A coupled reservoir-lake system dispatching method for excess flood distribution at component level

Abstract

The present invention discloses a reservoir-lake system dispatching method for coupled component-level excess flood volume allocation, including: collecting hydrological data and engineering data, calculating the reservoir inflow at each time step, the response process of the reservoir inflow at the control section, the response process of the measured outflow at the control section, the reservoir interval water inflow process, the lake interval water inflow process, and the system natural water inflow process; calculating the water volume of the system natural water inflow process exceeding the initial drainage capacity of the lake, defining the concept of the system initial excess flood volume; constructing a reservoir-lake system flood control and drainage joint dispatching model based on coupled component-level excess flood volume allocation, and using an improved whale-particle swarm hybrid optimization algorithm to obtain a system optimization dispatching plan. The present invention can quickly provide effective and reliable reference information for the reservoir-lake system flood control and drainage decision-making.

Description

Technical Field

[0001] The invention of the present application relates to the technical field of flood control and drainage scheduling, and in particular to a reservoir-lake system scheduling method for coupled component-level excess flood volume distribution. Background Art

[0002] The flood control and drainage system, which is basically composed of upstream reservoirs, midstream lakes and external drainage channels, usually needs to coordinate the storage and discharge methods of the system to facilitate timely flood discharge and drainage. The joint dispatch of "storage and discharge" requires reservoirs and lakes to implement integrated dispatch of flood control and drainage according to the water level of the external river. However, when the system encounters multiple or double encounters such as mountain torrents, large internal floods, and high water levels in external rivers, the system will face unfavorable situations such as flood clamping and blocked drainage channels, and there is a risk of flood disasters. Coordinating the joint dispatch of reservoirs, lakes and other engineering systems is an important measure to reduce disaster losses.

[0003] In essence, the combined dispatch of flood control and drainage is a type of large-scale, multi-objective, nonlinear optimization problem. At present, most of the modeling studies on similar problems adopt the method of directly constructing a multi-objective optimization model and solving it. However, direct multi-objective optimization has problems such as low computational efficiency, a large number of schemes, and difficulty in distinguishing the coordination relationship between floods and floods, which limits its practical application in the combined dispatch of flood control and drainage in the basin. In fact, in the actual dispatching process, when the upstream reservoir still has a certain flood control safety threshold, decision makers are often more concerned about the situation of downstream flooding and need to find a solution to reduce the losses of flooding as much as possible with an appropriate or minimum flood control storage capacity. However, it is difficult to quickly focus on this solution by directly constructing a multi-objective optimization model. The essence of integrated dispatching of flood control and drainage is to reasonably allocate excess flood volume, and to absorb excess flood volume by blocking upstream reservoirs, regulating midstream lakes, and discharging downstream rivers. Summary of the invention

[0004] The purpose of the invention is to provide a reservoir-lake system scheduling method for coupled component-level excess flood distribution to solve the above-mentioned problems existing in the prior art.

[0005] The technical solution is a reservoir-lake system dispatching method for coupling component-level excess flood volume allocation, comprising the following steps:

[0006] Step S1, collecting hydrological data and engineering data of the reservoir-lake system, and calculating the reservoir inflow at each time step, the response process of the reservoir inflow at the control section, the response process of the measured outflow at the control section, the water inflow process between the reservoirs, the water inflow process between the lakes, and the natural water inflow process of the system based on the above data;

[0007] Step S2: Based on the hydrological data and engineering data, the LSTM model of coupled physical principle is used for fitting, and the fitting results are arranged in time series to obtain the initial drainage capacity process of the lake. Based on the natural water inflow process of the system and the initial drainage capacity process of the lake, the initial excess flood volume of the system is calculated; wherein the initial excess flood volume of the system is defined as the amount of water that exceeds the initial drainage capacity process of the lake by the natural water inflow process of the system;

[0008] Step S3: Based on the hydrological data, engineering data and reservoir inflow, the water inflow of each reservoir's excess flood storage section, the remaining flood control storage capacity of each reservoir, the total water inflow of the lake's excess flood storage section, and the flood storage capacity corresponding to the lake warning water level are calculated, and the flood storage capacity of each reservoir and the flood storage capacity of the lake are calculated in turn based on the above data. Based on the natural water inflow process of the system, the improved XGBoost classification model based on the cost function is used to classify the floods. When 0≤excess flood volume< W 1, the flood is classified as the first magnitude type. W 1≤Excess Flood Volume< W 2:00 the flood is classified as the second magnitude type. W When 2≤excess flood volume, the flood is divided into the third magnitude type. Based on the initial excess flood volume of the system and the flood classification results, the excess flood volume generated by floods of different magnitude types is allocated, and the excess flood volume allocation results are calculated and verified to obtain the excess flood volume; wherein the excess flood volume is composed of the excess flood volume allocation results, and the natural water inflow process of the system includes at least one flood;

[0009] Step S4, based on the excess flood volume, obtain the water level constraint of the pre-constructed reservoir-lake system flood control and drainage joint scheduling model based on coupled component-level excess flood volume allocation, read the reservoir inflow, reservoir interval water inflow process, lake interval water inflow process input model, and use the improved whale-particle swarm hybrid optimization algorithm to solve and obtain the reservoir-lake system flood control and drainage joint scheduling result.

[0010] According to one aspect of the present application, the specific output of the present invention is the joint optimization scheduling result of flood control and drainage in the reservoir-lake system.

[0011] Beneficial effect: The present invention considers the excess flood allocation principle at the coupled component level of factors such as the spatial location of the reservoir and the idle storage capacity, transforms the multi-objective problem into a single-objective problem with excess flood allocation constraints, and constructs a reservoir-lake system flood control and drainage joint scheduling model based on the coupled component level excess flood allocation. The improved whale-particle swarm hybrid optimization algorithm is used to quickly and accurately calculate the reservoir-lake system flood control and drainage joint scheduling results, which can quickly provide effective and reliable reference information for the reservoir-lake system flood control and drainage decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1is a flow chart of the method of the present invention.

[0013] Figure 2 It is a schematic diagram of the topological structure of a typical flood control and drainage system.

[0014] Figure 3 It is a schematic diagram of excess flood volume of the flood control and drainage system.

[0015] Figure 4 This is a comparison chart between the reservoir water level and flow process after the method is used for scheduling and the current scheduling rules. DETAILED DESCRIPTION

[0016] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0017] like Figures 1 to 4 As shown, the following technical solution is proposed.

