A rapid response method and system for emergency dispatch of urban rainwater storage tanks

Abstract

The present invention discloses a rapid response method and system for emergency dispatch of urban rainwater storage tanks, including: the present invention accurately simulates and calculates the rainfall-induced waterlogging and storage tank dispatch process through a hydrological and hydrodynamic model, and then based on the simulation calculation results, uses an adaptive machine learning model to directly establish the relationship between rainfall parameters, storage tank operation rules and dispatch targets, and uses it to replace the time-consuming numerical model as an optimization solver to solve in the optimization algorithm, avoiding repeated calls to the time-consuming numerical model during the optimization process, thereby improving the solution efficiency. Under the premise of ensuring the calculation accuracy, the present invention greatly shortens the emergency dispatch response time, strives for a large amount of lead time for the dispatch decision of the rainwater storage tank, and has important reference value for the formulation of urban emergency response strategies.

Description

Technical Field

[0001] The invention belongs to the technical field of urban flood optimization dispatching, and relates to a rapid response method and system for emergency dispatching of urban rainwater storage tanks. Background Art

[0002] Rainwater storage tanks are usually used as a comprehensive storage facility, and have achieved good results in urban waterlogging prevention and control and urban rainwater pollution. In the optimization and scheduling of the operation rules of storage tanks, the scheme selection based on scenario analysis is still the main method, and the optimization effect depends on the experience of the operator, especially when dealing with multiple constraints. The optimization cannot play its due operational benefits. In addition, as an important response measure for urban flood control and drainage, there are few cases that directly link the operation rules of storage tanks with urban surface waterlogging.

[0003] Considering that the actual operation of the storage tank consumes a lot of costs, the optimization and scheduling of the operation rules of the storage tank still needs to be calculated with the help of numerical simulation. Urban areas have high building density and complex surface characteristics. If you want to characterize the real terrain to describe the surface flow, you need to accurately simulate and calculate through a high-precision two-dimensional surface hydrodynamic model. In order to obtain high-precision calculation results, the optimization solution of the operation and scheduling rules of flood control and drainage projects will inevitably cause a huge amount of calculation. Therefore, a rapid regulation method is urgently needed for urban rainwater storage tanks. Summary of the invention

[0004] The purpose of the present invention is to solve the problem in the prior art that the operating cost of the regulating reservoir is high and the optimization and scheduling of the operating rules of the regulating reservoir requires huge computing power, and to provide a rapid response method and system for emergency scheduling of urban rainwater regulating reservoirs.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A rapid response method for emergency dispatch of urban rainwater storage tanks, comprising:

[0007] Based on the operation modes of water inflow, emptying and flushing of the regulating reservoir, the operation scheduling model of the rainwater regulating reservoir is established by comprehensively considering the waterlogging risk level in the regulating range, the regulating operation efficiency of the regulating reservoir, the operation cost and the model constraint function.

[0008] Construct a particle swarm algorithm based on the mutant whale hunting operator;

[0009] Based on the case study hydrological and meteorological data, high-precision terrain image data, land use type and infiltration data, as well as pipe network and storage tank data, a hydrological and hydrodynamic model of the coupled storage tank is established;

[0010] The rainfall characteristic parameters and storage tank operation parameters are used as the model training input data set, and the objective function and constraint conditions of the rainwater storage tank operation scheduling model are used as the model training target data set. Based on the model training input data set, model training target data set and adaptive machine learning model, a machine learning model suitable for emergency scheduling of urban rainwater storage tanks with the highest accuracy is obtained.

[0011] An adaptive machine learning model is used instead of the hydrological and hydrodynamic model as the solver of the particle swarm algorithm of the mutant whale hunting operator. The population size of the particle swarm algorithm of the mutant whale hunting operator and the maximum running algebraic parameter settings are set to output the optimal objective function value and the optimal scheduling plan for the urban rainwater storage tank.

[0012] A further improvement of the present invention is:

[0013] Furthermore, based on the operating parameters of the regulating reservoir, the flood peak regulation operation efficiency, operation cost and model constraint function of the regulating reservoir are considered, specifically: based on the water inflow regulation, emptying and flushing operation modes of the regulating reservoir, the water level of the gate operation control node of each regulating reservoir is selected as the decision variable, and the primary goal is to minimize the risk of water accumulation within the regulation range of the regulating reservoir. On the premise of meeting the requirements of urban flood control, the minimization of the operating cost of the regulating reservoir is the secondary goal, and the goals are solved in sequence according to priority; the model constraint is: according to the sequential optimization strategy, after obtaining the optimal value of the primary goal, the optimal solution of the secondary goal is restricted by the primary goal.

