A gate-pump group joint optimization scheduling method for complex plain river network areas

Abstract

The application discloses a gate-pump group joint optimization scheduling method for complex plain river network areas and belongs to the technical field of flood control scheduling, which comprises the following steps: based on a long short-term memory network model, a response relationship model of a flood control point water level to meteorological and hydrological conditions and gate-pump engineering scheduling is constructed; a gate-pump group joint optimization scheduling model is constructed in combination with a target function and constraint conditions; the response relationship model and the gate-pump group joint optimization scheduling model are coupled to obtain a gate-pump group multi-objective optimization scheduling model for complex plain river network areas; a non-dominated sorting elitist algorithm for feasible search space optimization is used for multi-objective solution to obtain a non-inferior solution set; a cloud model evaluation method is used for multi-attribute decision of the non-inferior solution set, and a representative solution is preferably selected as a best gate-pump group joint scheduling scheme for complex plain river network areas. The method can solve the problem of excessive complexity caused by direct coupling of river network water dynamics simulation and optimization algorithms and realizes the optimality of gate-pump group joint flood control scheduling in complex plain river network areas.

Description

TECHNICAL FIELD

[0001] The present application relates to a complex plain river network area gate pump group joint optimization scheduling method, belongs to the water resources system planning and sustainable utilization, flood control scheduling technical field. BACKGROUND

[0002] Flood control optimization scheduling is a non-engineering technical means to ensure flood control safety, which is of great significance to fully play the benefit of flood control engineering, improve the flood control capacity of the basin, reduce the flood control pressure of the basin and reduce the loss of flood disaster. Through joint optimization scheduling, the various engineering measures in the flood control and drainage system can be coordinated and compensated, and the flood control effect of the whole is optimized.

[0003] The complex plain river network area has the characteristics of dense river network, numerous water conservancy projects and complex water flow, and the hydrological and hydrodynamic conditions and the flood control and drainage engineering system are extremely complex. The flood evolution process usually needs to be studied through hydrodynamic numerical simulation, and the joint optimization scheduling quantitative analysis is difficult. The numerical solution of Saint-Venant equation set is mostly used for river network hydrodynamic simulation, and the calculation process is complex. The optimization algorithm is to search for the optimal solution through multiple iterations, which usually needs to run thousands of times of loop, and the coupling calculation of the two is too complex and the calculation speed is too slow, so it cannot be applied to actual scheduling. At present, there are few studies on the best joint operation scheduling mode of gate pump group in complex plain river network area based on optimization method.

[0004] Therefore, there are numerous water conservancy engineering facilities such as gates and dams in the complex plain river network area, and there is a close relationship between flood movement and engineering operation. How to organically combine the flood movement simulation model and the engineering optimization scheduling model to realize the optimality of flood control scheduling is an urgent problem to be solved. SUMMARY

[0005] The purpose of the present application is to provide a complex plain river network area gate pump group joint optimization scheduling method, which can solve the problem of excessive complexity of direct coupling of river network hydrodynamic simulation and optimization algorithm, and realize the optimality of gate pump group joint flood control scheduling in complex plain river network area.

[0006] In order to achieve the above purpose, the present application provides the following technical scheme:

[0007] In the first aspect, the present application provides a complex plain river network area gate pump group joint optimization scheduling method, comprising:

[0008] The target area is generalized for flood control scheduling system, and the flood control points and gate pump engineering are determined;

[0009] Based on the long short-term memory network model, a response relationship model of flood control point water level to meteorological and hydrological conditions and gate pump engineering scheduling is constructed;

[0010] By combining the objective function and constraints, a joint optimization scheduling model for the gate pump group is constructed.

[0011] By coupling the response relationship model with the joint optimization scheduling model of gate and pump group, a multi-objective optimization scheduling model of gate and pump group in complex plain river network area is obtained.

[0012] A non-dominated sorting elite algorithm with feasible search space optimization is used to solve a multi-objective optimization scheduling model of gate and pump group in complex plain river network area, and a set of non-dominated solutions is obtained.

[0013] The cloud model evaluation method is used to make multi-attribute decisions on the non-dominated solution set, and the representative solution is selected as the best joint scheduling scheme for gate and pump group in complex plain river network areas.

[0014] In conjunction with the first aspect, further, the input features of the Long Short-Term Memory Network Model include the previous water level at the flood control point, the regional average rainfall, and the discharge flow of the gate pump project, while the output features include the current water level at the flood control point. The structure of the Long Short-Term Memory Network Model consists of two Long Short-Term Memory Network layers and one fully connected layer. It is trained using an adaptive moment estimation optimization algorithm and a regularized loss function. The training data includes historical flood measured data and simulated data after manual adjustment of the gate pump scheduling.

[0015] Building upon the first aspect, the objective function of the joint optimization scheduling model for the gate pump group is further defined as follows:

[0016] ;

[0017] in, This represents the decision variable, namely the discharge flow rate of the gate pump project. express Time period The corresponding flood control point water level;

[0018] The constraints of the joint optimization scheduling model for the gate pump group are:

[0019] ;

[0020] in, Indicates the first Water levels at each flood control point Indicates the first The maximum permissible or warning water level at each flood control control point. Indicates the first The discharge flow of the gate pump project Indicates the first The maximum design flow rate of a gate pump project Indicates the first The first gate pump project corresponding to the Water levels at each flood control point , Indicates the first The first gate pump project corresponding to the Minimum and maximum allowable water levels at each flood control point.

[0021] In conjunction with the first aspect, further feasible search space optimizations include:

[0022] Based on the known maximum design flow of the gate pump project, the precipitation in different time periods of the target area, the water level of the flood control point, and the response relationship between the water level of the flood control point and the discharge flow of the gate pump project, the upper and lower limits of the initial search space for each time period are derived.

[0023] Based on flood control scheduling rules, a flood control and drainage project group scheduling simulation model is used to simulate the discharge flow process of gate pump projects, and the simulated discharge flow of gate pump projects is used as the search benchmark of the feasible search space.

[0024] Based on the upper and lower limits of the initial search space for each time period, the upper and lower limits of the feasible search space for each time period are defined by the intersection of the dynamic search step size and the search benchmark.

[0025] The upper limit of the feasible search space is:

[0026] ;

[0027] in, express The upper limit of the feasible search space for a given time period. express The upper limit of the initial search space for the time period. express Search criteria for time periods Represents the dynamic search step size of the feasible search space;

[0028] The lower bound of the feasible search space is:

[0029] ;

[0030] in, express The lower bound of the feasible search space for a given time period. express The lower bound of the initial search space for the time period.

[0031] In addition to the first aspect, the dynamic search step size is further adjusted dynamically based on the maximum design flow of the gate pump project and historical flood scenarios.