[0018] According to one aspect of the present application, a reservoir-lake system scheduling method for coupled component-level excess flood volume allocation is provided, comprising the following steps:

[0019] Step S1, collecting hydrological data and engineering data of the reservoir-lake system, and calculating the reservoir inflow at each time step, the response process of the reservoir inflow at the control section, the response process of the measured outflow at the control section, the water inflow process between the reservoirs, the water inflow process between the lakes, and the natural water inflow process of the system based on the above data;

[0020] Step S2: Based on the hydrological data and engineering data, the LSTM model of coupled physical principle is used for fitting, and the fitting results are arranged in time series to obtain the initial drainage capacity process of the lake. Based on the natural water inflow process of the system and the initial drainage capacity process of the lake, the initial excess flood volume of the system is calculated; wherein the initial excess flood volume of the system is defined as the amount of water that exceeds the initial drainage capacity process of the lake by the natural water inflow process of the system;

[0021] Step S3: Based on the hydrological data, engineering data and reservoir inflow, the water inflow of each reservoir's excess flood storage section, the remaining flood control storage capacity of each reservoir, the total water inflow of the lake's excess flood storage section, and the flood storage capacity corresponding to the lake warning water level are calculated. Based on the above data, the flood storage capacity of each reservoir and the flood storage capacity of the lake are calculated in turn. Based on the natural water inflow process of the system, the improved XGBoost classification model based on the cost function is used to classify the flood. When 0≤excess flood volume <max{∑ Wd i,拦 , W 湖,拦}, the flood is classified as the first magnitude type, when max{∑ Wd i,拦 , W 湖,拦}≤Excess flood volume<∑(Wd i,拦 + W 湖,拦 ) the flood is classified as the second magnitude type, when ∑ ( Wd i,拦 + W 湖,拦 )≤excess flood volume, the flood is divided into the third magnitude type, and the excess flood volume generated by floods of different magnitude types is allocated based on the initial excess flood volume of the system and the flood classification results, and the excess flood volume allocation results are calculated and verified to obtain the excess flood volume; wherein the excess flood volume is composed of the excess flood volume allocation results, and the natural water inflow process of the system includes at least one flood;

[0022] Step S4, based on the excess flood volume, obtain the water level constraint of the pre-constructed reservoir-lake system flood control and drainage joint scheduling model based on coupled component-level excess flood volume allocation, read the reservoir inflow, reservoir interval water inflow process, lake interval water inflow process input model, and use the improved whale-particle swarm hybrid optimization algorithm to solve and obtain the reservoir-lake system flood control and drainage joint scheduling result.

[0023] In this embodiment, by constructing a coupled component-level excess flood distribution method and adopting an improved whale-particle swarm hybrid optimization algorithm, the calculation scale of multi-objective scheduling of flood control and drainage can be effectively reduced and the optimization efficiency can be improved.

[0024] According to one aspect of the present application, step S1 further comprises:

[0025] Step S11, collecting hydrological data and engineering data of the reservoir-lake system in the study area, the hydrological data including: river channel calculation parameters, reservoir storage, measured reservoir outflow, measured flow process of control section, lake storage, measured lake outflow and external river water level, the engineering data including: water level-storage capacity curve, water level-discharge capacity curve and engineering characteristic parameters of reservoirs and lakes;

[0026] In this embodiment, the above information can be obtained from public data, with the aim of gaining an in-depth understanding of the basic conditions and hydrological characteristics of the study area and providing basic information and data support for subsequent model design and construction.

[0027] Step S12: Based on the reservoir storage and the measured outflow of the reservoir, the reservoir inflow of each time step is calculated using the water balance principle, and the jagged edges of the reservoir inflow are eliminated using the sliding average method;

[0028] Step S13, using Muskingum method to calculate the measured outflow of the reservoir to the control section to obtain the response process of the measured outflow of the reservoir in the control section, and subtracting the measured flow process of the control section from the response process of the measured outflow of the reservoir in the control section to obtain the reservoir interval water inflow process of each time step;

[0029] Step S14: The Muskingum method is used to calculate the reservoir inflow to the control section to obtain the response process of the reservoir inflow at the control section. Based on the response process of the reservoir inflow at the control section, the reservoir interval water inflow process, the lake storage, and the measured lake outflow, the water balance method is used to deduce the lake interval water inflow process for each time step.

[0030] QLh t = ( Vl t+1 -Vl t ) / Δt+q n+1,t - QUh t t ∈[1 , T ];

[0031] Where: QLh t For the lake t The interval inflow process at the moment; Vl t+1 and Vl t The lakes are t +1 and t The accumulation of time; q n+1,t It is the outflow process of the lake; QUh t It is the upstream water process of the lake; T The scheduling period is long;

[0032] Step S15, calculate the system natural water inflow process for each time step, wherein the system natural water inflow process for each time step is the sum of the response process of the reservoir inflow at the control section, the reservoir interval water inflow process and the lake interval water inflow process for each time step.

[0033] In this embodiment, based on the hydrological data and engineering data and the reservoir inflow at each time step calculated based on the above data, the response process of the reservoir inflow at the control section, the response process of the measured reservoir outflow at the control section, the water inflow process between reservoirs, and the water inflow process between lakes, the natural water inflow process of the system is calculated to provide data support for subsequent calculations and modeling.

[0034] According to one aspect of the present application, step S12 further comprises:

[0035] Step S12a: Based on the reservoir storage and the measured outflow of the reservoir, the reservoir inflow at each time step is calculated using the water balance principle:

[0036] Q i,t = ( V i,t+1 -V i,t ) / Δt+q i,t i ∈[1 ,n ] t ∈[1 , T ];

[0037] Where: Q i,t For the reservoir t The flow of time; V i,t+1 and V i,t Reservoirs in t +1 and t The accumulation of time; Δt For a long period of time; q i,t For the reservoir t The outflow process at the moment; n is the number of reservoirs;

[0038] Step S12b, analyzing the runoff characteristics of the watershed, and determining the maximum value of the flow variation range of the reservoir inflow in adjacent time periods;

[0039] Step S12c: determine whether the calculated difference between the reservoir inflow at the current moment and the reservoir inflow at the next moment exceeds the constraint of the maximum value of the flow change range. If so, use the sliding average method to calculate the maximum value of the flow change range. t arrive t +1 The average value of the moment is replaced t The instantaneous value at the moment is eliminated as much as possible, otherwise it is equal to the original value, and the reservoir inflow is obtained after eliminating the aliasing using the sliding average method.

[0040] In this embodiment, the inferred reservoir inflow process often shows jagged jumps due to the influence of various factors. Therefore, it is necessary to use a sliding average method to eliminate the jaggedness and solve the problem of large fluctuations in the inferred reservoir inflow process, so that the inferred reservoir inflow process is more in line with general rules.

[0041] According to one aspect of the present application, step S13 is further:

[0042] Step S13a, using Muskingum method to calculate the measured outflow of the reservoir to the control section:

[0043] q j+1,t+1 = C 0q j,t+1 + C 1 q j,t + C 2 q j+1,t ;

[0044] Where:

[0045] C 0=(0.5 Δt-Kx ) / ( K-Kx+ 0.5 Δt );

[0046] C 1=(0.5 Δt+Kx ) / ( K-Kx+ 0.5 Δt );

[0047] C 2=( K-Kx- 0.5 Δt ) / ( K-Kx+ 0.5 Δt ) ;

[0048] and:

[0049] C 0+ C 1+ C 2=1.0;

[0050] Where: q j+1,t+1 For the river j+ Section 1 t +1 moment flow; q j,t+1 For the river j Section t +1 moment flow; q j,t For the river j Section t Time flow; q j+1,t For the river j+ Section 1 t Time flow; C 0. C 1. C 2 is the river routing parameter; ΔtThe calculation period is long;

[0051] Step S13b, subtract the measured flow process of the control section from the response process of the measured outflow of the reservoir at the control section to obtain the reservoir interval water inflow process of each time step:

[0052] QL i,t = QF i,t - QOm i,t i ∈[1 ,n ] t ∈[1 , T ];

[0053] Where: QL i,t It is the water inflow process between reservoirs; QL i,t To control the measured flow process of the section; QOm i,t It is the response process of the measured outflow of the reservoir at the control section.