[0014] Furthermore, the goals are solved in sequence according to the priority, specifically:

[0015] Combined with the distribution of flood-prone points within the regulation and storage area, the flood-prone area is divided according to the terrain grid units;

[0016] In the flood-prone area, the depth and duration of waterlogging are comprehensively considered, and the waterlogging risk is divided into four levels: IV, III, II, and I in the form of a piecewise function; the maximum risk level of waterlogging in each grid unit is calculated;

[0017]

[0018] Among them, V max B is the risk of water accumulation at the flood-prone point at the time of maximum flooding, Ⅳ , B Ⅲ , B Ⅱ , B Ⅰ are the number of grid cells of each risk level, k Ⅳ , k Ⅲ , k Ⅱ , k Ⅰ are the weight functions of their respective risk levels, and k is taken according to the regional waterlogging situation. Ⅳ , k Ⅲ, k Ⅱ , k Ⅰ are taken as 0.1, 0.2, 0.3 and 0.4 respectively, z is the unit grid cell length, h max,b is the maximum water depth of the b-th grid at the moment, t s The duration of water accumulation;

[0019] Integrating the risk of water accumulation at flood-prone points within the regulation and storage range, the optimal solution for the primary goal is:

[0020]

[0021] Among them, F1 is the primary target, I is the number of storage tanks; V max,i is the risk water volume at the maximum flooding moment within the storage range of the i-th storage pond;

[0022] After obtaining the optimal value of the primary objective, the optimal value of the primary objective is used as the constraint condition to solve the secondary objective:

[0023] F2=min{C} (2)

[0024] Among them, F2 is the secondary target, C is the operation cost function of the storage tank, and the operation cost function of the storage tank is shown in formula (3):

[0025]

[0026] Among them, the operation cost function C of the storage tank is the sum of the power consumption of the drainage pump and the flushing system; W m,t is the operating power consumption of the mth pump in the tth period; T m is the total operation time of the pump, which is mainly determined by the amount of water in the pool S i And pump flow Q p,m Calculated; P n is the power consumption of the nth flushing device; T n The total time the flushing equipment is running.

[0027] Furthermore, according to the sequential optimization strategy, after obtaining the optimal value of the primary objective, the optimal solution of the secondary objective is restricted by the primary objective. Specifically, the maximum water accumulation at the moment and the total water accumulation at each moment within the regulation range of the secondary objective are constrained by the primary objective.

[0028] V con ≤min{F1} (4)

[0029]

[0030] Among them, V con V is the risk water accumulation within the storage range of all storage tanks. This value is less than or equal to the maximum risk water accumulation at the time of flooding obtained by the first-level objective function;total,i V is the total amount of risk water accumulation at each moment in the entire rainfall drainage process within the storage range of the i-th storage tank; min is the minimum value of surface water volume within all storage ranges, and δ is the relaxation variable.

[0031] The constraints also include the infrastructure and operation rules of the reservoir itself:

[0032] S i ≤S allow,i (6)

[0033]

[0034]

[0035] Among them, the water volume of the regulating reservoir S i Need to be less than the maximum allowable volume S of the storage tank allow，i ;H on,j and H off,j are the gate opening and closing liquid level thresholds at the jth control point, and the value is the variable ΔH in the allowable range of each control point (H j,min , H j,max ) to change; considering that the judgment condition for closing the regulating reservoir is based on the saturation of the regulating reservoir volume, the closing liquid level threshold H is added off,j , when solving the optimal solution for the secondary objective, when the water volume S i Less than the maximum allowable volume S allow，i The closure of the storage tank is subject to the closure level threshold H off,j Control, when the water volume S i Equal to the maximum allowable volume S allow，i The storage tank is automatically closed because the maximum volume constraint is met. At this time, the closing liquid level threshold H off,j Take the lower limit of liquid level H j,min .