[0032] Building upon the first aspect, furthermore, the cloud model evaluation method is used to perform multi-attribute decision-making on the non-dominated solution set, and the optimal representative solution is selected as the best joint scheduling scheme for gate-pump groups in complex plain river network areas, including:

[0033] Based on the digital characteristics of the cloud model at each level of each evaluation index, cloud droplets are calculated using a positive cloud generator to construct the cloud model;

[0034] Based on the cloud model, the membership degree of each scheme in the non-dominated solution set corresponding to each level of each evaluation index is calculated, and the comprehensive membership degree of each scheme in the non-dominated solution set corresponding to each level is calculated according to the weight of each evaluation index.

[0035] Based on the comprehensive membership degree of each scheme at each level in the non-dominated solution set, the representative solution is selected as the best joint scheduling scheme for gate and pump group in complex plain river network areas.

[0036] The evaluation metrics include positive and negative metrics. The digital features of the cloud model include expectation, entropy, and hyperentropy. The difference between the maximum and minimum values ​​of each evaluation metric is used as the metric. As the interval length, each evaluation indicator is divided into Each level.

[0037] Combining the first aspect, the further expected calculation formula is as follows:

[0038] ;

[0039] in, Indicates the first The first evaluation indicator Expectations at each level , Indicates the first The first evaluation indicator The upper and lower limits of each level;

[0040] The formula for calculating entropy is:

[0041] ;

[0042] in, Indicates the first The first evaluation indicator Entropy at each level, Indicates the first The first evaluation indicator Expectations at each level;

[0043] The formula for calculating hyperentropy is:

[0044] ;

[0045] in, Indicates the first The first evaluation indicator The level of hyperentropy, Indicates the first The first evaluation indicator Entropy at each level;

[0046] The formula for calculating cloud droplets is:

[0047] ;

[0048] in, Represents a random number. Indicates using The first fine-tuning The first evaluation indicator Entropy at each level, Indicates using The first fine-tuning The first evaluation indicator Expectations at each level;

[0049] For positive indicators, the formula for calculating membership degree includes:

[0050] when hour:

[0051] ;

[0052] when hour:

[0053] ;

[0054] when hour:

[0055] ;

[0056] For inverse indicators, the formula for calculating membership degree includes:

[0057] when hour:

[0058] ;

[0059] when hour:

[0060] ;

[0061] when hour:

[0062] ;

[0063] in, Indicates the solution set corresponding to the first solution. the first evaluation index, the membership of the first level, the value of the first evaluation index, , the expectation of the first evaluation index, the total number of levels; the total number of levels; The calculation formula of the comprehensive membership is:

[0064]

[0065]

[0066] wherein, the comprehensive membership vector sequence of the scheme in the non-inferior solution set corresponding to each level, , , the comprehensive membership of the scheme in the non-inferior solution set corresponding to the first, second, …, the nth level, the weight of the first evaluation index, , , the membership of the first evaluation index of the scheme in the non-inferior solution set corresponding to the first, second, …, the nth level, the total number of evaluation indexes. In combination with the first aspect, further, the positive index includes the discharge flow of the gate pump project, and the negative index includes the highest water level of the flood control point and the number of days that the water level of the flood control point exceeds the warning water level. In combination with the first aspect, further, the interval length is 1 / 5 of the difference between the maximum value and the minimum value of each evaluation index, each evaluation index is divided into five levels, and the interval of each level is I (0, a), II (a, b), III (b, c), IV (c, d) and V (d, ∞), wherein a represents the minimum value of the evaluation index plus one interval length, b represents the minimum value of the evaluation index plus two interval lengths, c represents the minimum value of the evaluation index plus three interval lengths, and d represents the minimum value of the evaluation index plus four interval lengths.

[0067] In combination with the first aspect, further, according to the comprehensive membership of each scheme in the non-inferior solution set corresponding to each level, the preferred representative solution is used as the best gate pump group joint scheduling scheme in the complex plain river network area, which includes:

[0068] In combination with the first aspect, further, according to the comprehensive membership of each scheme in the non-inferior solution set corresponding to each level, the preferred representative solution is used as the best gate pump group joint scheduling scheme in the complex plain river network area, which includes:

[0069] In combination with the first aspect, further, according to the comprehensive membership of each scheme in the non-inferior solution set corresponding to each level, the preferred representative solution is used as the best gate pump group joint scheduling scheme in the complex plain river network area, which includes:

[0070] ​​​​​For each scheme in the non-inferior solution set, the highest comprehensive membership degree corresponding to the level is taken as the evaluation level of the scheme;

[0071] The scheme with the highest evaluation level is screened as the preferred scheme;

[0072] The scheme with the highest comprehensive membership degree in the preferred scheme is screened as the best joint scheduling scheme of the gate pump group in the complex plain river network area.

[0073] In a second aspect, the present application provides a joint optimization scheduling system of gate pump groups in a complex plain river network area, comprising:

[0074] a generalization module for generalizing the flood control scheduling system of the target area, determining the flood control points and the gate pump projects;

[0075] a response relationship model construction module for constructing a response relationship model of the water level of the flood control points to the meteorological and hydrological conditions and the scheduling of the gate pump projects based on a long short-term memory network model;

[0076] an optimization scheduling model construction module for constructing a joint optimization scheduling model of the gate pump groups in combination with the objective function and the constraint conditions;

[0077] a model coupling module for coupling the response relationship model and the joint optimization scheduling model of the gate pump groups to obtain a multi-objective optimization scheduling model of the gate pump groups in the complex plain river network area;

[0078] a solution module for performing multi-objective solution on the multi-objective optimization scheduling model of the gate pump groups in the complex plain river network area by using a non-dominated sorting elitism algorithm optimized by a feasible search space to obtain a non-inferior solution set;

[0079] an evaluation module for performing multi-attribute decision on the non-inferior solution set by using a cloud model evaluation method, and optimizing a representative solution as the best joint scheduling scheme of the gate pump groups in the complex plain river network area.

[0080] In a third aspect, the present application provides a computer device, comprising:

[0081] a storage medium for storing a computer program;

[0082] a processor for executing the computer program to realize the joint optimization scheduling method of the gate pump groups in the complex plain river network area according to any one of the first aspect.

[0083] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to realize the joint optimization scheduling method of the gate pump groups in the complex plain river network area according to any one of the first aspect.

[0084] In a fifth aspect, the present application provides a computer program product comprising a computer program which, when executed by a processor, implements the complex plain river network area gate pump group joint optimization scheduling method of any one of the first aspect.

