff change in the time dimension is extracted. By calculating the risk probability of each reservoir in the sub-basin at the same time point and the change in risk probability of the mainstream reservoir affected by it, the risk spatial transfer probability matrix is constructed to obtain the risk transfer mechanism, that is, the evolution of flood control risk of the reservoir group under the condition of runoff change in the spatial dimension, and complete the identification of the spatiotemporal evolution of flood control risk of the reservoir group under the condition of runoff change.
[0070] According to one aspect of the present application, step S1 further comprises:
[0071] Step S11, collect historical runoff data of the study area, identify the watersheds and rivers in the study area, determine the source of the rivers in the study area, obtain the main stream and n tributaries, take each tributary basin as a sub-basin, and obtain that the study area consists of n sub-basins and the main stream, and the main stream and each sub-basin each include a number of reservoirs, and n is a positive integer;
[0072] In this embodiment, the main stream and tributaries are identified by determining the river source, providing a basis for dividing the study area.
[0073] Step S12, extract the historical runoff of each sub-basin to construct a corresponding historical runoff sequence, the sequence length is Ti, Ti is a positive integer, i is the sub-basin number, and each historical runoff sequence is divided into a natural period and several change periods based on the historical runoff mutation point.
[0074] In this embodiment, a high-precision watershed and river identification algorithm based on DEM data combined with a main stream path identification method of multi-index joint judgment is used to achieve accurate division of the study area. Not only the terrain characteristics are taken into account, but also the hydrological principles are integrated to make the division of sub-basins more in line with the actual hydrological process. The mutation point detection method introduced with multi-dimensional indicators realizes the accurate segmentation of the runoff sequence by constructing a time series point in 7-dimensional space and calculating the centroid Euclidean distance. Compared with the traditional single indicator analysis, this method can more comprehensively reflect the changes in runoff characteristics and thus more accurately identify the natural period and change period of runoff. Through these technical innovations, this step can provide more accurate and reliable input data for the subsequent hydrological model construction and risk identification, significantly improving the applicability and accuracy of the entire method. Especially in complex watershed systems, this refined division and segmentation method can effectively capture local changes in hydrological characteristics and provide more detailed and accurate basic information for subsequent flood control risk assessment.
[0075] According to one aspect of the present application, the step S11 is further:
[0076] Step S11a, read the DEM data of the study area, traverse each grid, take the current grid center as the center, take the larger value between the resolution of the DEM data and the recognition accuracy requirement as the radius to draw a circle, include the grids completely covered by the circle into the inspection range, calculate the absolute value of the maximum elevation difference between the center grid and the grids within the detection range, and record the positive and negative signs of the maximum elevation difference, with the grid center being higher being positive and lower being negative, and record the vector direction at the same time;
[0077] Different from the existing method which takes the current grid as the center and calculates the surrounding eight grids, the method adopted in this embodiment can take into account both the resolution and recognition accuracy requirements.
[0078] Step S11b, identifying a grid whose maximum elevation difference absolute value is greater than the watershed threshold and whose maximum elevation difference sign is positive and whose adjacent grid maximum elevation difference sign is negative, as a watershed grid, drawing a watershed line based on all watershed grids, and dividing the sub-basins by the watershed line;
[0079] Step S11c, identifying a grid whose maximum elevation difference absolute value is greater than a river threshold and whose maximum elevation difference sign is negative and whose adjacent grid maximum elevation difference sign is positive, as a river grid, drawing rivers based on all river grids, and determining the flow direction based on the vector direction of the river grid;
[0080] Step S11d, moving upstream along the river from the exit of the study area, grid by grid, when the river intersection is identified, implement multi-index joint judgment of the main stream path, identify the main stream path, and continue to move upstream along the main stream, grid by grid, until the main stream end;
[0081] In a certain embodiment, specifically:
[0082] Traverse every possible river path above the intersection point and calculate its cumulative length and cumulative flow respectively;
[0083] Normalize the cumulative length and cumulative flow of each path to aj and bj, where j represents the jth path, and construct the judgment function sj=αaj+βbj;
[0084] Given the initial weight range, according to the known main and branch ranges of the large basin or adjacent basins where the study basin is located, the parameter optimization model is established with the highest accuracy in identifying the main stream as the objective function, and the values of parameters α and β are calibrated;
[0085] The judgment function with the determined parameters is used to identify the main stream path, and the main stream continues to move upstream grid by grid until the end of the main stream.
[0086] In a certain embodiment, specifically:
[0087] For a certain watershed, two potential main stream paths were identified by moving from the outlet of the study area to the upstream grid by grid.