[0036] Furthermore, a particle swarm algorithm based on the mutant whale hunting operator is constructed, specifically:

[0037] The well-known particle swarm algorithm speed and position replacement formula is as follows: In the tth iteration, the speed and position of the i-th particle in the d-dimensional search space are updated:

[0038]

[0039]

[0040] Among them, ω is the inertia weight; c 1 and c 2 are the individual learning and social learning factors of particles; r 1 and r2 A random number between 0 and 1; pbest i d is the individual optimal solution of the i-th particle; gbest d is the global optimal solution for all particles in the d-th dimension space; x and V are the position and velocity of the particle respectively;

[0041] Construct a mutant whale hunting operator. After finding prey, the whale will surround the prey. The expression when searching and hunting is as follows:

[0042] X(t+1)=X rand (t)-M|C·X rand (t)-X(t)| (11)

[0043] X(t+1)=X best (t)-M|C·X best (t)-X(t)| (12)

[0044] Among them, X rand (t) is the position of a randomly selected individual whale in the population at time t; X bset (t) is the position of the best individual in the population at time t; X(t) is the position of the current whale individual at time t; C is the swing factor between [0,2]; M is the variant whale hunting operator, expressed as:

[0045] M=m(2r-1) (13)

[0046]

[0047] Where r is a random number in (0,1); T max is the maximum number of iterations; m is the mutation control factor, and P is the mutation probability;

[0048] When hunting prey, whales use a bubble net hunting method, including: encirclement and contraction and spiral bubble hunting. The mathematical model formula for this stage is as follows:

[0049] X(t+1)=|X best (t)-X(t)|·e lb ·cos(2πl)+X best (t) (15)

[0050] Among them, X bset (t) is the position of the optimal individual in the population at time t; b is the spiral constant; l is a random number between -1 and 1.

[0051] The mutant whale hunting operator is coupled into the particle swarm algorithm speed update formula, and formula (9) is updated as follows:

[0052]

[0053] Furthermore, an adaptive machine learning model suitable for emergency dispatching of the regulating reservoir is constructed, specifically: rainfall characteristic parameters and regulating reservoir operation parameters are used as the model training input data set, and the objective function and constraint conditions of the rainwater regulating reservoir operation dispatching model are used as the model training target data set; the model training input data set is normalized;

[0054] Based on the K nearest neighbor model, back propagation neural network model and convolutional neural network model, the mapping relationship between the normalized model training input data set and the model training target data set is respectively obtained; the model training results are statistically analyzed and the optimal training model is selected.

[0055] Furthermore, the rainfall characteristic parameters and the operation parameters of the storage tank are used as the input data set for model training, and the objective function and constraint conditions of the rainwater storage tank operation scheduling model are used as the target data set for model training. Specifically, the correlation analysis between the rainfall characteristic parameters and the output is performed based on the Pearson term coefficient, and the calculation formula is as follows:

[0056]

[0057] cov(x,y)=E(xy)-E(x)E(y) (18)

[0058]

[0059] Among them, ρ xy is the Pearson correlation coefficient; cov(x,y) is the covariance of parameters x and y; σ x and σ y Represent the variance of x and y respectively; E(x), E(y) and E(xy) are the mathematical expectations of x, y and xy respectively;

[0060] The normalization process of the model training input data set is specifically as follows:

[0061] Before training the reservoir operation parameters and rainfall characteristic parameters, the training data needs to be normalized:

[0062]

[0063] Where: x *i is the normalized value of a certain type of impact factor; x i is the original value; x i min and x i max are the minimum and maximum values ​​of this type of influencing factors respectively.

[0064] Furthermore, the optimal training model is selected, specifically: according to the mean absolute percentage error MAPE and the coefficient of determination R 2 Determine the applicability of the K nearest neighbor model, the back propagation neural network model, and the convolutional neural network model; if the K nearest neighbor model, the back propagation neural network model, and the convolutional neural network model all meet the conditions, that is, the MAPE value is less than 5% and R 2 The value is greater than 0.9, indicating that the three models have good adaptability. The model with the smallest MAPE value and R 2 The model with the largest value is used for model integration and error correction:

[0065]

[0066] Among them, P C is the prediction result of the modified model, P 1 To select the prediction result of the first model, M 1 and R 1 The MAPE value and R of the first model are 2 Value, P 2 To select the prediction results of the second model, M 2 and R 2 The MAPE and R values ​​of the second model are 2 value; if two of the K nearest neighbor model, back propagation neural network model, and convolutional neural network model meet the conditions, and one does not, it means that there are two models with good adaptability. These two models are selected to use formula (21) for model integration and error correction; if only one model meets the conditions, the result of this model is directly used as the final result. If all three models do not meet the conditions, it means that the number of data sets is insufficient, and retraining is performed after supplementing the data.