[0085] Compared with the prior art, the present application has the beneficial effects that:

[0086] The complex plain river network area gate pump group joint optimization scheduling method provided by the present application establishes a deep learning model of the response relationship between the flood control control point and the meteorological and hydrological conditions and the gate pump engineering scheduling, to replace the complex hydrodynamic numerical simulation, and then coupled with the optimization algorithm for optimization scheduling calculation, to construct a complex plain river network area gate pump group joint optimization scheduling model, and use the NSGA-II algorithm of feasible search space optimization to solve multiple objectives, and use the cloud model evaluation method to optimize the representative solution of the obtained non-inferior solution set, to solve the problem of excessive complexity of direct coupling of river network hydrodynamic simulation and optimization algorithm, and realize the optimality of gate pump group joint flood control scheduling in the complex plain river network area. BRIEF DESCRIPTION OF DRAWINGS

[0087] Figure 1 is an LSTM model structure schematic diagram provided by an embodiment of the present application;

[0088] Figure 2 is an improved NSGA-II algorithm flowchart provided by an embodiment of the present application;

[0089] Figure 3 is a cloud model schematic diagram provided by an embodiment of the present application;

[0090] Figure 4 is a generalization diagram of a Taihu Basin flood control multi-objective optimization scheduling system provided by an embodiment of the present application;

[0091] Figure 5 is a Taihu Basin flood control optimization scheduling model structure schematic diagram provided by an embodiment of the present application;

[0092] Figure 6 is a target space distribution diagram of a non-inferior solution set provided by an embodiment of the present application, wherein (a) corresponds to a "southern type" design storm, (b) corresponds to a "northern type" design storm, and (c) corresponds to a "whole basin type" design storm;

[0093] Figure 7 is a two-dimensional matrix diagram of a Taihu Basin flood control multi-objective competition relationship provided by an embodiment of the present application, wherein (a) corresponds to a "southern type" design storm, (b) corresponds to a "northern type" design storm, and (c) corresponds to a "whole basin type" design storm;

[0094] Figure 8is a comparison diagram of the optimization scheme and the current rule water level process under the "southern type" design rainstorm scenario provided by the embodiment of the application, wherein (a) corresponds to the simulation process of the water level of Taihu Lake in the non-inferior solution set, (b) corresponds to the simulation process of the water level of Ganlu in the non-inferior solution set, and (c) corresponds to the simulation process of the water level of Pingwang in the non-inferior solution set;

[0095] Figure 9 is a comparison diagram of the optimization scheme and the current rule water level process under the "northern type" design rainstorm scenario provided by the embodiment of the application, wherein (a) corresponds to the simulation process of the water level of Taihu Lake in the non-inferior solution set, (b) corresponds to the simulation process of the water level of Ganlu in the non-inferior solution set, and (c) corresponds to the simulation process of the water level of Pingwang in the non-inferior solution set;

[0096] Figure 10 is a comparison diagram of the optimization scheme and the current rule water level process under the "full-basin type" design rainstorm scenario provided by the embodiment of the application, wherein (a) corresponds to the simulation process of the water level of Taihu Lake in the non-inferior solution set, (b) corresponds to the simulation process of the water level of Ganlu in the non-inferior solution set, and (c) corresponds to the simulation process of the water level of Pingwang in the non-inferior solution set. DETAILED DESCRIPTION

[0097] The technical solutions of the present application will be further described in detail below with reference to specific embodiments.

[0098] The embodiments of the present application will be described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference signs represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be understood as a limitation on the present application. The technical features in the embodiments of the present application and the embodiments can be combined with each other without conflict.

[0099] The embodiment of the present application provides a complex plain river network area gate pump group joint optimization scheduling method, comprising:

[0100] The flood control scheduling system of the target area is generalized, and the flood control control point and the gate pump project are determined;

[0101] Based on the long short-term memory network model, a response relationship model of the water level of the flood control control point to the meteorological and hydrological conditions and the gate pump project scheduling is constructed;

[0102] In combination with the objective function and the constraint condition, a gate pump group joint optimization scheduling model is constructed;

[0103] The response relationship model and the gate pump group joint optimization scheduling model are coupled to obtain a complex plain river network area gate pump group multi-objective optimization scheduling model;

[0104] The non-dominated sorting genetic algorithm II (NSGA-II) with feasible search space optimization is used to solve the multi-objective optimization model of the gate-pump group in the complex plain river network area, and a non-inferior solution set is obtained.

[0105] The cloud model evaluation method is used for multi-attribute decision-making of the non-inferior solution set, and the representative solution is optimized as the best joint scheduling scheme of the gate-pump group in the complex plain river network area.

[0106] In one possible embodiment, the joint optimization scheduling method of the gate-pump group in the complex plain river network area specifically includes the following steps:

[0107] Step 1: Generalization of the flood control scheduling system in the complex plain river network area.

[0108] First, the main flood control points are found by comprehensive analysis, which are used as the representative points of the basin flood control scheduling. Then, the gate-pump projects that play a major role in the flood control of the basin are selected as the objects of optimization scheduling by considering the hydraulic connection of each region. Accordingly, the system generalization diagram of the joint optimization scheduling of the gate-pump group in the complex plain river network area is obtained.

[0109] Step 2: Simulation of the response of the flood control points to the scheduling of the gate-pump projects.

[0110] The responses of the flood control points to the same discharge process of the same gate-pump project are different under different pre-flood levels and different weather conditions. If the hydrological and hydrodynamic model can be directly coupled with the scheduling model, and the simulation results of the hydrological and hydrodynamic model are directly used to evaluate the pros and cons of the scheduling scheme in the optimization scheduling, it is the most accurate and direct way. However, the direct coupling of the river network hydrodynamics calculation and the optimization method is too complex and the calculation cost is too high to be applied to practical applications. Therefore, the Long Short Term Memory (LSTM) model is used in this embodiment to simulate the response of the flood control points to different discharge conditions of each gate-pump project by using historical flood and corresponding scheduling data, that is, to establish a response relationship model between the flood control points and the meteorological and hydrological conditions and the scheduling of the gate-pump projects, to replace the complex hydrodynamic numerical simulation.

[0111] The LSTM model is a deep learning algorithm that uses a threshold mechanism to control the flow and loss of information, which well solves the long-term dependence problem of the Recurrent Neural Network (RNN). The structure of the LSTM model is shown in Figure 1 The LSTM model introduces three thresholds: the input gate , the forget gate , and the output gate ; introduces the cell state representing long-term memory and the candidate state waiting to be stored in long-term memory:

[0112] ;

[0113] where, input gate determines how much information will be stored in the current cell state; forget gate selectively forgets information in the cell state; output gate selectively outputs information in the cell state. All three gates are functions of the current input features and the short-term memory from the previous time step. 、 and are the weight matrices to be trained, 、 and are the bias terms to be trained. is the sigmoid activation function, which makes the range of the gate between 0 and 1.

[0114] ;

[0115] where, is the memory, representing the short-term memory, which is the current cell state after passing through the output gate. The candidate state represents the new knowledge to be stored in the cell state, which is a function of the current input features and the short-term memory from the previous time step. The cell state represents the long-term memory, which is equal to the long-term memory from the previous time step, passing through the forget gate, and the new knowledge induced at the current time step, passing through the input gate.