[0088] Path 1, cumulative length is 80km, cumulative flow is 60m 3 / s;
[0089] Path 2, cumulative length is 120km, cumulative flow is 40m 3 / s;
[0090] After normalization calculation, a1=0.4, b1=0.6; a2=0.6, b2=0.4;
[0091] Construct judgment function sj=αaj+βbj;
[0092] According to the situation of the main and tributary rivers in the adjacent basins of this river basin, the parameters α=0.4 and β=0.6 are determined, so the judgment function is sj=0.4aj+0.6bj. Then the judgment function values of path 1 and path 2 are s1=0.5 and s2=0.48 respectively, so path 1 is judged to be the main stream.
[0093] Step S11e, take the tributaries of the main source of the river as the main stream of the study area, and the other tributaries as the tributaries of the study area, to obtain a main stream and n tributaries, and take each tributary basin as a sub-basin, so that the study area consists of the main stream and n sub-basins, the main stream and each sub-basin each include a number of reservoirs, and n is a positive integer.
[0094] In this embodiment, through innovative terrain analysis and river identification algorithms, high-precision basin division of the study area is achieved, laying a solid foundation for subsequent flood control risk identification. Specifically, this step first adopts a high-precision terrain analysis method based on DEM data. By setting the dynamic inspection range and calculating the elevation difference, this method can more accurately identify terrain features, especially in complex terrain areas, where the advantages of this method are more obvious. A dual identification mechanism of watersheds and rivers is introduced. By setting the watershed threshold and the river threshold, and combining the positive and negative sign judgment of the grid elevation difference, this method can not only accurately identify the watershed position, but also determine the river position and flow direction at the same time. The dual identification mechanism significantly improves the accuracy of basin division, especially in areas with complex terrain and dense river networks, which can better reflect the actual hydrological characteristics. In addition, a multi-index joint judgment method for the main stream path is introduced. Not only the spatial position of the river is considered, but also the hydrological principle is incorporated, which can more accurately identify the main stream path, effectively solving the problem that the traditional method is difficult to accurately identify the main and tributary streams in a complex river network. In short, this step achieves the accurate division of the study area and decomposes the complex basin system into the main stream and multiple sub-basins. This high-precision watershed division method provides a more reliable spatial basis for subsequent hydrological simulation and risk assessment, and significantly improves the applicability of the entire method in complex watershed systems. Especially in large-scale watersheds and complex terrain areas, this refined division method can effectively capture local hydrological characteristics and provide more accurate and detailed spatial information support for the subsequent identification of the spatiotemporal evolution of flood control risks.
[0095] According to one aspect of the present application, step S12 further comprises:
[0096] Step S12a, for each sub-basin's historical runoff, calculate the annual mean, the annual daily maximum, daily minimum, monthly maximum, monthly minimum, the annual daily runoff variance, and the monthly runoff variance, and obtain a time series of seven indicators, where the sequence length is Ti, where Ti is a positive integer and i is the sub-basin number;
[0097] Step S12b, construct each time point in the research period of each sub-basin, that is, each year, as a point in a 7-dimensional space, each dimension corresponds to each indicator, and assign the calculated 7 indicator values to each time point;
[0098] Step S12c, for each sub-basin, construct the monitoring point of the time series of the study period from 2 to Ti-1, divide the study period into 2 sections with the monitoring point as the boundary, and calculate the centroid coordinates of the two sections of the sub-sequences before and after the monitoring point in the 7-dimensional space and the Euclidean distance between the two centroids;
[0099] Step S12d: set a threshold, and select monitoring points whose two centroid Euclidean distances are greater than the threshold as mutation points.
[0100] The runoff mutation point is the time point or interval in the hydrological time series where the runoff flow changes significantly. By identifying the runoff mutation point, the time point when the runoff changes can be found, and the runoff sequence can be divided into segments with different runoff changes, providing a time series for the subsequent identification of the spatiotemporal evolution of flood control risks of reservoir groups under the background of runoff changes.
[0101] Most existing research methods conduct mutation analysis on the annual mean runoff sequence to find mutation points, but ignore extreme values, variances, and runoff changes on different time scales such as month by month and day by day. These runoff changes often have a great impact on the flood control risks of reservoirs and river basins. Therefore, the present invention creatively considers the multi-dimensional indicator characteristics of runoff changes, and adopts a mutation point test method that integrates multi-dimensional indicator characteristics to provide an accurate segmentation basis for subsequent risk analysis.