[0067] A rapid response system for emergency dispatch of urban rainwater storage tanks, comprising:

[0068] The first building module establishes a rainwater storage tank operation scheduling model based on the water inflow, emptying and flushing operation modes of the storage tank, taking into account the waterlogging risk level in the storage range, the storage operation efficiency, the operation cost and the model constraint function of the storage tank;

[0069] A second building block, wherein the second building block builds a particle swarm algorithm based on a mutated whale hunting operator;

[0070] A third construction module, wherein the third construction module establishes a hydrological and hydrodynamic model of a coupled storage tank based on the case hydrological and meteorological data, high-precision terrain image data, land use type and infiltration data, and pipe network and storage tank data;

[0071] An acquisition module, wherein the acquisition module uses rainfall characteristic parameters and storage tank operation parameters as model training input data sets, uses objective functions and constraints of the rainwater storage tank operation scheduling model as model training target data sets, and acquires a machine learning model suitable for emergency scheduling of urban rainwater storage tanks and with the highest accuracy based on the model training input data sets, the model training target data sets, and the adaptive machine learning model;

[0072] An output module replaces the hydrological and hydrodynamic model with an adaptive machine learning model as a solver of the particle swarm algorithm of the mutant whale hunting operator, sets the population size of the particle swarm algorithm of the mutant whale hunting operator and the maximum running algebraic parameter setting, and outputs the optimal objective function value and the optimal scheduling plan for the urban rainwater storage tank.

[0073] Compared with the prior art, the present invention has the following beneficial effects:

[0074] The present invention uses a hydrological and hydrodynamic model to accurately simulate and calculate the rainfall-induced waterlogging and storage tank scheduling process, and then based on the simulation calculation results, uses an adaptive machine learning model to directly establish the relationship between rainfall parameters, storage tank operation rules and scheduling targets, and uses it to replace the time-consuming numerical model as an optimization solver to solve in the optimization algorithm, avoiding repeated calls to the time-consuming numerical model during the optimization process, thereby improving the solution efficiency. Under the premise of ensuring the calculation accuracy, the present invention greatly shortens the emergency dispatch response time, gains a lot of lead time for the scheduling decision of the rainwater storage tank, and has important reference value for the formulation of urban emergency response strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0076] Figure 1 It is a schematic flow chart of a rapid response method for emergency dispatch of an urban rainwater storage tank of the present invention;

[0077] Figure 2 It is a structural schematic diagram of a rapid response system for emergency dispatch of urban rainwater storage tanks of the present invention;

[0078] Figure 3 It is the invention case of the rapid response method for emergency dispatch of urban rainwater storage tanks of the present invention, and the distribution map of the drainage pipe network and storage tanks;

[0079] Figure 4 It is a schematic diagram of an algorithm flow of a rapid response method for emergency dispatch of an urban rainwater storage tank according to the present invention;

[0080] Figure 5 It is a schematic diagram of a convergence curve of a regulation reservoir dispatching rule of a rapid response method for emergency dispatching of an urban rainwater regulation reservoir of the present invention;

[0081] Figure 6 It is a comparison chart of the invention case results of the rapid response method for emergency dispatch of urban rainwater storage tanks of the present invention. DETAILED DESCRIPTION

[0082] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0083] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0084] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0085] In the description of the embodiments of the present invention, it should be noted that if the terms "upper", "lower", "horizontal", "inner", etc. indicate an orientation or positional relationship based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the product of the invention is usually placed when in use, it is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0086] In addition, if the term "horizontal" appears, it does not mean that the component must be absolutely horizontal, but can be slightly tilted. For example, "horizontal" only means that its direction is more horizontal than "vertical", which does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0087] In the description of the embodiments of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal connection of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0088] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0089] See also Figure 1 The present invention discloses a rapid response method for emergency dispatch of urban rainwater storage tanks, comprising:

[0090] S101, based on the water inflow, storage, emptying and flushing operation modes of the regulating reservoir, comprehensively considering the waterlogging risk level in the regulating range, the regulating operation efficiency of the regulating reservoir, the operation cost and the model constraint function, establish a rainwater regulating reservoir operation scheduling model.