[0054] In this embodiment, the Muskingum method is used to calculate the measured outflow of the reservoir to the control section, and the evolution law of the measured outflow of the reservoir is taken into account, so that the calculated water inflow process of the reservoir interval is more in line with the actual situation.

[0055] According to one aspect of the present application, step S2 further comprises:

[0056] Step S21, reading the lake storage, external river water level, lake measured outflow and lake water level-discharge capacity curve data from the hydrological data and engineering data;

[0057] Step S22, arranging the lake storage, external river water level and lake measured outflow data into the form of LSTM model input data as samples of the LSTM model;

[0058] Step S23: Based on the samples of the LSTM model and the water level-discharge capacity curve of the lake, the LSTM model based on the coupled physical principle is used to mine the mapping relationship between the lake storage, the water level of the external river and the measured outflow of the lake, and the initial drainage capacity of the lake at each time step is obtained by fitting, and the initial drainage capacity of the lake is arranged in a continuous time series based on the initial drainage capacity of the lake to obtain the initial drainage capacity process of the lake;

[0059] Step S24, calculating the system initial excess flood volume based on the system natural water inflow process and the lake initial flood drainage capacity process; wherein the system initial excess flood volume is defined as the water volume of the system natural water inflow process exceeding the lake initial flood drainage capacity process:

[0060] W超额 =∑[ Qr t - q n+1,t ] Δt t ∈[ t 0 , t 1];

[0061] q n+1,t =f 1( Vl t ,Zr t ) t ∈[ t 0 , t 1];

[0062] Where: Qr t It is the natural water inflow process of the system; q n+1,t It is the process of initial drainage capacity of the lake; q n+1,t Lake storage Vl t and outer river water level Zr t related; Δt For a long period of time; t 0 is Qr t >q n+1,t The starting timing of t 1 for Qr t >q n+1,t The termination sequence of the system initial excess flood is Figure 3 The total area of ​​the natural water inflow process in the middle system above the initial drainage capacity of the lake, as shown in the figure WR The flood control capacity of the reservoir group, WL To prevent flooding in lakes, W Res is the remaining excess flood volume;

[0063] In this embodiment, the initial excess flood volume of the system is calculated based on the natural water inflow process of the system and the initial drainage capacity process of the lake, providing an initial value for subsequent iterative calculation of the excess flood volume.

[0064] According to one aspect of the present application, step S23 is further:

[0065] Step S23a, based on the sample of the LSTM model and the water level-discharge capacity curve of the lake, build an LSTM model and construct a loss function, wherein the LSTM model includes a forget gate, an input gate, a long-term state unit of the current input, a long-term state unit at the current moment, and an output gate;

[0066] Step S23b, adding a discharge capacity constraint penalty term and a non-negative constraint penalty term to the loss function, and establishing an LSTM model of coupled physical principles;

[0067] Step S23c, dividing the samples of the LSTM model into a training period and a verification period, using the lake storage capacity and the water level of the external river as the independent variable factors of the LSTM model of the coupled physical principle, and the measured outflow of the lake as the dependent variable factor of the LSTM model of the coupled physical principle, inputting the independent variable factors and the dependent variable factors of the training period into the LSTM model of the coupled physical principle for training, and obtaining the mapping relationship between the independent variable factors and the dependent variable factors;

[0068] Step S23d, input the independent variable factors of the verification period into the model to calculate the prediction results of the corresponding dependent variable factors, that is, obtain the initial drainage capacity of the lake at each time step, arrange the initial drainage capacity of the lake in time series, and obtain the initial drainage capacity process of the lake.

[0069] In this embodiment, an LSTM model based on coupled physical principles is used to quickly calculate the initial drainage capacity of the lake that is more in line with actual laws. The fitting results are arranged in time series to obtain a more accurate process of the initial drainage capacity of the lake, thereby reducing the computational time cost of subsequent iterative calculations of excess flood volume.

[0070] According to one aspect of the present application, step S23b is further:

[0071] Step S23b1, reading the lake storage and the lake water level-storage capacity curve data in the sample of the LSTM model, obtaining the lake water level based on the above data, and obtaining the lake discharge capacity based on the lake water level and the lake water level-discharge capacity curve;

[0072] Step S23b2: Setting parameters y 1. Parameters y 2, where the parameters y 1 is the parameter to be assigned to the value of the initial drainage capacity of the lake predicted in step S23c, and the parameter y 2 is the parameter to be assigned to the discharge capacity of the lake;

[0073] Step S23b3: Compare parameters y 1 and parameters y 2, if the parameter y 1 is greater than the parameter y2, then calculate the root mean square error of the two and obtain the discharge capacity constraint penalty term. If the parameter y 1 is less than the parameter y 2, then the discharge capacity constraint penalty term is 0;

[0074] Step S23b4: Compare parameters y The size of 1 and 0, the absolute value of the smaller value between the two and the larger value O Multiply them together to get the non-negative constraint penalty term;

[0075] Step S23b5, extract the LSTM model loss function, add the discharge capacity constraint penalty term and the non-negative constraint penalty term to the LSTM model loss function to form a new loss function, and establish an LSTM model of coupled physical principles.

[0076] Compared with the artificially assumed mapping relationship and the basic LSTM method of traditional water conservancy models, this application adopts the LSTM model of coupled physical principles to explore the mapping relationship between lake storage, external river water level and actual lake outflow, and adds discharge capacity constraints and non-negative constraint penalty terms to the model loss function. Through its special gating mechanism and data-driven method, it effectively captures the time series correlation of lake storage, external river water level and actual lake outflow, effectively solves the problem of LSTM model fitting values ​​exceeding the discharge capacity constraint and appearing negative values, reduces the deviation of artificial assumptions, and reduces the computational time cost of subsequent iterative calculations of excess flood volume.