[0102] In this embodiment, a multidimensional index system is introduced, including 7 indicators such as annual mean, extreme value, and variance. It not only fully reflects the magnitude characteristics of runoff, but also includes the time distribution characteristics, which can more comprehensively characterize the dynamic changes of runoff. Compared with the traditional single index analysis, it can better capture the complex change characteristics of runoff, especially the runoff evolution process under the dual influence of climate change and human activities. A time series point construction method based on 7-dimensional space and a sliding monitoring point technology are introduced. By constructing each time point as a point in 7-dimensional space and using sliding monitoring points to divide subsequences, this method can fully consider the synchronous changes of multiple indicators, and effectively overcome the limitation that the traditional single index analysis method is difficult to identify comprehensive changes. Especially when the runoff change process is complex and a single indicator is difficult to reflect the whole picture, the advantages of this method are more obvious. A mutation point identification method based on Euclidean distance is adopted. By calculating the centroid coordinates and Euclidean distance of the two subsequences before and after the monitoring point in 7-dimensional space, and setting a threshold to screen mutation points, the mutation characteristics of the runoff sequence can be more accurately identified. It not only takes into account the comprehensive changes of multiple indicators, but also can effectively filter out the impact of short-term fluctuations and identify the real long-term change trend. In short, the accurate segmentation of historical runoff sequences is achieved, and the complex runoff evolution process is divided into natural periods and multiple change periods. This high-precision runoff sequence segmentation method provides a more reliable time frame for subsequent hydrological model parameter calibration and flood risk assessment, and significantly improves the accuracy and reliability of the entire method in long-term series analysis. Especially in basins where climate change and human activities are significantly affected, this refined segmentation method can effectively identify runoff characteristics in different periods, and provide key time scale information support for accurately assessing the flood risk of reservoir groups under runoff changes.
[0103] According to one aspect of the present application, step S2 further comprises:
[0104] Step S21, constructing an improved SWAT model of the study area;
[0105] Step S22, taking the hydrological model parameters that obey uniform distribution within their value range as the prior distribution of the first runoff sequence, using the DREAM algorithm with parameter adaptive optimization to perform Bayesian inference on the prior distribution of the first runoff sequence, and obtaining the posterior probability density function of the hydrological model as the model parameters of the first runoff sequence and the prior distribution of the next runoff sequence, and optimizing the DREAM algorithm parameters based on the posterior distribution of the first runoff sequence;
[0106] Step S23, using the posterior distribution of the previous runoff sequence as the prior distribution of the current runoff sequence for Bayesian inference, using the obtained posterior distribution as the prior distribution of the next runoff sequence, and sequentially calculating the posterior probability density function of the hydrological model corresponding to each runoff sequence;
[0107] Step S24, based on the posterior probability density function of the model parameters, randomly sample the massive parameters corresponding to each historical runoff sequence segment of each sub-basin, construct a massive parameter set for the study area, and use the parameter value combination of the massive parameter set to set the improved SWAT model of the study area respectively, simulate the runoff to obtain the massive runoff scenario and the reservoir flood control indicators and reservoir discharge flow of each reservoir under the massive runoff scenario.
[0108] In this embodiment, the method of constructing the sub-basin SWAT model in parallel and optimizing the reservoir operation module enables the model to better adapt to the complex basin system and reservoir group operation characteristics. In particular, the use of the AHPSO algorithm to solve the reservoir optimization operation model effectively improves the accuracy and efficiency of reservoir operation. The parameter adaptive optimization and Bayesian inference method based on the DREAM algorithm are introduced. Not only can the historical runoff information be fully utilized, but also the time-varying characteristics of the parameters can be captured through continuous Bayesian updates. By using the posterior distribution of the previous runoff sequence as the prior distribution of the next segment, this method realizes the dynamic optimization of parameters and effectively solves the problem that the traditional SWAT model is difficult to reflect the runoff changes in parameter calibration. In addition, the introduced DREAM algorithm parameter optimization model further improves the accuracy and efficiency of parameter estimation. By constructing a massive parameter set and simulating a massive runoff scenario, the impact of parameter uncertainty on the model output can be comprehensively evaluated, providing a reliable basis for subsequent risk assessment. In summary, the above innovations enable the improved SWAT model to more accurately simulate the hydrological processes of complex watershed systems under conditions of runoff changes, especially in long time series and multi-scale spaces, providing high-quality simulation data support for the identification of the spatiotemporal evolution of flood control risks in reservoir groups.
[0109] According to one aspect of the present application, step S21 is further:
[0110] Step S21a, constructing the sub-basin SWAT hydrological model in parallel, and optimizing the model according to the actual conditions of the sub-basin boundaries, rivers, reservoirs, etc.;
[0111] Step S21b, construct an optimization dispatching model for each reservoir respectively, replace the reservoir calculation module of the original SWAT model, use the AHPSO algorithm to solve the model, and set flood control parameters;
[0112] Step S21c, respectively use local sensitivity analysis method, variance decomposition method, Monte Carlo simulation and LH-OTA method to perform sensitivity analysis on the model parameters of each historical runoff segment of each sub-basin, screen out the sensitive parameters of the SWAT model, and use them as key parameters together with the parameters of the reservoir optimization scheduling model to calibrate the key parameters.