[0116] The LSTM model constructed in this embodiment includes two LSTM layers and one fully connected layer, and the optimal number of memories is determined by trial and error. The parameter weights and biases are calculated by the backpropagation through time algorithm (BBTT). The mean square error is used as the loss function, and the adaptive moment estimation (Adam) optimization method is used to minimize the mean square error. In order to make all data located in the central region of the sigmoid function, all inputs are normalized to [0, 1] before training the model, avoiding the problem of output signal saturation. At the same time, regularization is used in the training process to avoid overfitting, a penalty term is added to the parameters, and the penalty term and the loss function are used as the optimization target together. The loss function after regularization is:

[0117] ;

[0118] where, is the initial loss function; is the regularization term, which is the sum of the squares of all weights divided by the training sample size , to weight and The ratio of 1 / 2 is for the convenience of differential calculation.

[0119] Step 3: Construct the joint optimization scheduling model of gate-pump group in complex plain river network area.

[0120] Combined with the actual situation of the study area, the objective function and constraint conditions of the joint optimization scheduling of gate-pump group in complex plain river network area are determined. The depth learning model of flood control point response to gate-pump engineering scheduling is coupled with the optimization scheduling to construct the multi-objective optimization scheduling model of gate-pump group in complex plain river network area.

[0121] The objective function of the joint optimization scheduling model of gate-pump group is:

[0122] ;

[0123] Wherein, represents the decision variable, i.e. the discharge of gate-pump engineering, represents the water level of the flood control point corresponding to the time period .

[0124] The constraint conditions of the joint optimization scheduling model of gate-pump group are:

[0125] ;

[0126] Wherein, represents the water level of the th flood control point, represents the highest allowable water level or warning water level of the th flood control point, represents the discharge of the th gate-pump engineering, represents the maximum design flow of the th gate-pump engineering, represents the water level of the th flood control point corresponding to the th gate-pump engineering, , represents the lowest allowable water level or highest allowable water level of the th flood control point corresponding to the th gate-pump engineering.

[0127] Step 4: Multi-objective solution is carried out by using improved non-dominated sorting genetic algorithm II (NSGA-II).

[0128] The NSGA-II algorithm employs a fast non-dominated sorting algorithm, which significantly reduces computational complexity compared to the NSGA algorithm. It utilizes crowding degree and crowding degree comparison operators instead of specifying a shared radius, and uses them as the winning criterion in peer comparisons after quicksort, thus maintaining population diversity. Furthermore, it introduces an elite retention strategy, expands the sampling space, prevents the loss of the best individual, and improves the algorithm's computational speed and robustness.

[0129] When using the NSGA-II algorithm for multi-dimensional variable multi-objective optimization, an excessively large search space can reduce search efficiency; and an unreasonable evolutionary direction can lead to premature convergence and getting trapped in local optima. Therefore, this embodiment introduces feasible search space optimization techniques based on the traditional NSGA-II algorithm, proposing an improved NSGA-II algorithm that improves computational efficiency while maintaining the algorithm's global search capability. The feasible search space optimization process is as follows:

[0130] Step 1: Determine the initial search space.

[0131] In this embodiment, the decision variable is the discharge flow of the main gate pumping station. Given the constraints of the maximum discharge flow of the gate pumping station, the regional precipitation in the basin during different time periods, the relationship between the water level and the discharge flow at the flood control point, and the water level constraints at the flood control point, the initial upper and lower limits of the discharge flow for each time period are calculated. , .

[0132] Step 2: Obtain the search baseline.

[0133] Based on the current flood control scheduling rules of the Taihu Lake Basin, a simulation model of the flood control and drainage engineering group scheduling was used to simulate the discharge flow process of the main gate pumping projects under the same design rainstorm scenario. The simulated discharge flow process was used as the search benchmark in the feasible search space of the multi-objective optimization model.

[0134] Step 3: Determine the feasible search space.

[0135] Select dynamic search step size This is then combined with the search baseline to construct a temporary search space. The intersection of the initial search space and the temporary search space is taken as the upper and lower boundaries of the feasible search space for each time period. The upper limit of the feasible search space is:

[0136] ;

[0137] in, express The upper limit of the feasible search space for a given time period. express The upper limit of the initial search space for the time period. express Search criteria for time periods Dynamic search step length representing the feasible search space.

[0138] The lower limit of the feasible search space is:

[0139] ;

[0140] Wherein, represents The lower limit of the time period feasible search space, represents The lower limit of the time period initial search space.

[0141] The feasible search space optimization technique is applied to improve the NSGA-II algorithm, and the specific solving steps are shown in Figure 2 .

[0142] In one possible embodiment, the dynamic search step length is dynamically adjusted according to the maximum design flow of the gate pump project and the historical flood scenario.

[0143] Step 5: The cloud model evaluation method is used to select the representative solution from the obtained non-inferior solution set.

[0144] The cloud model is a model for converting the uncertainty between the qualitative concept and its quantitative numerical representation, and has better characteristics of describing the randomness and fuzziness of variables. The specific evaluation steps are as follows:

[0145] Step 1: Divide each evaluation index into 5 levels with 1 / 5 of the difference between the maximum and minimum value as the interval length, and the interval of each level is I (0, a), II (a, b), III (b, c), IV (c, d) and V (d, ∞), wherein a represents the minimum value of the evaluation index plus 1 interval length, b represents the minimum value of the evaluation index plus 2 interval lengths, c represents the minimum value of the evaluation index plus 3 interval lengths, and d represents the minimum value of the evaluation index plus 4 interval lengths. The evaluation indexes include positive indexes and reverse indexes, the positive indexes include the discharge of the gate pump project, and the reverse indexes include the highest water level of the flood control point and the duration of the water level of the flood control point exceeding the warning water level. Determine the cloud model digital characteristics of each level of each index:

[0146] The expected calculation formula is:

[0147] ;

[0148] Wherein, represents the expected value of the first evaluation index in the first level, , represents the expected value of the first evaluation index in the first level, , ​​The upper and lower limits of each level.

[0149] The formula for calculating entropy is:

[0150] ;

[0151] in, Indicates the first The first evaluation indicator Entropy at each level, Indicates the first The first evaluation indicator Expectations at each level.

[0152] The formula for calculating hyperentropy is:

[0153] ;

[0154] in, Indicates the first The first evaluation indicator The level of hyperentropy, Indicates the first The first evaluation indicator Entropy at each level. The threshold can be adjusted according to the fuzzy threshold of the variable; in this embodiment, it is set to 1. .

[0155] Step 2: Calculate cloud droplets based on the forward cloud generator and construct a cloud model:

[0156] ;

[0157] in, Represents a random number. Indicates using The first fine-tuning The first evaluation indicator Entropy at each level, Indicates using The first fine-tuning The first evaluation indicator The expected value is calculated at each level. This calculation is repeated 1000 times to obtain 1000 cloud droplets, forming a cloud model, the schematic of which is shown below. Figure 3 As shown. Figure 3 middle, Expressing expectations, Represents entropy, This represents hyperentropy.