[0091] Based on the water inflow regulation, emptying and flushing operation modes of the regulating reservoir, the water level of the gate operation control node of each regulating reservoir is selected as the decision variable, and the primary goal is to minimize the risk of water accumulation within the regulating range of the regulating reservoir. Under the premise of meeting the requirements of waterlogging prevention and control, the minimization of the operating cost of the regulating reservoir is taken as the secondary goal, and the goals are solved in sequence according to the priority; the constraint condition of the model is: according to the sequential optimization strategy, after obtaining the optimal value of the primary goal, the optimal solution of the secondary goal is restricted by the primary goal.

[0092] Combined with the distribution of flood-prone points within the regulation and storage area, the flood-prone area is divided according to the terrain grid units;

[0093] Taking into account the depth and duration of waterlogging, the waterlogging risk is divided into four levels: IV, III, II, and I in the form of a piecewise function; the maximum risk level of waterlogging in each grid unit is calculated;

[0094]

[0095] Among them, V max B is the risk of water accumulation at the flood-prone point at the time of maximum flooding, Ⅳ , B Ⅲ , B Ⅱ , B Ⅰ are the number of grid cells of each risk level, k Ⅳ , k Ⅲ , k Ⅱ , k Ⅰ are the weight functions of their respective risk levels, and k is taken according to the regional waterlogging situation.Ⅳ , k Ⅲ , k Ⅱ , k Ⅰ are taken as 0.1, 0.2, 0.3 and 0.4 respectively, z is the unit grid cell length, h max,b is the maximum water depth of the bth grid at the moment, t s is the duration of water accumulation; integrating the risk water accumulation at flood-prone points within the storage range, the optimal solution for the primary goal is:

[0096]

[0097] Among them, F1 is the primary target, I is the number of storage tanks; V max,i is the risk water volume at the maximum flooding moment within the storage range of the i-th storage pond;

[0098] After obtaining the optimal value of the primary objective, the optimal value of the primary objective is used as the constraint condition to solve the secondary objective:

[0099] F2=min{C} (2)

[0100] Among them, F2 is the secondary target, C is the operation cost function of the storage tank, and the operation cost function of the storage tank is shown in formula (3):

[0101]

[0102] Among them, the operation cost function C of the storage tank is the sum of the power consumption of the drainage pump and the flushing system; W m,t is the operating power consumption of the mth pump in the tth period; T m is the total operation time of the pump, which is mainly determined by the amount of water in the pool S i And pump flow Q p,m Calculated; P n is the power consumption of the nth flushing device; T n The total time the flushing equipment is running.

[0103] According to the sequential optimization strategy, after obtaining the optimal value of the primary goal, the optimal solution of the secondary goal is restricted by the primary goal. Specifically, the maximum water accumulation at any time and the total water accumulation at each time within the regulation range of the secondary goal are constrained by the primary goal.

[0104] V con ≤min{F1} (4)

[0105]

[0106] Among them, V con V is the risk water accumulation within the storage range of all storage tanks. This value is less than or equal to the maximum risk water accumulation at the time of flooding obtained by the first-level objective function; total,iV is the total amount of risk water accumulation at each moment in the entire rainfall drainage process within the storage range of the i-th storage tank; min is the minimum value of surface water volume within all storage areas, and the slack variable δ is a slack variable. The constraints also include the infrastructure and operation rules of the storage pond itself:

[0107] S i ≤S allow,i (6)

[0108]

[0109]

[0110] Among them, the water volume of the regulating reservoir S i Need to be less than the maximum allowable volume S of the storage tank allow，i ;H on,j and H off,j are the gate opening and closing liquid level thresholds at the jth control point, and the value is the variable ΔH in the allowable range of each control point (H j,min , H j,max ) to change; considering that the judgment condition for closing the regulating reservoir is based on the saturation of the regulating reservoir volume, the closing liquid level threshold H is added off,j , when solving the optimal solution for the secondary objective, when the water volume S i Less than the maximum allowable volume S allow，i The closure of the storage tank is subject to the closure level threshold H off,j Control, when the water volume S i Equal to the maximum allowable volume S allow，i The storage tank is automatically closed because the maximum volume constraint is met. At this time, the closing liquid level threshold H off,j Take the lower limit of liquid level H j,min .

[0111] S102, construct a particle swarm algorithm based on the mutant whale hunting operator.