[0077] According to one aspect of the present application, step S3 further comprises:

[0078] Step S31, calculate the water inflow of each reservoir's excess flood storage section based on the reservoir inflow, wherein the period when the system's natural water inflow process is greater than the lake's initial drainage capacity process is the reservoir's excess flood storage section:

[0079] W i =∑ Qm i,t Δt t ∈[ t 0 , t 1];

[0080] Where: W i For the upstream i The amount of water inflow into the excess flood storage section of the reservoir;

[0081] Step S32, calculating the remaining flood control storage capacity of each reservoir based on the reservoir storage capacity, the water level-storage capacity curve of the reservoir and the engineering characteristic parameters;

[0082] Step S33, calculate the flood storage capacity of each reservoir, wherein the flood storage capacity of each reservoir is the smaller of the water inflow of the excess flood storage section of each reservoir and the remaining flood control storage capacity of each reservoir:

[0083] Wd i,拦 = min{ W i , Vf i} i ∈[1 ,n ];

[0084] Where: Vf i For the upstream i Remaining flood control capacity of the reservoir; Wd i,拦 For the i The flood storage capacity of the reservoir;

[0085] Step S34, calculating the flood storage capacity of the reservoir group, wherein the flood storage capacity of the reservoir group is the sum of the flood storage capacities of each reservoir;

[0086] Step S35, based on the reservoir inflow, calculate the water inflow of each reservoir's excess flood storage section and the flood storage capacity of each reservoir, obtain the outflow of the upstream reservoir, and calculate the total inflow of the lake's excess flood storage section based on the outflow of the upstream reservoir and the water inflow process between the lakes, where the period when the system's natural water inflow process is greater than the lake's initial drainage capacity process is the lake's excess flood storage section;

[0087] W lake =∑( W i - Wd i,拦 )+∑ QLh t Δt i ∈[1 ,n ] t ∈[ t 0 , t 1];

[0088] Where: W lake The total water inflow of the lake's excess flood storage section;

[0089] Step S36: Calculate the flood storage capacity corresponding to the lake warning water level based on the lake water level-storage capacity curve and engineering characteristic parameters W P ;

[0090] Step S37, calculate the flood storage capacity of the lake, where the flood storage capacity of the lake is the smaller of the total water inflow of the flood storage section of the lake for excess flood volume and the flood storage capacity corresponding to the lake warning water level:

[0091] W 湖,拦 = min{ W lake , W P};

[0092] Where: W P The warning water level of the lake corresponds to the flood storage capacity; W 湖,拦 for the flood storage capacity of the lake;

[0093] Step S38: Based on the natural water flow process of the system, the improved XGBoost classification model based on the cost function is used to classify the flood. When 0≤excess flood volume <max{∑ Wd i,拦 , W 湖,拦}, the flood is classified as the first magnitude type, when max{∑ Wd i,拦 , W 湖,拦}≤Excess flood volume<∑( Wd i,拦 + W 湖,拦 ) the flood is classified as the second magnitude type, when ∑ ( Wd i,拦 + W 湖,拦 ) ≤ excess flood volume, the flood is classified into the third magnitude type. Based on the system's initial excess flood volume and the flood classification results, the excess flood volume generated by floods of different magnitude types is allocated, the excess flood volume allocation results are calculated and verified, and the excess flood volume is obtained; wherein, the excess flood volume is composed of the excess flood volume allocation results, and the system's natural water inflow process includes at least one flood.

[0094] In this embodiment, excess flood volume is obtained based on hydrological data, engineering data and reservoir inflow, and water level constraints are provided for a subsequent reservoir-lake system flood control and drainage joint scheduling model based on coupled component-level excess flood volume allocation.

[0095] According to one aspect of the present application, step S38 is further:

[0096] Step S38a, normalizing the system's natural water inflow process as a sample of the improved XGBoost classification model based on the cost function; wherein the system's natural water inflow process includes at least one flood;

[0097] Step S38b, constructing an objective function, setting the number of model iterations and iteration convergence conditions, wherein the objective function includes a loss function and a regularization term;

[0098] Step S38c, using the gradient boosting algorithm to iteratively construct a decision tree, and using the cost function to improve the XGBoost classification method, and constructing a classification model by accumulating the outputs of multiple decision trees;

[0099] Step S38d, dividing the samples into a training period and a validation period, inputting the samples in the training period into the improved XGBoost classification model based on the cost function for training to obtain the classification principle;

[0100] Step S38e, input the samples of the validation period into the improved XGBoost classification model based on the cost function, and obtain the flood classification results of the first magnitude type, the second magnitude type, and the third magnitude type;

[0101] Step S38f, based on the system initial excess flood volume and flood classification results, use the coupled component-level excess flood volume allocation method to allocate excess flood volume to floods of different magnitudes and types, and calculate the excess flood volume allocation results;

[0102] Step S38g, verify the excess flood volume allocation result. When the difference between the lake water level after the iteration and the lake water level at the beginning of this iteration is less than the convergence accuracy index, the iteration ends and the excess flood volume is obtained. Otherwise, the lake water level is updated and the step S38f is turned to perform iterative calculation. The excess flood volume is composed of the excess flood volume allocation result.

[0103] Vl t,k+1 = Vl t,k +( q n+1,t,k - q n+1,t,k+1 ) Δt t ∈[1 , T ];

[0104] Zl t,k+1 = f 2 ( Vl t,k+1 ) t ∈[1 , T ];

[0105] Where: Vl t,k , Vl t,k+1 Before and after iterative update t Lake storage capacity at any given moment; f 2 It is the water level-storage capacity curve data of the lake;

[0106] In this embodiment, based on the natural water inflow process of the system, an improved XGBoost classification model based on the cost function is adopted to obtain the flood classification results, and based on the initial excess flood volume of the system and the flood classification results, a coupled component-level excess flood volume allocation method is proposed for floods of different magnitudes and types to obtain the excess flood volume. The excess flood volume provides water level constraints for the subsequent constructed reservoir-lake system flood control and drainage joint scheduling model based on the coupled component-level excess flood volume allocation, thereby improving the computational efficiency of the model.

[0107] According to one aspect of the present application, step S38c is further:

[0108] Step S38c1, based on the sample, objective function, number of model iterations, and iterative convergence conditions, the average value of the objective function is set as the initial prediction value, and a decision tree is constructed based on the above data using a gradient boosting algorithm;

[0109] Step S38c2: Setting parameters y 3. Parameters y 4, where the parameters y 3 is the parameter to be assigned to the actual magnitude type of the sample. y 4 is the parameter to be assigned to the magnitude type obtained by sample training classification in step S38d;

[0110] Step S38c3: Based on parameters y 3 and parameters y 4. Calculate the first-order derivative and second-order derivative of the loss function for each sample. The first-order derivative guides the growth of the new decision tree.