[0113] Since different sensitivity analysis methods are sensitive to different aspects of the model, using a single method for sensitivity analysis may lead to inaccurate results. Therefore, in this embodiment, the simultaneous use of multiple sensitivity analysis methods can provide a more comprehensive understanding of the model's response to input changes and provide a more reliable data reference for screening sensitive parameters.
[0114] Step S21d: Based on the source-sink relationship of each sub-basin, the calibrated SWAT models of each sub-basin are coupled to form an improved SWAT model of the entire study area.
[0115] The original SWAT model provides two methods for reservoir outflow calculation:
[0116] (1) Given the reservoir discharge;
[0117] (2) Given a storage capacity, outflow occurs when the amount exceeds the storage capacity.
[0118] Both of these two methods cannot simulate the process of reservoir optimization scheduling and cannot meet the flood control risk calculation. Therefore, in this embodiment, the original SWAT model is improved. Since the runoff changes, sensitive parameters, and parameter values of different sub-basins are different, the method of modeling and coupling the sub-basins separately is adopted.
[0119] In this embodiment, based on the method of coupling each sub-basin model with the source-sink relationship, a systematic simulation of the hydrological process in the entire study area is achieved, which enables the improved SWAT model to more accurately simulate the complex hydrological process under the condition of runoff variation, and provides a reliable basis for the flood control risk assessment of the reservoir group.
[0120] According to one aspect of the present application, step S22 is further:
[0121] Step S22a, using DREAM algorithm parameters and hyperparameters as variables, their initial value ranges as constraints, and the probability distribution of the first p candidate points obtained by the DREAM algorithm being closest to the posterior distribution of the first segment of parameters as the objective function, constructing a DREAM algorithm parameter optimization model, where p is a positive integer greater than 100;
[0122] Step S22b, using a variety of multi-objective optimization algorithms to solve the DREAM algorithm parameter optimization model, to obtain a plurality of non-inferior solution sets, and taking the union of the plurality of non-inferior solution sets as a comprehensive non-inferior solution set of the model parameters;
[0123] Step S22c, construct a multi-attribute decision-making model using multiple methods to obtain the optimal solution of the comprehensive non-inferior solution set of model parameters. When the results obtained by multiple methods are inconsistent and y optimal solutions are obtained, the centroid of the y optimal solutions is obtained as the optimal solution.
[0124] The DREAM algorithm is a multi-chain MCMC algorithm that exhibits excellent sampling efficiency in complex, high-dimensional, and multimodal target distributions. It has been proven that it can provide better solutions than commonly used optimization algorithms in high-dimensional search spaces. However, the algorithm requires a large number of iterative calculations and has low computational efficiency. In order to improve the efficiency of the algorithm, the present invention proposes two improvements:
[0125] One is to use the posterior distribution of the previous segment as the posterior distribution of the next segment. The main improvement is not the algorithm, but the initial distribution of the parameters, which converges to the posterior distribution of the parameters more quickly.
[0126] The second improvement to the algorithm is to propose parameter adaptive optimization. According to the posterior distribution of the first segment, the initial parameters of the DREAM algorithm are optimized so that the candidate points it generates are closer to the posterior distribution. This can reduce iterations when performing posterior distribution inference in the next segment.
[0127] Using the DREAM algorithm for Bayesian inference can not only obtain the posterior probability density function of the parameters, but also make full use of prior information to improve the accuracy of parameter estimation. Secondly, the practice of taking the posterior distribution of the previous runoff sequence as the prior distribution of the next section realizes the dynamic optimization of the parameters and effectively captures the time-varying characteristics of the parameters. This is particularly important for simulating runoff changes in long time series. The introduced DREAM algorithm parameter optimization model further improves the robustness of parameter estimation through multi-objective optimization and multi-attribute decision-making. In particular, when the results of multiple methods are inconsistent, the method of solving the center of gravity as the optimal solution can effectively reduce the uncertainty of parameter estimation. These technical innovations enable model parameters to better adapt to runoff change conditions, significantly improve the simulation accuracy of the SWAT model in complex watershed systems, and provide more reliable hydrological simulation results for subsequent flood risk assessment.
[0128] According to one aspect of the present application, step S3 further comprises:
[0129] Step S31, calculating the water level and flow of the flood control section based on the reservoir discharge;
[0130] Step S32: Calculate the flood control risk corresponding to each historical runoff sequence segment of each reservoir in each sub-basin in turn.