[0112] The well-known particle swarm algorithm speed and position replacement formula is as follows: In the tth iteration, the speed and position of the i-th particle in the d-dimensional search space are updated:

[0113]

[0114]

[0115] Among them, ω is the inertia weight; c 1 and c 2 are the individual learning and social learning factors of particles; r 1 and r 2 A random number between 0 and 1; pbest id is the individual optimal solution of the i-th particle; gbest d is the global optimal solution for all particles in the d-th dimension space; x and V are the position and velocity of the particle respectively;

[0116] Construct a mutant whale hunting operator. After finding prey, the whale will surround the prey. The expression when searching and hunting is as follows:

[0117] X(t+1)=X rand (t)-M|C·X rand (t)-X(t)| (11)

[0118] X(t+1)=X best (t)-M|C·X best (t)-X(t)| (12)

[0119] Among them, X rand (t) is the position of a randomly selected individual whale in the population at time t; X bset (t) is the position of the best individual in the population at time t; X(t) is the position of the current whale individual at time t; C is the swing factor between [0,2]; M is the variant whale hunting operator, expressed as:

[0120] M=m(2r-1) (13)

[0121]

[0122] Where r is a random number in (0,1); T max is the maximum number of iterations; m is the mutation control factor, and P is the mutation probability;

[0123] When hunting prey, whales use a bubble net hunting method, including: encirclement and contraction and spiral bubble hunting. The mathematical model formula for this stage is as follows:

[0124] X(t+1)=|X best (t)-X(t)|·e lb ·cos(2πl)+X best (t) (15)

[0125] Among them, X bset (t) is the position of the optimal individual in the population at time t; b is the spiral constant; l is a random number between -1 and 1.

[0126] The mutant whale hunting operator is coupled into the particle swarm algorithm speed update formula, and formula (9) is updated as follows:

[0127]

[0128] S103, based on the case study hydrological and meteorological data, high-precision terrain image data, land use type and infiltration data, as well as pipe network and storage tank data, establish a hydrological and hydrodynamic model of the coupled storage tank.

[0129] S104, using rainfall characteristic parameters and storage tank operation parameters as model training input data sets, using the objective function and constraints of the rainwater storage tank operation scheduling model as model training target data sets, based on the model training input data sets, model training target data sets and adaptive machine learning model, obtain a machine learning model suitable for emergency scheduling of urban rainwater storage tanks and with the highest accuracy.

[0130] The rainfall characteristic parameters and storage tank operation parameters are used as the model training input data set, and the objective function and constraint conditions of the rainwater storage tank operation scheduling model are used as the model training target data set; the model training input data set is normalized; the mapping relationship between the normalized model training input data set and the model training target data set is respectively calculated based on the K nearest neighbor model, the back propagation neural network model and the convolutional neural network model; the model training results are statistically analyzed and the optimal training model is selected.

[0131] The rainfall characteristic parameters and storage tank operation parameters are used as the model training input data set, and the objective function and constraint conditions of the rainwater storage tank operation scheduling model are used as the model training target data set. Specifically, the correlation analysis between the rainfall characteristic parameters and the output is performed based on the Pearson term coefficient. The calculation formula is as follows:

[0132]

[0133] cov(x,y)=E(xy)-E(x)E(y) (18)

[0134]

[0135] Among them, ρ xy is the Pearson correlation coefficient; cov(x,y) is the covariance of parameters x and y; σ x and σ y Represent the variance of x and y respectively; E(x), E(y) and E(xy) are the mathematical expectations of x, y and xy respectively;

[0136] The normalization process of the model training input data set is specifically as follows:

[0137] Before training the reservoir operation parameters and rainfall characteristic parameters, the training data needs to be normalized:

[0138]

[0139] Where: x *iis the normalized value of a certain type of impact factor; x i is the original value; x i min and x i max are the minimum and maximum values ​​of this type of influencing factors respectively.

[0140] Select the most optimized training model, specifically: according to the mean absolute percentage error MAPE and the coefficient of determination R 2 Determine the applicability of the K nearest neighbor model, the back propagation neural network model, and the convolutional neural network model; if the K nearest neighbor model, the back propagation neural network model, and the convolutional neural network model all meet the conditions, that is, the MAPE value is less than 5% and R 2 The value is greater than 0.9, indicating that the three models have good adaptability. The model with the smallest MAPE value and R 2 The model with the largest value is used for model integration and error correction:

[0141]

[0142] Among them, P C is the prediction result of the modified model, P 1 To select the prediction result of the first model, M 1 and R 1 The MAPE value and R of the first model are 2 Value, P 2 To select the prediction results of the second model, M 2 and R 2 The MAPE and R values ​​of the second model are 2 value; if two of the K nearest neighbor model, back propagation neural network model, and convolutional neural network model meet the conditions, and one does not, it means that there are two models with good adaptability. These two models are selected to use formula (21) for model integration and error correction; if only one model meets the conditions, the result of this model is directly used as the final result. If all three models do not meet the conditions, it means that the number of data sets is insufficient, and retraining is performed after supplementing the data.