[0111] Step S38c4, based on the first-order derivative and the second-order derivative of the loss function, draw a ROC curve, each point under the ROC curve corresponds to a different classification threshold, and the true positive rate (TPR) and false positive rate (FPR) of the model are calculated based on the classification threshold;

[0112] Step S38c5, setting the weights of missed judgments and misjudgments, and based on the weights of missed judgments and misjudgments and the number of missed judgments and misjudged samples, using the cost function to optimize the classification threshold, the optimized classification threshold is the split point of the decision tree, wherein judging the third magnitude type as the second magnitude type and the second magnitude type as the first magnitude type is a missed judgment, judging the first magnitude type as the second magnitude type and the second magnitude type as the third magnitude type is a misjudgment, and the cost function is calculated according to the following formula:

[0113] COST = d 1. Number of missed samples + d 2. Number of misjudged samples;

[0114] Where: d 1.d 2 is the weight of missed judgment and misjudgment. Since the cost of missed judgment is greater, d 1> d 2. Select the point with the minimum total cost by adjusting the threshold;

[0115] Assume that the classifier outputs the category c The probability of P ( y = c | x ), define different thresholds τ for each category c , the threshold is the splitting point of the decision tree:

[0116] y=argmax[ P ( y = c | x )>τ c ];

[0117] Step S38c6: obtain a prediction value based on the split point, add the prediction value of the new decision tree to the previous prediction value, and update the overall prediction value;

[0118] Step S38c7, judging whether the number of iterations reaches the set number of model iterations or whether the predicted value satisfies the iterative convergence condition; if the number of iterations does not reach the set number of model iterations and the predicted value does not satisfy the iterative convergence condition, then go to step S38c3 to perform iterative calculation, otherwise the iteration is terminated;

[0119] Step S38c8: weighted sum of the outputs of all decision trees to form a classification model, where the output of the decision tree is the predicted value.

[0120] In this embodiment, an improved XGBoost classification model based on a cost function is used to divide floods into three different magnitude types. XGBoost is a widely used gradient boosting decision tree algorithm that effectively handles complex classification problems by integrating the prediction results of multiple decision trees in classification tasks. By default, the decision threshold of XGBoost classification is 0.5. Classifying floods of different magnitude types with traditional decision thresholds does not meet the needs of actual conditions. This application combines the cost function and reduces the misjudgment rate and missed judgment rate of the classification model through the cost function method, thereby optimizing the performance of the classification model to meet the needs of accurate flood classification.

[0121] According to one aspect of the present application, the step S38f is further:

[0122] Step S38f1, judging the flood magnitude type based on the system initial excess flood volume and flood classification results;

[0123] Step S38f2, if 0≤W 超额，k <max{∑ Wd i,拦 , W 湖,拦}(in k is the number of iterations), perform this step:

[0124] Considering the principle of equal sharing of flood control safety, based on relevant simulation data, the excess flood volume is distributed according to the proportion of reservoir flood storage capacity, where the reservoir flood storage capacity takes into account the inflow of excess flood storage sections of each reservoir and the remaining flood control storage capacity of each reservoir. At this time, the flood interception capacity of the reservoir group is:

[0125] WR =∑ Wd i,拦 i ∈[1 ,n ];

[0126] The flood control capacity of the lake is:

[0127] WL =min{ W 超额，k -∑ Wd i,拦 , W 湖,拦} i ∈[1 ,n ];

[0128] Obtain the excess flood allocation result, where the excess flood allocation result is composed of the flood control capacity of the reservoir group and the flood control capacity of the lake, otherwise skip this step and proceed to the next step;

[0129] Step S38f3, if max{∑ Wd i,拦 , W 湖,拦}≤ W 超额，k <∑( Wd i,拦 + W 湖,拦 )(in k is the number of iterations), perform this step:

[0130] The flood interception capacity of the lake is obtained based on the flood storage capacity of the lake, and the remaining excess flood volume is calculated based on the flood interception capacity of the lake and the excess flood volume:

[0131] WL= W 湖,拦

[0132] W Res = W 超额，k - WL

[0133] Based on relevant simulation data, the remaining excess flood volume is distributed according to the proportion of reservoir flood storage capacity, where the reservoir flood storage capacity takes into account the inflow of each reservoir's excess flood storage section and the remaining flood control storage capacity of each reservoir. At this time, the flood control capacity of the reservoir group is:

[0134] WR =min{ W 剩余 ,∑ Wd i,拦} i ∈[1 ,n ];

[0135] The flood control capacity of each reservoir is calculated based on the flood control capacity of the reservoir group and the flood storage capacity of the reservoir:

[0136] α i = Wd i,拦 / ∑ Wd i,拦 i ∈[1 ,n ];

[0137] WR i = α i · WR i ∈[1 ,n ];

[0138] Obtain the excess flood allocation result, where the excess flood allocation result is composed of the flood control capacity of the reservoir group and the flood control capacity of the lake, otherwise skip this step and proceed to the next step;

[0139] Step S38f4, if ∑( Wd i,拦 + W 湖,拦 ) ≤ W 超额，k (in k is the number of iterations), perform this step:

[0140] The flood interception capacity of the reservoir is obtained based on the relevant simulation data and the flood storage capacity of the reservoir. The flood storage capacity of the reservoir takes into account the water inflow of the excess flood storage section of each reservoir and the remaining flood control storage capacity of each reservoir. The flood interception capacity of each reservoir is summed up to get the flood interception capacity of the reservoir group:

[0141] WR i = Wd i,拦 i ∈[1 ,n ];

[0142] WR= ∑ WR i i ∈[1 ,n ];

[0143] Based on the flood control capacity and excess flood capacity of the reservoir group, the flood control capacity of the lake is calculated:

[0144] WL= W 超额，k – WR

[0145] The excess flood volume allocation result is obtained, where the excess flood volume allocation result is composed of the flood control capacity of the reservoir group and the flood control capacity of the lake.

[0146] In this embodiment, based on the system's initial excess flood volume and flood classification results, a coupled component-level excess flood volume allocation method is used to allocate the excess flood volumes of floods of different magnitudes and types, and the excess flood volume allocation results are obtained to provide data support for the calculation of excess flood volume.

[0147] According to one aspect of the present application, step S4 is further:

[0148] Step S41, based on the excess flood volume, obtain the water level constraint of the reservoir-lake system flood control and drainage joint scheduling model based on the coupled component-level excess flood volume allocation, read the reservoir inflow, reservoir interval water inflow process, and lake interval water inflow process input model, wherein the objective function of the model is to minimize the excess flood volume, and the constraints include: water balance constraint, water level constraint, scheduling period initial condition constraint, scheduling period boundary condition constraint, outflow discharge constraint, discharge amplitude constraint, and non-negative variable constraint;

[0149] Step S42, setting the parameters of the whale optimization algorithm and the particle swarm optimization algorithm, constructing the whale optimization algorithm model and the particle swarm optimization algorithm model, and using the improved whale-particle swarm hybrid optimization algorithm to solve and obtain the joint optimization scheduling result of flood control and drainage of the reservoir-lake system.

[0150] In this embodiment, a reservoir-lake system flood control and drainage joint scheduling model based on coupled component-level excess flood volume allocation is constructed, and the model is solved by using an improved whale-particle swarm hybrid optimization algorithm to quickly and accurately calculate the reservoir-lake system flood control and drainage joint scheduling results.

[0151] According to one aspect of the present application, step S42 is further:

[0152] Step S42a: Setting the convergence factor of the whale optimization algorithm model a max , control the decay speed uThe initial value of the factor; Set the maximum value of the inertia weight of the particle swarm optimization algorithm w max , minimum w min and the maximum number of iterations T ;

[0153] Step S42b, constructing a whale optimization algorithm model based on the convergence factor a max , control the decay speed u The initial value of the factor and the current number of iterations t , using the exponential function to calculate the new nonlinear dynamic convergence factor:

[0154] a=a max · e -ut ;

[0155] In this embodiment, the convergence factor of the whale optimization algorithm is improved by exponential decrease in order to speed up the convergence speed in the later stage, maintain a large exploration ability in the early stage of the algorithm to avoid falling into the local optimum, and quickly reduce it in the later stage to enhance the development ability and improve the convergence accuracy; and the rapid decay characteristics of the exponential function enable the algorithm to enter the local search stage more quickly, shorten the optimization time, and effectively avoid oscillation in the later stage to enhance stability.