[0131] Flood risk mainly considers the conditions of key prevention and control sections and water levels downstream of the reservoir. If the flow or water level of a section of the main river exceeds the flood limit water level or guaranteed water level of the river, flood risk will occur. In theory, the most accurate way to simulate flood evolution is to use a hydrodynamic model. Since the calculation process of the hydrodynamic model is very long, if the hydrodynamic model is coupled with the optimization scheduling model, it is almost unsolvable. Therefore, in this embodiment, the flow of the flood control section downstream of the reservoir is calculated by the discharge flow and interval inflow of the upstream reservoir. Since the water flow requires time to propagate, there is actually a lag. In order to use historical data to calculate the lag, a machine learning model is constructed to fit the flow of the section through the discharge flow of the upstream reservoir, the interval inflow, and the distance from the upstream reservoir to the downstream prevention and control section.
[0132] When the test section is affected by individual or combined factors, the relationship between water level and flow will no longer be a single-valued function relationship, specifically:
[0133] When floods rise and fall, the water surface gradient will change, thus affecting the relationship between water level and flow. Changes in the downstream water body will change the water surface gradient of the test section, making the relationship between water level and flow complicated. When the section is affected by scouring, siltation, freezing or plant growth, the shape of the test section and the roughness of the river channel will change, which will affect the relationship between water level and flow, making the water level too large or too small at the same flow.
[0134] In this embodiment, whether a risk occurs is first determined by the water level, and the probability of the risk occurring is obtained by calculating the ratio of the number of times the risk occurs to the total number of times in a massive runoff scenario.
[0135] In this embodiment, a method for simulating flow and water level of flood control section based on machine learning is introduced. This method not only takes into account the downstream flow and interval inflow of the upstream reservoir, but also innovatively introduces the distance from the reservoir to the section as an input variable to effectively capture the spatial variation characteristics of the water flow evolution process. By optimizing the parameters of the machine learning model, this method can quickly and accurately predict the hydrological elements of the flood control section, providing important support for real-time flood control decision-making. Secondly, the flood control risk calculation method adopted in step S32 comprehensively evaluates the flood control risk of the reservoir group under different hydrological conditions by comprehensively considering multiple historical runoff sequence segments and massive runoff scenarios. This method can not only reflect the flood control capacity of a single reservoir, but also evaluate the overall flood control effect of the reservoir group, providing a scientific basis for the joint dispatch of the reservoir group. Especially under the condition of runoff change, this method can capture the dynamic change characteristics of flood control risk and effectively identify high-risk periods and areas. Through these technical innovations, this step realizes the seamless connection from hydrological simulation to flood control risk assessment, greatly improving the accuracy and timeliness of flood control risk assessment. This has important practical significance for flood control scheduling and risk management of complex river basin systems, and can provide flood control decision makers with more reliable and comprehensive risk information, effectively improving flood control and disaster reduction capabilities.
[0136] According to one aspect of the present application, step S32 is further:
[0137] Step S31a, constructing a machine learning model for simulating flow and water level of the downstream flood control section;
[0138] Step S31b, extract the historical upstream reservoir discharge, interval inflow and the distance from the upstream reservoir to the downstream prevention and control section and input them into the machine learning model, calculate the simulated flow and water level of the section, compare the actual flow of the section with the simulated flow and water level, and optimize the model parameters;
[0139] Step S31c, extracting the reservoir discharge flow and the distance from the upstream reservoir to the downstream prevention and control section and inputting them into the optimized machine learning model to calculate the flow and water level of the flood control section;
[0140] Step S31d, the water level is calculated. The flow and water level of the flood control section are the flood control safety indicators of the control section.
[0141] According to one aspect of the present application, step S4 is further:
[0142] Step S41, extracting the historical runoff mutation time points of n sub-basins and taking their union as the historical runoff mutation time points of the mainstream, to obtain m historical runoff mutation time points of the mainstream, where m is a positive integer;
[0143] Step S42, sequentially extract all reservoirs in the sub-basin where runoff changes occurred at the time point of historical runoff mutation of each mainstream and the mainstream reservoirs affected by them, and calculate the corresponding flood control risks respectively;
[0144] Step S43, based on the flood control risks of all reservoirs in the sub-basin where runoff changes occurred at a historical runoff mutation time point of a main stream and the main stream reservoirs affected by it, calculate the flood control risk transfer probability from the sub-basin reservoir to the main stream reservoir, and calculate the flood control risk transfer probability from the sub-basin reservoir to the main stream reservoir corresponding to m historical runoff mutation time points of the main stream in turn.