[0158] Step 3: Calculate the membership degree of each indicator for each scheme at each level:

[0159] For positive indicators, the formula for calculating membership degree includes:

[0160] When ;

[0161] ;

[0162] When ;

[0163] ;

[0164] When ;

[0165] ;

[0166] For the reverse index, the calculation formula of membership degree includes:

[0167] When ;

[0168] ;

[0169] When ;

[0170] ;

[0171] When ;

[0172] ;

[0173] wherein, represents the membership degree of the scheme corresponding to the first evaluation index in the non-inferior solution set, represents the value of the first evaluation index, , represents the expectation of the first evaluation index in the first grade, represents the expectation of the first evaluation index in the second grade,

[0174] represents the total number of grades. Repeat 1000 times for each scheme, obtain 1000

[0175] , and determine the membership degree of each index under each grade by averaging:

[0176] Step 4: Calculate the comprehensive membership degree of each scheme to each grade.

[0177] First, normalize the membership degrees of the five grades, and then assign the index weight value, let be the membership degree vector sequence of the first index, and let is the weight vector of each index, and the comprehensive membership vector sequence is synthesized as follows:

[0178]

[0179] wherein, represents the comprehensive membership vector sequence of the scheme in the non-inferior solution set corresponding to each level, , represents the comprehensive membership of the scheme in the non-inferior solution set corresponding to the 1st, 2nd, …, represents the weight of the 1st evaluation index, , represents the membership of the scheme in the non-inferior solution set corresponding to the 1st, 2nd, …, represents the total number of evaluation indexes.

[0180] Step 5: According to the comprehensive membership of each scheme in the non-inferior solution set corresponding to each level, the representative solution is optimized as the best joint scheduling scheme of the complex plain river network area.

[0181] For each scheme in the non-inferior solution set, the level corresponding to the highest comprehensive membership is taken as its evaluation level; the scheme with the highest evaluation level is selected as the preferred scheme; the scheme with the highest comprehensive membership in the preferred scheme is selected as the best joint scheduling scheme of the complex plain river network area.

[0182] Step 6: The water power model is called again to simulate the optimization scheme of the representative solution to verify whether the actual scheduling result is reasonable and feasible.

[0183] In one possible embodiment, the joint optimization scheduling method of the complex plain river network area lock-pump group specifically includes the following steps:

[0184] Step 1: System generalization.

[0185] ​​​​​​​The flood control scheduling of Taihu Basin mainly focuses on the safe storage and discharge of flood in Taihu Lake. Firstly, the safety of the dikes around Taihu Lake is ensured, and the safety of Shanghai, Suzhou, Wuxi, Changzhou, Zhenjiang, Hangzhou, Jiaxing, Huzhou and other cities and important facilities are protected. Influenced by the terrain, the flood in the west of Taihu Lake is basically discharged into Taihu Lake; part of the flood in the west of Zhejiang Province is discharged into Hangzhou Bay through Hangjiahu District, but a large part of the flood is directly discharged into Taihu Lake. The other four subareas have water exchange with Taihu Lake. When the basin flood occurs, the water level of each subarea rises rapidly, and a large amount of water is discharged into Taihu Lake at this time. There are two main flood discharge channels of Taihu Lake: through Wangting Water Control Center to discharge into Wangyu River, and then through Changshu Jiangbian Water Control Center to discharge into Yangtze River; through Taipu Gate to discharge into Taipu River, and then through Huangpu River to discharge into Yangtze River. At the same time, a small part of the flood can be discharged into Wusong River through Guajingkou Water Control Center, and then discharged into Taihu Lake.

[0186] (1) Select the flood control point.

[0187] The flood control point is selected from the representative station and the control section: the representative station can represent the flood control situation of the area where it is located. If the representative station meets the flood control requirements, the area it represents can also meet the flood control requirements. In order to meet the flood control requirements of some cities or regions, the flood control requirements of the control section of these cities or regions must be met.

[0188] In this embodiment, the water level of Taihu Lake is selected first. The water level of Taihu Lake is the main standard for measuring the flood situation of Taihu Lake, and determines the operation mode of Wangyu River, Taipu River and surrounding regional gate pump engineering, which is particularly important in the flood control scheduling of Taihu Basin. Secondly, the water level representative points of Wangyu River and Taipu River, Ganlu and Pingwang, are selected. The water level of these stations determines the operation mode of the surrounding basin-level flood control water control center. In addition, considering the similarity of the location of the station and the flood control situation of the area where the station is located, the water level of Changzhou, Wuxi, Suzhou, Xiangcheng, Chenmu, Jiaxing and Hangchangqiao is selected for key calculation.

[0189] (2) Select the main gate pump engineering.

[0190] There are thousands of large and small gate pump engineering in Taihu Basin. Different gate pump engineering has different effects and importance on the flood control of the basin. Some gate pumps can only affect the water level and flow of local area for a short time, and some gate pumps can control the flood situation of a large area for a long time.

[0191] In this embodiment, Wangyu River and Taipu River are the main flood discharge channels of Taihu Lake, and the operation of water conservancy projects on them is the key to flood control in the basin. The water conservancy projects on Wangyu River mainly include Wangting Interchange Water Conservancy Hub, gates and pumps on both banks of Wangyu River, and Changshu Jiang Edge Hub. Among them, Wangting Interchange Water Conservancy Hub is directly connected with Taihu Lake, and Changshu Jiang Edge Hub is directly connected with Yangtze River. The joint operation of these two projects determines the amount of flood discharged by Wangyu River; the operation of gates and pumps on both banks determines the amount of flood discharged by Wangyu River to the two banks. Taipu River mainly relies on the operation of Taipu Gate to control the flood discharged by Taihu Lake. Therefore, in this embodiment, Wangting Interchange Water Conservancy Hub, Changshu Jiang Edge Hub and Taipu Gate are selected as the main gate and pump projects at the basin level to carry out the calculation of optimal flood control scheduling.

[0192] Based on the above analysis, the system generalization diagram of the multi-objective optimal scheduling model for flood control in Taihu Basin is as shown in Figure 4

[0193] Step 2: Simulate the response of flood control points to the scheduling of gate and pump projects.

[0194] Using LSTM deep learning algorithm, water level simulation models of 10 key flood control points (Taihu, Ganlu, Pingwang, Changzhou, Wuxi, Suzhou, Xiangcheng, Chenmu, Jiaxing, Hangchangqiao) are established. The input factors of the model include: the previous water level of Taihu, Ganlu, Pingwang, Changzhou, Wuxi, Suzhou, Xiangcheng, Chenmu, Jiaxing, Hangchangqiao; the current period average rainfall of the west of the lake, the west of Zhejiang, Taihu area, Wuchengxiyu area, Yangchengdenglai area, Hangjiahu area; the current period discharge of Wangting Interchange, Changshu Hub, Taipu Gate. The output factor is the water level of the key flood control point in the current period. Select the measured data from 1990 to 2017 and the simulation data after artificially modifying the discharge of gate and pump projects, 70% for training and 30% for testing. The accuracy of the 10 water level simulation models in the calibration period and the verification period is shown in Table 1.