[0156] Step S42c: construct a particle swarm optimization algorithm model based on the maximum value of the inertia weight w max , minimum w min , current iteration number t and the maximum number of iterations T , the cosine function is used to calculate the new nonlinear dynamic inertia weight factor:

[0157] w ( t )= w min +( w max - w min ) / 2·(1+cos(π· t / T ));

[0158] In this embodiment, the inertia weight factor of the particle swarm optimization algorithm is improved by nonlinear dynamic adjustment of the cosine function, so that the inertia weight maintains a high value in the early stage of the search to promote global search; it gradually decreases in the later stage to enhance local development capabilities, and the smooth change of the cosine function is more natural than linear attenuation, avoiding oscillation or premature convergence problems caused by too rapid changes in weight during the search process.

[0159] Step S42d, set the threshold, divide the reservoir inflow into high-flow population A and low-flow population B, calculate the new nonlinear dynamic convergence factor based on the exponential function, use the whale optimization algorithm to solve population A, and calculate the new nonlinear dynamic inertia weight factor based on the cosine function, use the particle swarm optimization algorithm to solve population B, and obtain the joint optimization scheduling result of flood control and drainage of the reservoir-lake system.

[0160] According to one aspect of the present application, the step S42d is further:

[0161] Step S42d1, given a threshold ζ, the time period when the reservoir inflow is higher than the threshold ζ is divided into a high-flow population A, and the time period when the reservoir inflow is lower than the threshold ζ is divided into a low-flow population B;

[0162] Step S42d2: Based on the new nonlinear dynamic convergence factor calculated by the exponential function, the whale optimization algorithm is used to optimize the high-flow population A: Initialize the search space dimension D , population size M , maximum number of iterations T , initialize the whale population position X and fitness value, record the optimal position in the population X * , a new nonlinear dynamic convergence factor calculated using an exponential function is used to calculate each whale i of A and C :

[0163] A= 2 ar – a ;

[0164] C= 2 r ;

[0165] Where: r is a random number in [0,1];

[0166] If | A |<1, the update position is close to the current optimal solution:

[0167] X i = X * - A|CX * - X i | ;

[0168] If | A |≥1, randomly select a position in the population Xrand :

[0169] X i = X rand - A|CX rand - X i | ;

[0170] Choose the spiral update with a probability of 50%:

[0171] X i = De bl cos(2π l )+ X * ;

[0172] Where: D=| X * - X i | is the spiral path distance; b is a constant that controls the shape of the spiral; l is a random number between [-1,1];

[0173] If the fitness value of a whale's position is better, update X * When updating, it is important to ensure that the position of each whale is within the allowed range. If the maximum number of iterations is reached or the accuracy requirement is met, the optimization is stopped and the optimal position is output. X * and the optimal fitness value, which is the optimal solution of the scheduling result;

[0174] Based on the cosine function, a new nonlinear dynamic inertia weight factor is calculated, and the particle swarm optimization algorithm is used to optimize the low-flow population B: N , upper and lower limits of each particle's speed, acceleration factor c 1 and c 2. Random Numbers r 1 and r 2. Initialize the parameters of each particle's position, speed adaptation value, and individual optimal position. p best,i and the global optimal position g best Initialize and update particles i Speed:

[0175] v i = w (t ) v i + c 1. r 1·( p best,i - x i ) + c 2. r 2.( g best - x i );

[0176] Where: v i is the particle velocity; x i is the particle position;

[0177] Update Particles i Location:

[0178] x i = x i +v i ;

[0179] If the fitness value of the particle's new position is better than p best,i , then update p best,i If the particle p best,i Better than g best , then update g best During the update, ensure that the particle position and velocity are within the allowable range. If the maximum number of iterations is reached or the accuracy requirement is met, stop the optimization and output the global optimal position. g best and the optimal fitness value, which is the optimal solution of the scheduling result;

[0180] Step S42d3, arrange the optimal solution of the scheduling result in time series to obtain the joint optimization scheduling result of flood control and drainage of the reservoir-lake system.

[0181] In this embodiment, the task is split into subtasks through a clustering strategy. The high-flow process has complex nonlinear characteristics, and the objective function has a large range of variation. The whale optimization algorithm has a strong global search capability, especially in high-dimensional nonlinear problems, and can better handle complex flood scheduling needs; the characteristics of the low-flow process are relatively stable, and the particle swarm optimization algorithm has a high computational efficiency when dealing with continuous optimization problems. It can quickly find feasible solutions and reduce the waste of computing resources. Selecting appropriate optimization algorithms for different problems can effectively avoid the defects of a single algorithm and improve the overall optimization performance.

[0182] In the study area, based on the classification results of the improved XGBoost classification model based on the cost function, three historical floods of different magnitudes, namely 20190606, 20210710 and 20200707, were selected for scheduling research. For reservoirs, the highest water levels of the three floods of different magnitudes were lower than those of regular scheduling after the scheduling of this method, which were reduced by 0.39 m, 0.11 m and 0.03 m respectively. For lakes, the river discharge increased after the scheduling of the 20190606 flood using this method compared with the regular scheduling. After the scheduling of the 20210710 flood, the over-warning water level operation using the current rules was changed to the lake water level not exceeding the warning level during the entire scheduling period. After the scheduling of the 20200707 flood, the highest water level of the lake was reduced by 0.01m.

[0183] In summary, by applying the coupled component-level excess flood allocation method for the reservoir-lake system scheduling proposed in this application, the mapping relationship between the lake storage, the external river water level and the measured outflow of the lake can be quickly and accurately excavated to fit the initial drainage capacity of the lake, and floods can be classified scientifically and reasonably. The coupled component-level excess flood allocation method is used to allocate excess floods to floods of different magnitudes and types. The improved whale-particle swarm hybrid optimization algorithm is used to quickly and accurately calculate the joint scheduling results of flood control and drainage of the reservoir-lake system, providing effective and reliable reference information for flood control and drainage decision-making of the reservoir-lake system.

[0184] According to one aspect of the present application, the specific output of the present invention is the joint optimization scheduling result of flood control and drainage in the reservoir-lake system.

[0185] In this embodiment, after all the above steps, the combined optimization scheduling result of the reservoir-lake system flood control and drainage can be obtained, which provides effective and reliable reference information for the reservoir-lake system flood control and drainage decision-making, and has high technical innovation and practicality. The preferred implementation of the present invention is described in detail above, but the present invention is not limited to the specific details of the above implementation. Within the technical concept of the present invention, the technical solution of the present invention can be subjected to a variety of equivalent transformations, and these equivalent transformations all belong to the protection scope of the present invention.