[0145] In this embodiment, a certain runoff change moment is first extracted, and the water levels of reservoirs in the sub-basin and the mainstream reservoirs are calculated under the massive runoff scenario. The water level exceeding the threshold is regarded as a flood risk. The probability of risk occurring in each reservoir and flood control section is obtained by calculating the ratio of the number of runoff scenarios with risks to the total number of runoff scenarios. The data of all reservoirs in the sub-basin that affect the water level of the mainstream reservoir are extracted, and the contribution of each reservoir to the risk of the mainstream reservoir is calculated to obtain the probability of flood risk transfer from the basin reservoir to the mainstream reservoir.
[0146] Step S44: construct a risk space transfer probability matrix based on the flood control risk transfer probability from the sub-basin reservoir to the mainstream reservoir corresponding to the m mainstream historical runoff mutation time points to obtain the risk transfer mechanism.
[0147] In this embodiment, the hydrological change characteristics of the entire basin system are effectively captured by extracting the union of the historical runoff mutation time points of the sub-basin as the mutation time point of the mainstream. This method not only considers the local changes in the sub-basin but also reflects the comprehensive impact of these changes on the mainstream, providing a reasonable time frame for subsequent risk transfer analysis. The calculation method of the flood risk transfer probability introduced. By analyzing the flood risk changes of the sub-basin reservoirs and the mainstream reservoirs near the runoff mutation time point, this method can quantify the risk transfer relationship between different reservoirs. This probability-based risk transfer analysis method not only considers the physical connection between reservoirs but also reflects the impact of hydrological changes on risk transfer, effectively solving the problem that traditional methods are difficult to accurately describe the risk transfer law in complex basin systems. By constructing the risk space transfer probability matrix, a systematic expression of the flood risk transfer mechanism of the entire reservoir group is achieved. It not only intuitively shows the risk transfer path in space but also can quantitatively describe the intensity of risk transfer. The comprehensive identification of the spatiotemporal evolution of flood control risks of the reservoir group is achieved. In particular, the ability to accurately capture the dynamic characteristics of risk transfer under the condition of runoff change provides a scientific basis for the joint dispatch of reservoir groups and the formulation of regional flood control strategies.
[0148] The risk transmission mechanism, that is, the evolution of flood control risks of reservoir groups under the condition of runoff changes in the spatial dimension, completes the identification of the spatiotemporal evolution of flood control risks of reservoir groups under the condition of runoff changes.
[0149] In summary, the present invention breaks through the limitations of existing methods, identifies the spatial evolution characteristics of complex reservoir group risks caused by inconsistent changes in runoff, finds spatial risk weaknesses, and identifies the temporal evolution characteristics and development trends of reservoir group risks.
[0150] According to another aspect of the present application, a system for identifying the spatiotemporal evolution of flood control risks of a reservoir group under runoff variation conditions is provided, which is characterized by comprising:
[0151] at least one processor; and
[0152] a memory communicatively connected to at least one of the processors; wherein,
[0153] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the method for identifying the spatiotemporal evolution of flood control risks of a reservoir group under runoff change conditions as described in any of the above items.
[0154] The preferred embodiments of the present invention are described in detail above; however, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all belong to the protection scope of the present invention.
Claims
1. A method for identifying the spatiotemporal evolution of flood control risks of reservoir groups under runoff variation conditions, characterized in that: The steps include: Step S1, collect historical runoff data of the study area, divide the study area into the main stream and several sub-basins, construct the historical runoff series of each sub-basin and divide it into a natural period and several change periods; Step S2, constructing an improved SWAT model for the study area, calculating the posterior probability density function of the model parameters corresponding to each runoff sequence, randomly sampling to obtain a massive parameter set corresponding to each historical runoff sequence segment of each sub-basin, using the parameter value combination of the massive parameter set to set the improved SWAT model respectively, simulating the runoff to obtain the massive runoff scenario corresponding to each historical runoff sequence segment of each sub-basin; Step S3, using the massive runoff scenarios of each historical runoff sequence segment of the sub-basin to drive the improved reservoir optimization scheduling model in SWAT, obtain the reservoir flood control index value and reservoir discharge flow of the sub-basin, calculate the water level and flow of the flood control section, and calculate the flood control risk corresponding to each reservoir in each historical runoff sequence