[0195] Table 1: Accuracy of 10 water level simulation models in the calibration period and the verification period.

[0196]

[0197] As shown in Table 1, the water level simulation accuracy of the flood control points based on deep learning is high, and the determinacy coefficients in the verification period can all reach more than 0.9, which can be used to replace the hydrological and hydrodynamic models to simulate the water level of each scheduling mode considered in the optimal scheduling. Among them, the water level simulation models of Taihu, Ganlu and Pingwang are applied to calculate the objective function and the constraint condition, and the other water level simulation models are applied to calculate the constraint condition to restrict the water level of each flood control point not to exceed the warning water level or the highest water level allowed.

[0198] Step 3: Build a joint optimal scheduling model for complex plain river network area gate and pump groups.​

[0199] In order to carry out the flood control optimization scheduling of Taihu Basin, the main gate and pump projects in the basin are coordinated and cooperated, so as to achieve the overall optimal flood control effect. The objective function and constraint conditions are as follows:

[0200] (1) Objective function.

[0201] The objective function of the embodiment is the highest water level of Taihu, the representative station Gannu of Wangyu River and the representative station Pingwang of Taipu River, and the highest water level of the three points is minimum:

[0202] ;

[0203] Wherein, The decision variable in the embodiment is the discharge process of three main gate and pump projects (Wangting Interchange, Jiangbian Hub of Changshu and Taipu Gate). The Taihu water level at each time period when the three gate discharge decision series is The Gannu water level at each time period when the three gate discharge decision series is The Pingwang water level at each time period when the three gate discharge decision series is

[0204] (2) Constraint conditions.

[0205] ① Highest water level constraint:

[0206] For the flood control of Taihu Basin, the flood control of each flood control point must be met.

[0207] ;

[0208] Wherein, Taihu, Gannu, Pingwang, Changzhou, Wuxi, Suzhou, Xiangcheng, Chenmiao, Jiaxing and Hangchangqiao respectively represent the flood control points.

[0209] ② Discharge constraint:

[0210] ;

[0211] Wherein, Wangting interchange discharge, Changshu Jiangbian hub discharge and Taipu gate discharge respectively represent.

[0212] ③ Gate and pump project water level constraint:

[0213] .

[0214] ④ Variable non-negative constraint:

[0215] ​​​The above variables should satisfy the non-negative requirement.

[0216] (3) Model structure.

[0217] The present embodiment is directed to the problem of excessive complexity of directly coupling river network hydrodynamic simulation and optimization algorithm. The above LSTM deep learning algorithm is used to replace river network hydrodynamic simulation, and then coupled with the optimization algorithm for optimization scheduling calculation. The structure of the flood control optimization scheduling model of the Taihu Lake Basin constructed in the present embodiment is shown in Figure 5 .

[0218] Step 4: Multi-objective solution is performed by using the improved non-dominated sorting genetic algorithm (NSGA-II).

[0219] Three design rainstorm scenarios are set, which are the “southern type”, “northern type” and “whole basin type” 100-year design rainstorm processes from June 1st to July 31st in the flood season of the Taihu Lake Basin. The scheduling period is 8 hours. The improved NSGA-II algorithm is used to solve the multi-objective optimization scheduling model of the Taihu Lake Basin flood control, the population size is set to 100, the iteration number is 2000, the crossover probability is 0.5, and the mutation probability is 0.1. The distribution of non-inferior solution set in the target space under the three design rainstorm scenarios is shown in Figure 6 .

[0220] Figure 7 The box plot (diagonal) of the value range of the non-inferior solution set of the three targets of the highest water level of the Taihu Lake, the highest water level of the Ganlu, and the highest water level of the Pingwang under the three design rainstorm scenarios, and the scatter plot of the competition relationship between each other.

[0221] As shown in Figure 7 (a), under the “southern type” design rainstorm scenario, the non-inferior solution set of the highest water level of the Taihu Lake has a target value range of 4.904~5.076m (average value of 4.982m), the non-inferior solution set of the highest water level of the Ganlu has a target value range of 4.180~4.396m (average value of 4.277m), and the non-inferior solution set of the highest water level of the Pingwang has a target value range of 4.240~4.504m (average value of 4.365m). As shown in Figure 7 (b), under the “northern type” design rainstorm scenario, the non-inferior solution set of the highest water level of the Taihu Lake has a target value range of 4.936~5.181m (average value of 5.039m), the non-inferior solution set of the highest water level of the Ganlu has a target value range of 4.248~4.531m (average value of 4.402m), and the non-inferior solution set of the highest water level of the Pingwang has a target value range of 4.032~4.159m (average value of 4.098m). As shown in Figure 7From the middle (c), it can be seen that under the design storm scenario of "full basin type", the non-inferior solution set of the highest water level of Taihu Lake is 4.732~4.978 m (average value is 4.861 m), the non-inferior solution set of the highest water level of Ganlu is 4.041~4.385 m (average value is 4.201 m), and the non-inferior solution set of the highest water level of Pingwang is 3.934~4.073 m (average value is 3.977 m). The analysis of the competition relationship between the targets shows that with the increase of the highest water level of Taihu Lake, the highest water level of Pingwang decreases significantly, and the highest water level of Ganlu also has a general decreasing trend, that is, the projection of the non-inferior solution set on the Taihu Lake-Pingwang plane is relatively concentrated; the projection on the Taihu Lake-Ganlu plane is slightly dispersed, but there is a certain rule; and the projection on the Ganlu-Pingwang plane is relatively scattered. The above phenomenon shows that under different design storm scenarios, there is a strong competition relationship between the highest water level of Taihu Lake and the highest water level of Pingwang, and there is a certain competition relationship between the highest water level of Taihu Lake and the highest water level of Ganlu.

[0222] In fact, under the non-inferior solution set of any design storm scenario, the reduction of any water level target will inevitably sacrifice other water level targets, and the difference is only in the degree of target replacement, which is the embodiment of the competition and game relationship between the multi-objective water levels of the key flood control points in Taihu Basin. The reduction of the highest water level of Taihu Lake will cause the simultaneous increase of the highest water levels of Pingwang and Ganlu, because the increased discharge of Taihu Lake will inevitably cause the increase of the water levels of the two representative stations. The joint regulation of Changshu hub in the lower reaches of Wangyu River can eliminate the influence of the increased discharge of Wanting overpass on the water level of Ganlu to a certain extent. However, there is no hub project for joint regulation in the lower reaches of Taipu River, so when the discharge of Taipu Gate is increased, the water level of Pingwang increases significantly.

[0223] Step 5: Optimize the representative solution by using cloud model evaluation method for the obtained non-inferior solution set.

[0224] According to the cloud model evaluation method, the non-inferior solution set under different design storm scenarios is subjected to multi-attribute decision, and the top ten non-inferior solutions under each scenario are selected, as shown in Table 2.