Claims

1. A reservoir-lake system dispatching method for coupled component-level excess flood distribution, characterized in that: The following steps are involved: Step S1, collecting hydrological data and engineering data of the reservoir-lake system, and calculating the reservoir inflow at each time step, the response process of the reservoir inflow at the control section, the response process of the measured outflow at the control section, the reservoir interval water inflow process, the lake interval water inflow process, and the system natural water inflow process based on the above data; Step S2: Based on the hydrological data and engineering data, the LSTM model of coupled physical principle is used for fitting, and the fitting results are arranged in time series to obtain the initial drainage capacity process of the lake. Based on the natural water inflow process of the system and the initial drainage capacity process of the lake, the initial excess flood volume of the system is calculated; wherein the initial excess flood volume of the system is defined as the amount of water that exceeds the initial drainage capacity process of the lake by the natural water inflow process of the system; Step S3, based on hydrological data and engineering data, reservoir inflow, calculate the inflow of excess flood storage sections of each reservoir, the remaining flood control storage capacity of each reservoir, the total inflow of excess flood storage sections of lakes, and the flood storage capacity corresponding to the warning water level of lakes, and calculate the flood storage capacity of each reservoir and the flood storage capacity of lakes in turn based on the above data, based on the natural water inflow process of the system, use the improved XGBoost classification model based on the cost function to divide the flood into the first magnitude type, the second magnitude type, and the third magnitude type, based on the initial excess flood volume of the system and the flood classification results, distribute the excess flood volume generated by floods of different magnitude types, calculate the excess flood volume distribution results and verify them, and obtain the excess flood volume; wherein, the excess flood volume is composed of the excess flood volume distribution results, and the natural water inflow process of the system includes at least one flood; Step S4, based on the excess flood volume, obtain the water level constraint of the reservoir-lake system flood control and drainage joint scheduling model based on the coupled component-level excess flood volume allocation, read the reservoir inflow, reservoir interval water inflow process, and lake interval water inflow process input model, and use the improved whale-particle swarm hybrid optimization algorithm to solve and obtain the reservoir-lake system flood control and drainage joint scheduling result; Step S4 includes the following steps: Step S41, based on the excess flood volume, obtain the water level constraint of the reservoir-lake system flood control and drainage joint scheduling model based on the coupled component-level excess flood volume allocation, read the reservoir inflow, reservoir interval water inflow process, and lake interval water inflow process input model, wherein the objective function of the model is to minimize the excess flood volume, and the constraints include: water balance constraint, water level constraint, scheduling period initial condition constraint, scheduling period boundary condition constraint, outflow discharge constraint, discharge amplitude constraint, and non-negative variable constraint; Step S42, setting parameters of the whale optimization algorithm and the particle swarm optimization algorithm, constructing the whale optimization algorithm model and the particle swarm optimization algorithm model, and using the improved whale-particle swarm hybrid optimization algorithm to solve and obtain the reservoir-lake system flood control and drainage joint optimization scheduling result; The step S42 specifically includes: Step S42a, setting the convergence factor of the whale optimization algorithm model and the initial value of the factor controlling the decay speed; setting the maximum value, minimum value and maximum number of iterations of the inertia weight of the particle swarm optimization algorithm; Step S42b, constructing a whale optimization algorithm model, and using an exponential function to calculate a new nonlinear dynamic convergence factor based on the convergence factor, the initial value of the factor controlling the decay speed, and the current number of iterations; Step S42c, constructing a particle swarm optimization algorithm model, and using a cosine function to calculate a new nonlinear dynamic inertia weight factor based on the maximum value, minimum value, and maximum number of iterations of the inertia weight; Step S42d, set the threshold, divide the reservoir inflow into high-flow population A and low-flow population B, calculate the new nonlinear dynamic convergence factor based on the exponential function, use the whale optimization algorithm to solve population A, and calculate the new nonlinear dynamic inertia weight factor based on the cosine function, use the particle swarm optimization algorithm to solve population B, and obtain the joint optimization scheduling result of flood control and drainage of the reservoir-lake system.

2. The reservoir-lake system dispatching method for coupled component-level excess flood distribution according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11, collecting hydrological data and engineering data of the reservoir-lake system in the study area, the hydrological data including: river channel calculation parameters, reservoir storage, measured reservoir outflow, measured flow process of control section, lake storage, measured lake outflow and external river water level, the engineering data including: water level-storage capacity curve, water level-discharge capacity curve and engineering characteristic parameters of reservoirs and lakes; Step S12: Based on the reservoir storage and the measured outflow of the reservoir, the reservoir inflow of each time step is calculated using the water balance principle, and the jagged edges of the reservoir inflow are eliminated using the sliding average method; Step S13, using Muskingum method to calculate the measured outflow of the reservoir to the control section to obtain the response process of the measured outflow of the reservoir in the control section, and subtracting the measured flow process of the control section from the response process of the measured outflow of the reservoir in the control section to obtain the reservoir interval water inflow process of each time step; Step S14, using Muskingum method to calculate the reservoir inflow to the control section to obtain the response process of the reservoir inflow in the control section, based on the response process of the reservoir inflow in the control section, the reservoir interval water inflow process, the lake storage, and the measured lake outflow, the water balance method is used to deduce the lake interval water inflow process for each time step; Step S15, calculate the system natural water inflow process for each time step, wherein the system natural water inflow process for each time step is the sum of the response process of the reservoir inflow at the control section, the reservoir interval water inflow process and the lake interval water inflow process for each time step.

3. The reservoir-lake system dispatching method for coupled component-level excess flood distribution according to claim 2 is characterized in that: Step S2 includes the following steps: Step S21, reading the lake storage, external river water level, lake measured outflow and lake water level-discharge capacity curve data from the hydrological data and engineering data; Step S22, arranging the lake storage, external river water level and lake measured outflow data into the form of LSTM model input data as samples of the LSTM model; Step S23: Based on the samples of the LSTM model and the water level-discharge capacity curve of the lake, the LSTM model based on the coupled physical principle is used to mine the mapping relationship between the lake storage, the water level of the external river and the measured outflow of the lake, and the initial drainage capacity of the lake at each time step is obtained by fitting, and the initial drainage capacity of the lake is arranged in a continuous time series based on the initial drainage capacity of the lake to obtain the initial drainage capacity process of the lake; Step S24, calculating the system's initial excess flood volume based on the system's natural water inflow process and the lake's initial flood drainage capacity process; wherein the system's initial excess flood volume is defined as the amount of water that exceeds the lake's initial flood drainage capacity process by the system's natural water inflow process.