segment of the sub-basin; Step S4: based on the historical runoff mutation time of each sub-basin, m historical runoff mutation time points of the mainstream are obtained, and the flood control risks corresponding to all reservoirs in the sub-basin where the runoff changes occur at each historical runoff mutation time point of the mainstream and the mainstream reservoirs affected by them are calculated in turn, and the flood control risk transfer probability from the sub-basin reservoir to the mainstream reservoir is calculated, and the risk space transfer probability matrix is constructed to obtain the risk transfer mechanism; The step S2 is further as follows: Step S21, constructing an improved SWAT model of the study area; Step S22, taking the hydrological model parameters that obey uniform distribution within their value range as the prior distribution of the first runoff sequence, using the DREAM algorithm with parameter adaptive optimization to perform Bayesian inference on the prior distribution of the first runoff sequence, and obtaining the posterior probability density function of the hydrological model as the model parameters of the first runoff sequence and the prior distribution of the next runoff sequence, and optimizing the DREAM algorithm parameters based on the posterior distribution of the first runoff sequence; Step S23, using the posterior distribution of the previous runoff sequence as the prior distribution of the current runoff sequence for Bayesian inference, using the obtained posterior distribution as the prior distribution of the next runoff sequence, and sequentially calculating the posterior probability density function of the hydrological model corresponding to each runoff sequence; Step S24, based on the posterior probability density function of the model parameters, randomly sampling to obtain the massive parameters corresponding to each historical runoff sequence segment of each sub-basin, constructing the massive parameter set of the study area, using the parameter value combination of the massive parameter set to set the improved SWAT model of the study area respectively, simulating the runoff to obtain the massive runoff scenario and the reservoir flood control index and reservoir discharge flow of each reservoir under the massive runoff scenario; The step S3 is further as follows: Step S31, calculating the water level and flow of the flood control section based on the reservoir discharge; Step S32, sequentially calculating the flood control risk corresponding to each historical runoff sequence segment of each reservoir in each sub-basin; The step S21 is further as follows: Step S21a, constructing the sub-basin SWAT hydrological model in parallel, and optimizing the model according to the actual conditions of the sub-basin boundaries, rivers, and reservoirs; Step S21b, construct an optimization dispatching model for each reservoir respectively, replace the reservoir calculation module of the original SWAT model, use the AHPSO algorithm to solve the model, and set flood control parameters; Step S21c, using a variety of sensitivity analysis methods to perform sensitivity analysis on the model parameters of each historical runoff segment of each sub-basin, screen out the sensitive parameters of the SWAT model, and use them together with the parameters of the reservoir optimization operation model as key parameters to calibrate the key parameters; Step S21d, coupling the SWAT models of each sub-basin based on the source-sink relationship of each sub-basin to form an improved SWAT model of the entire study area; The step S22 is further as follows: Step S22a, using DREAM algorithm parameters and hyperparameters as variables, their initial value ranges as constraints, and the probability distribution of the first p candidate points obtained by the DREAM algorithm being closest to the posterior distribution of the first segment of parameters as the objective function, constructing a DREAM algorithm parameter optimization model, where p is a positive integer greater than 100; Step S22b, using a variety of multi-objective optimization algorithms to solve the DREAM algorithm parameter optimization model, to obtain a plurality of non-inferior solution sets, and taking the union of the plurality of non-inferior solution sets as a comprehensive non-inferior solution set of the model parameters; Step S22c, construct a multi-attribute decision-making model using multiple methods to obtain the optimal solution of the comprehensive non-inferior solution set of model parameters. When the results obtained by multiple methods are inconsistent and y optimal solutions are obtained, the centroid of the y optimal solutions is obtained as the optimal solution.
2. The method for identifying the spatiotemporal evolution of flood control risks of reservoir groups under runoff variation conditions as claimed in claim 1, characterized in that: The step S1 is further as follows: Step S11, collect historical runoff data of the study area, identify the watersheds and rivers in the study area, determine the source of the rivers in the study area, obtain the main stream and n tributaries, take each tributary basin as a sub-basin, and obtain that the study area consists of n sub-basins and the main stream, and the main stream and each sub-basin each include a number of reservoirs, and n is a positive integer; Step S12, extract the historical runoff of each sub-basin to construct a corresponding historical runoff sequence, the sequence length is Ti, Ti is a positive integer, i is the sub-basin number, and each historical runoff sequence is divided into a natural period and several change periods based on the historical runoff mutation point.