[0225] Table 2: Comparison of the top ten non-inferior solutions and simulation results.

[0226]

[0227] Step 6: Re-call the water dynamic model to simulate the optimization scheme of the representative solution to verify whether the actual regulation result is reasonable and feasible.

[0228] Since the multi-objective optimization model only considers the case of 10 key flood control points, some schemes may be harmful to the areas not considered, although they are beneficial to the key flood control points, and are not suitable as the optimal scheduling scheme. Therefore, it is necessary to apply a hydrological and hydrodynamic model to simulate the scheduling scheme obtained by the multi-objective optimization, to comprehensively understand the scheduling results of each element in the basin by the scheme, so as to serve as a basis for scheme evaluation. At the same time, the embodiment adopts deep learning to replace hydrological and hydrodynamic simulation for optimization, and is not actually simulated scheduling and flood process. The hydrodynamic simulation of the obtained scheduling scheme can also serve the purpose of verifying whether the representative non-inferior solution is feasible. Therefore, the embodiment uses the hydrodynamic model to simulate the optimization scheme corresponding to the top ten non-inferior solutions to verify whether the optimal scheduling is reasonable, and the corresponding simulation scheduling results are shown in Table 2.

[0229] The results of the hydrodynamic simulation in Table 2 show that the optimization schemes obtained in the embodiment are feasible, and the non-inferior solutions and the simulation results are not much different. Therefore, the optimization scheduling technology coupled by deep learning and multi-objective evolutionary algorithm is feasible, and the optimization schemes corresponding to the non-inferior solution set can be used as scheduling plans for decision makers to refer to, and the simulation results can be used as a basis for decision-making of different scheduling plans.

[0230] The embodiment draws the optimization water level process line of the top ten schemes under each scenario, and compares it with the water level process line of the current rule scheduling, as shown in Figures 8 to 10 The water level process line of the current rule scheduling refers to the water level process line after scheduling under each design rainstorm scenario according to the current flood and water scheduling scheme of the Taihu Basin; the optimization water level process line of the top ten schemes refers to the water level process line obtained after hydrodynamic simulation of the optimization scheme corresponding to the top ten non-inferior solutions; the shaded part is the optimization water level process line range under different scenarios, which is obtained by taking the outer envelope of all water level optimization process lines.

[0231] As can be seen from Figures 8 to 10 , through optimization scheduling, the Taihu flood peak water level is reduced, and the overall water level is reduced; the Ganlu flood peak water level is reduced; the Pingwang overall water level is slightly raised. It shows that the Taihu flood can be discharged through the optimization scheduling of the Wangting Interchange and the Taipu Gate, so as to reduce the water level of the Taihu. Increasing the discharge flow, the water level of the representative station of the Wangyu River and the Taipu River will inevitably rise, however, through the joint optimization scheduling of the Wangting Interchange and the Changshu Hub, the influence of increasing the discharge on the Ganlu water level can be eliminated to a large extent. However, there is a lack of gate and pump engineering for joint regulation and control in the lower reaches of the Taipu River, so the Pingwang water level is sensitive to the Taipu gate discharge. When the discharge is increased, the Pingwang water level rises to a certain extent, but it is within an acceptable range.

[0232] The embodiment of the present application provides a gate and pump group joint optimization scheduling system in a complex plain river network area, which comprises:

[0233] an abstraction module configured to abstract a flood control scheduling system for a target region, and determine a flood control point and a gate-pump project;

[0234] a response relationship model construction module configured to construct a response relationship model of a water level of the flood control point to meteorological and hydrological conditions and gate-pump project scheduling based on a long short-term memory network model;

[0235] an optimization scheduling model construction module configured to construct a joint optimization scheduling model of the gate-pump group in combination with an objective function and a constraint condition;

[0236] a model coupling module configured to couple the response relationship model and the joint optimization scheduling model of the gate-pump group to obtain a gate-pump group multi-objective optimization scheduling model for a complex plain river network region;

[0237] a solution module configured to perform multi-objective solution on the gate-pump group multi-objective optimization scheduling model for the complex plain river network region by using a non-dominated sorting genetic algorithm II (NSGA-II) to obtain a non-inferior solution set;

[0238] an evaluation module configured to perform multi-attribute decision on the non-inferior solution set by using a cloud model evaluation method, and optimally select a representative solution as a best joint scheduling scheme of the gate-pump group for the complex plain river network region.

[0239] The gate-pump group joint optimization scheduling system for the complex plain river network region provided in this embodiment can execute the gate-pump group joint optimization scheduling method provided in any embodiment of the present application, and has the corresponding function modules and beneficial effects of the execution method.

[0240] The embodiment of the present application provides a computer device, comprising:

[0241] a storage medium configured to store a computer program;

[0242] a processor configured to execute the computer program to implement the gate-pump group joint optimization scheduling method provided in any embodiment of the present application.

[0243] The embodiment of the present application provides a computer readable storage medium having a computer program stored thereon, and the computer program is executed by a processor to implement the gate-pump group joint optimization scheduling method provided in any embodiment of the present application.

[0244] The embodiment of the present application provides a computer program product comprising a computer program, and the computer program is executed by a processor to implement the gate-pump group joint optimization scheduling method provided in any embodiment of the present application.

[0245] Those skilled in the art will appreciate that embodiments of the application can be readily used as software, hardware, or a combination of software and hardware. In a software embodiment, various elements are implemented in hardware, software, or a combination of both hardware and software. In a software embodiment, the software implementation can include but is not limited to a bit stream, a machine executable file, a program, or a software application. In a software embodiment, the various elements are implemented as software programs to provide the functionality described herein. It will be apparent to those skilled in the art that substantial equivalents of the structures described herein can be substituted for the structures described and described herein without deviating from the spirit and scope of the application.

[0246] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0247] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0248] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0249] The above merely preferred embodiments of the present application and it is to be understood that those skilled in the technical field of the present technology can make a number of improvements and variations thereto without departing from the technical principles of the present application, which improvements and variations should also be considered as falling within the scope of the present application.

Claims

1. A method for joint optimization scheduling of sluice gate and pump group in complex plain river network areas, characterized in that, include: A flood control dispatch system for the target area was generalized, and flood control control points and gate / pump projects were identified. Based on the long short-term memory network model, a response relationship model of flood control point water level to meteorological and hydrological conditions and gate pump project scheduling is constructed. By combining the objective function and constraints, a joint optimization scheduling model for the gate pump group is constructed. By coupling the response relationship model with the joint optimization scheduling model of gate and pump group, a multi-objective optimization scheduling model of gate and pump group in complex plain river network area is obtained. A non-dominated sorting elite algorithm with feasible search space optimization is used to solve a multi-objective optimization scheduling model of gate and pump group in complex plain river network area, and a set of non-dominated solutions is obtained. The cloud model evaluation method is used to make multi-attribute decisions on the non-dominated solution set, and the representative solution is selected as the best joint scheduling scheme for gate and pump group in complex plain river network areas. The objective function of the joint optimization scheduling model for gate pump groups is: ； in, This represents the decision variable, namely the discharge flow rate of the gate pump project. express Time period The corresponding flood control point water level; The constraints of the joint optimization scheduling model for the gate pump group are: ； in, Indicates the first Water levels at each flood control point Indicates the first The maximum permissible or warning water level at each flood control control point. Indicates the first The discharge flow of the gate pump project Indicates the first The maximum design flow rate of a gate pump project Indicates the first The first gate pump project corresponding to the Water levels at each flood control point , Indicates the first The first gate pump project corresponding to the Minimum and maximum allowable water levels at each flood control point.