4. The reservoir-lake system dispatching method for coupled component-level excess flood distribution according to claim 3, characterized in that: The step S23 specifically includes: Step S23a, based on the sample of the LSTM model and the water level-discharge capacity curve of the lake, build an LSTM model and construct a loss function, wherein the LSTM model includes a forget gate, an input gate, a long-term state unit of the current input, a long-term state unit at the current moment, and an output gate; Step S23b, adding a discharge capacity constraint penalty term and a non-negative constraint penalty term to the loss function, and establishing an LSTM model of coupled physical principles; Step S23c, dividing the samples of the LSTM model into a training period and a verification period, using the lake storage capacity and the water level of the external river as the independent variable factors of the LSTM model of the coupled physical principle, and the measured outflow of the lake as the dependent variable factor of the LSTM model of the coupled physical principle, inputting the independent variable factors and the dependent variable factors of the training period into the LSTM model of the coupled physical principle for training, and obtaining the mapping relationship between the independent variable factors and the dependent variable factors; Step S23d, input the independent variable factors of the verification period into the model to calculate the prediction results of the corresponding dependent variable factors, that is, obtain the initial drainage capacity of the lake at each time step, arrange the initial drainage capacity of the lake in time series, and obtain the initial drainage capacity process of the lake.

5. The reservoir-lake system dispatching method for coupled component-level excess flood distribution according to claim 4 is characterized in that: Step S3 includes the following steps: Step S31, calculating the water inflow of each reservoir's excess flood storage section based on the reservoir inflow, wherein the period when the system's natural water inflow process is greater than the lake's initial drainage capacity process is the reservoir's excess flood storage section; Step S32, calculating the remaining flood control storage capacity of each reservoir based on the reservoir storage capacity, the water level-storage capacity curve of the reservoir and the engineering characteristic parameters; Step S33, calculating the flood storage capacity of each reservoir, wherein the flood storage capacity of each reservoir is the smaller of the water inflow of the excess flood storage section of each reservoir and the remaining flood control storage capacity of each reservoir; Step S34, calculating the flood storage capacity of the reservoir group, wherein the flood storage capacity of the reservoir group is the sum of the flood storage capacities of each reservoir; Step S35, based on the reservoir inflow, calculate the water inflow of each reservoir's excess flood storage section and the flood storage capacity of each reservoir, obtain the water outflow of the upstream reservoir, and calculate the total water inflow of the lake's excess flood storage section based on the water outflow of the upstream reservoir and the water inflow process between the lakes, wherein the period when the system's natural water inflow process is greater than the lake's initial drainage capacity process is the lake's excess flood storage section; Step S36, calculating the flood storage capacity corresponding to the lake warning water level based on the lake water level-storage capacity curve and engineering characteristic parameters; Step S37, calculating the flood storage capacity of the lake, wherein the flood storage capacity of the lake is the smaller of the total water inflow of the flood storage section of the lake for excess flood volume and the flood storage capacity corresponding to the lake warning water level; Step S38: Based on the system's natural water inflow process, an improved XGBoost classification model based on a cost function is used to divide floods into first magnitude type, second magnitude type, and third magnitude type; based on the system's initial excess flood volume and the flood classification results, the excess flood volume generated by floods of different magnitude types is distributed; the excess flood volume distribution results are calculated and verified to obtain the excess flood volume; wherein, the excess flood volume is composed of the excess flood volume distribution results, and the system's natural water inflow process includes at least one flood.

6. The reservoir-lake system dispatching method for coupled component-level excess flood volume allocation according to claim 5, characterized in that: The step S38 specifically includes: Step S38a, normalizing the system's natural water inflow process as a sample of the improved XGBoost classification model based on the cost function; wherein the system's natural water inflow process includes at least one flood; Step S38b, constructing an objective function, setting the number of model iterations and iteration convergence conditions, wherein the objective function includes a loss function and a regularization term; Step S38c, using the gradient boosting algorithm to iteratively construct a decision tree, and using the cost function to improve the XGBoost classification method, and constructing a classification model by accumulating the outputs of multiple decision trees; Step S38d, dividing the samples into a training period and a validation period, inputting the samples in the training period into the improved XGBoost classification model based on the cost function for training to obtain the classification principle; Step S38e, input the samples of the validation period into the improved XGBoost classification model based on the cost function, and obtain the flood classification results of the first magnitude type, the second magnitude type, and the third magnitude type; Step S38f, based on the system initial excess flood volume and flood classification results, use the coupled component-level excess flood volume allocation method to allocate excess flood volume to floods of different magnitudes and types, and calculate the excess flood volume allocation results; Step S38g, verify the excess flood volume allocation result. When the difference between the lake water level after the iteration and the lake water level at the beginning of this iteration is less than the convergence accuracy index, the iteration ends and the excess flood volume is obtained. Otherwise, the lake water level is updated and the step S38f is performed for iterative calculation. Among them, the excess flood volume is composed of the excess flood volume allocation result.

7. The reservoir-lake system dispatching method for coupled component-level excess flood volume allocation according to claim 6, characterized in that: The step S38f specifically includes: Step S38f1, judging the flood magnitude type based on the system initial excess flood volume and flood classification results; Step S38f2, if the flood is of the first magnitude type, execute this step: calculate the flood interception capacity of the reservoir based on the flood storage capacity of the reservoir, sum the flood interception capacity of each reservoir to obtain the flood interception capacity of the reservoir group, calculate the flood interception capacity of the lake based on the flood interception capacity of the reservoir group, the excess flood volume and the flood storage capacity of the lake, and obtain the excess flood volume allocation result, wherein the excess flood volume allocation result is composed of the flood interception capacity of the reservoir group and the flood interception capacity of the lake; otherwise skip this step and proceed to the next step; Step S38f3, if the flood is of the second magnitude type, execute this step: obtain the lake flood interception capacity based on the lake flood storage capacity, calculate the reservoir group flood interception capacity based on the lake flood interception capacity, excess flood volume and reservoir flood storage capacity, calculate the flood interception capacity of each reservoir based on the reservoir group flood interception capacity and reservoir flood storage capacity, and obtain the excess flood volume allocation result, wherein the excess flood volume allocation result is composed of the reservoir group flood interception capacity and the lake flood interception capacity, otherwise skip this step and proceed to the next step; Step S38f4, if the flood is of the third magnitude type, execute this step: obtain the reservoir flood control capacity based on the reservoir flood storage capacity, sum up the flood control capacity of each reservoir to obtain the flood control capacity of the reservoir group, calculate the lake flood control capacity based on the reservoir group flood control capacity and excess flood volume, and obtain the excess flood volume allocation result, where the excess flood volume allocation result is composed of the reservoir group flood control capacity and the lake flood control capacity.

8. The reservoir-lake system dispatching method for coupled component-level excess flood volume allocation according to claim 1, characterized in that: The step S42d specifically includes: Step S42d1, given a threshold, dividing the period when the reservoir inflow is higher than the threshold into a high-flow population A, and dividing the period when the reservoir inflow is lower than the threshold into a low-flow population B; Step S42d2, based on the new nonlinear dynamic convergence factor calculated by the exponential function and the new nonlinear dynamic inertia weight factor calculated by the cosine function, the whale optimization algorithm and the particle swarm optimization algorithm are used to optimize and solve the high flow population A and the low flow population B respectively; Step S42d3, arrange the solution results in time series to obtain the joint optimization scheduling result of flood control and drainage of the reservoir-lake system.

Images