3. The method for identifying the spatiotemporal evolution of flood control risk of a reservoir group under runoff variation conditions as claimed in claim 2 is characterized in that: The step S11 is further as follows: Step S11a, read the DEM data of the study area, traverse each grid, take the current grid center as the center, take the larger value between the resolution of the DEM data and the recognition accuracy requirement as the radius to draw a circle, include the grids completely covered by the circle into the inspection range, calculate the absolute value of the maximum elevation difference between the center grid and the grids within the detection range, and record the positive and negative signs of the maximum elevation difference, with the grid center being higher being positive and lower being negative, and record the vector direction at the same time; Step S11b, identifying a grid whose maximum elevation difference absolute value is greater than the watershed threshold and whose maximum elevation difference sign is positive and whose adjacent grid maximum elevation difference sign is negative, as a watershed grid, drawing a watershed line based on all watershed grids, and dividing the sub-basins by the watershed line; Step S11c, identifying a grid whose maximum elevation difference absolute value is greater than a river threshold and whose maximum elevation difference sign is negative and whose adjacent grid maximum elevation difference sign is positive, as a river grid, drawing rivers based on all river grids, and determining the flow direction based on the vector direction of the river grid; Step S11d, moving upstream along the river from the exit of the study area, grid by grid, when the river intersection is identified, implement multi-index joint judgment of the main stream path, identify the main stream path, and continue to move upstream along the main stream, grid by grid, until the main stream end; Step S11e, take the tributaries of the main source of the river as the main stream of the study area, and the other tributaries as the tributaries of the study area, to obtain a main stream and n tributaries, and take each tributary basin as a sub-basin, so that the study area consists of the main stream and n sub-basins, the main stream and each sub-basin each include a number of reservoirs, and n is a positive integer.
4. The method for identifying the spatiotemporal evolution of flood control risks of a reservoir group under runoff variation conditions as claimed in claim 2, characterized in that: The step S12 is further as follows: Step S12a, for each sub-basin's historical runoff, calculate the annual mean, the annual daily maximum, daily minimum, monthly maximum, monthly minimum, the annual daily runoff variance, and the monthly runoff variance, and obtain a time series of seven indicators, where the sequence length is Ti, where Ti is a positive integer and i is the sub-basin number; Step S12b, construct each time point in the research period of each sub-basin, that is, each year, as a point in a 7-dimensional space, each dimension corresponds to each indicator, and assign the calculated 7 indicator values to each time point; Step S12c, for each sub-basin, construct the monitoring point of the time series of the study period from 2 to Ti-1, divide the study period into 2 sections with the monitoring point as the boundary, and calculate the centroid coordinates of the two sections of the sub-sequences before and after the monitoring point in the 7-dimensional space and the Euclidean distance between the two centroids; Step S12d: set a threshold, and select monitoring points whose two centroid Euclidean distances are greater than the threshold as mutation points.
5. The method for identifying the spatiotemporal evolution of flood control risk of a reservoir group under runoff variation conditions as claimed in claim 1, characterized in that: The step S31 is further as follows: Step S31a, constructing a machine learning model for simulating flow and water level of the downstream flood control section; Step S31b, extract the historical upstream reservoir discharge, interval inflow and the distance from the upstream reservoir to the downstream prevention and control section and input them into the machine learning model, calculate the simulated flow and water level of the section, compare the actual flow of the section with the simulated flow and water level, and optimize the model parameters; Step S31c, extracting the reservoir discharge flow and the distance from the upstream reservoir to the downstream prevention and control section and inputting them into the optimized machine learning model to calculate the flow and water level of the flood control section; Step S31d, the water level is calculated. The flow and water level of the flood control section are the flood control safety indicators of the control section.
6. The method for identifying the spatiotemporal evolution of flood control risks of a reservoir group under runoff variation conditions as claimed in claim 1, characterized in that: The step S4 is further as follows: Step S41, extracting the historical runoff mutation time points of n sub-basins and taking their union as the historical runoff mutation time points of the mainstream, to obtain m historical runoff mutation time points of the mainstream, where m is a positive integer; Step S42, sequentially extract all reservoirs in the sub-basin where runoff changes occurred at the time point of historical runoff mutation of each mainstream and the mainstream reservoirs affected by them, and calculate the corresponding flood control risks respectively; Step S43: based on the flood control risks of all reservoirs in the sub-basin where the runoff changes occurred at a certain mainstream historical runoff mutation time point and the mainstream reservoirs affected by the sub-basin, the flood control risk transfer probability from the sub-basin reservoir to the mainstream reservoir is calculated, and the flood control risk transfer probability from the sub-basin reservoir to the mainstream reservoir corresponding to m mainstream historical runoff mutation time points is calculated in turn; Step S44: construct a risk space transfer probability matrix based on the flood control risk transfer probability from the sub-basin reservoir to the mainstream reservoir corresponding to the m mainstream historical runoff mutation time points to obtain the risk transfer mechanism.
7. The spatiotemporal evolution identification system of flood control risk of reservoir groups under runoff change conditions is characterized by: include: at least one processor; as well as a memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the method for identifying the spatiotemporal evolution of flood control risks of a reservoir group under runoff change conditions as described in any one of claims 1 to 6.