2. The method for joint optimization scheduling of gate and pump groups in complex plain river network areas according to claim 1, characterized in that, The input features of the Long Short-Term Memory (LSTM) network model include the previous water level at the flood control point, the regional average rainfall, and the discharge flow of the gate pump project. The output features include the current water level at the flood control point. The structure of the LTM network model consists of two LTM network layers and one fully connected layer. It is trained using an adaptive moment estimation optimization algorithm and a regularized loss function. The training data includes historical flood measured data and simulated data after manual adjustment of the gate pump scheduling.

3. The method for joint optimization scheduling of gate and pump groups in complex plain river network areas according to claim 1, characterized in that, Feasible search space optimizations include: Based on the known maximum design flow of the gate pump project, the precipitation in different time periods of the target area, the water level of the flood control point, and the response relationship between the water level of the flood control point and the discharge flow of the gate pump project, the upper and lower limits of the initial search space for each time period are derived. Based on flood control scheduling rules, a flood control and drainage project group scheduling simulation model is used to simulate the discharge flow process of gate pump projects, and the simulated discharge flow of gate pump projects is used as the search benchmark of the feasible search space. Based on the upper and lower limits of the initial search space for each time period, the upper and lower limits of the feasible search space for each time period are defined by the intersection of the dynamic search step size and the search benchmark. The upper limit of the feasible search space is: ； in, express The upper limit of the feasible search space for a given time period. express The upper limit of the initial search space for the time period. express Search criteria for time periods Represents the dynamic search step size of the feasible search space; The lower bound of the feasible search space is: ； in, express The lower bound of the feasible search space for a given time period. express The lower bound of the initial search space for the time period.

4. The method for joint optimization scheduling of gate and pump groups in complex plain river network areas according to claim 3, characterized in that, The dynamic search step size is dynamically adjusted based on the maximum design flow of the gate pump project and historical flood scenarios.

5. The method for joint optimization scheduling of gate and pump groups in complex plain river network areas according to claim 1, characterized in that, The cloud model evaluation method is used to perform multi-attribute decision-making on the non-dominated solution set, and the optimal representative solution is selected as the best joint scheduling scheme for gate and pump group in complex plain river network areas, including: Based on the digital characteristics of the cloud model at each level of each evaluation index, cloud droplets are calculated using a positive cloud generator to construct the cloud model; Based on the cloud model, the membership degree of each scheme in the non-dominated solution set corresponding to each level of each evaluation index is calculated, and the comprehensive membership degree of each scheme in the non-dominated solution set corresponding to each level is calculated according to the weight of each evaluation index. Based on the comprehensive membership degree of each scheme at each level in the non-dominated solution set, the representative solution is selected as the best joint scheduling scheme for gate and pump group in complex plain river network areas. The evaluation metrics include positive and negative metrics. The digital features of the cloud model include expectation, entropy, and hyperentropy. The difference between the maximum and minimum values ​​of each evaluation metric is used as the metric. As the interval length, each evaluation indicator is divided into Each level.

6. The method for joint optimization scheduling of gate and pump groups in complex plain river network areas according to claim 5, characterized in that, The formula for calculating the expected value is: ； in, Indicates the first The first evaluation indicator Expectations at each level , Indicates the first The first evaluation indicator The upper and lower limits of each level; The formula for calculating entropy is: ； in, Indicates the first The first evaluation indicator Entropy at each level, Indicates the first The first evaluation indicator Expectations at each level; The formula for calculating hyperentropy is: ； in, Indicates the first The first evaluation indicator The level of hyperentropy, Indicates the first The first evaluation indicator Entropy at each level; The formula for calculating cloud droplets is: ； in, Represents a random number. Indicates using The first fine-tuning The first evaluation indicator Entropy at each level, Indicates using The first fine-tuning The first evaluation indicator Expectations at each level; For positive indicators, the formula for calculating membership degree includes: when hour: ； when hour: ； when hour: ； For inverse indicators, the formula for calculating membership degree includes: when hour: ； when hour: ； when hour: ； in, Indicates the solution set corresponding to the first solution. The first evaluation indicator Membership degree at each level, Indicates the first The value of each evaluation indicator , Indicates the first The first evaluation indicator Expectations at each level Indicates the total number of grades; The formula for calculating the overall membership degree is: ； in, This represents the sequence of comprehensive membership vectors for each level of the solutions in the non-dominated solution set. , … This indicates that the solutions in the non-dominated solution set correspond to the 1st, 2nd, ..., The overall membership degree of each level, Indicates the first The weight of each evaluation indicator, , … Indicates the solution set corresponding to the first solution. The first, second, ..., evaluation indicators Membership degree at each level, This indicates the total number of evaluation indicators.

7. The method for joint optimization scheduling of gate and pump groups in complex plain river network areas according to claim 5, characterized in that, Positive indicators include the discharge flow of gate pump projects, while negative indicators include the highest water level at flood control points and the number of consecutive days that the water level at flood control points exceeds the warning level.

8. The method for joint optimization scheduling of gate and pump groups in complex plain river network areas according to claim 5, characterized in that, Using 1 / 5 of the difference between the maximum and minimum values ​​of each evaluation indicator as the interval length, the evaluation indicators are divided into 5 levels. The intervals for each level are I (0, a), II (a, b), III (b, c), IV (c, d) and V (d, ∞), where a represents the minimum value of the evaluation indicator plus 1 interval length, b represents the minimum value of the evaluation indicator plus 2 interval lengths, c represents the minimum value of the evaluation indicator plus 3 interval lengths, and d represents the minimum value of the evaluation indicator plus 4 interval lengths.

9. The method for joint optimization scheduling of gate and pump groups in complex plain river network areas according to claim 5, characterized in that, Based on the comprehensive membership degree of each scheme at each level in the non-dominated solution set, the preferred representative solutions as the optimal joint scheduling schemes for gate-pump groups in complex plain river network areas include: For each solution in the set of non-dominated solutions, the level corresponding to its highest comprehensive membership degree is taken as its evaluation level; The solution with the highest evaluation rating is selected as the preferred solution; The scheme with the highest comprehensive membership degree among the selected schemes is chosen as the optimal joint scheduling scheme for gate and pump groups in complex plain river network areas.

Images