[0143] S105, replace the hydrological and hydrodynamic model with an adaptive machine learning model as the solver of the particle swarm algorithm of the mutant whale hunting operator, set the population size of the particle swarm algorithm of the mutant whale hunting operator and the maximum running algebraic parameter setting, and output the optimal objective function value and the optimal scheduling plan for the urban rainwater storage tank.

[0144] See also Figure 2 The present invention discloses a rapid response system for emergency dispatch of urban rainwater storage tanks, comprising:

[0145] The first building module establishes a rainwater storage tank operation scheduling model based on the water inflow, emptying and flushing operation modes of the storage tank, taking into account the waterlogging risk level in the storage range, the storage operation efficiency, the operation cost and the model constraint function of the storage tank;

[0146] A second building block, wherein the second building block builds a particle swarm algorithm based on a mutated whale hunting operator;

[0147] A third construction module, wherein the third construction module establishes a hydrological and hydrodynamic model of a coupled storage tank based on the case hydrological and meteorological data, high-precision terrain image data, land use type and infiltration data, and pipe network and storage tank data;

[0148] An acquisition module, wherein the acquisition module uses rainfall characteristic parameters and storage tank operation parameters as model training input data sets, uses objective functions and constraints of the rainwater storage tank operation scheduling model as model training target data sets, and acquires a machine learning model suitable for emergency scheduling of urban rainwater storage tanks and with the highest accuracy based on the model training input data sets, the model training target data sets, and the adaptive machine learning model;

[0149] An output module replaces the hydrological and hydrodynamic model with an adaptive machine learning model as a solver of the particle swarm algorithm of the mutant whale hunting operator, sets the population size of the particle swarm algorithm of the mutant whale hunting operator and the maximum running algebraic parameter setting, and outputs the optimal objective function value and the optimal scheduling plan for the urban rainwater storage tank.

[0150] Example

[0151] The present invention discloses a rapid response method for emergency dispatch of urban rainwater storage tanks, comprising:

[0152] Step 1: Based on the water inflow, storage, emptying and flushing operation modes of the regulating reservoir, a rainwater regulating reservoir operation scheduling model is established by comprehensively considering the waterlogging risk level in the regulating range, the regulating operation efficiency of the regulating reservoir, the operation cost and the model constraint function.

[0153] Step 2, construct a particle swarm algorithm based on a mutated whale hunting operator, which is used to automatically find the optimal emergency dispatching solution for the storage tank. In order to make up for the balance between the initial global search capability and the later local search capability of the particle swarm algorithm, the present invention designs a mutated whale hunting operator to guide particle update.

[0154] Step 3: Use the case study hydrological and meteorological data, high-precision terrain image data, land use type and infiltration data, and pipe network and storage tank data to establish a hydrological and hydrodynamic model of the coupled storage tank. The pipe network storage tank data required for the model establishment in this case are as follows: Figure 3 shown.

[0155] Step 4: Establish an adaptive machine learning model suitable for emergency dispatch of the reservoir. Take the rainfall characteristic parameters and reservoir operation parameters as the model training input data set, and take the calculation parameters reflecting the objective function and constraint conditions of the emergency dispatch of the reservoir in step 1 as the model training target data set. Establish a mapping relationship between the two, and use an adaptive method to compare and select to find the most suitable machine learning model for emergency dispatch of the reservoir. The calculation parameters used for the model training target data set are calculated and statistically obtained by the hydrological and hydrodynamic model in step 3.

[0156] Step 5: Replace the hydrological and hydrodynamic model described in step 3 with the adaptive machine learning model described in step 4 as the solver of the optimization algorithm in step 2 to drive the algorithm evolution process. This avoids calling the time-consuming hydrological and hydrodynamic model, thereby greatly improving the optimization efficiency. Figure 4 shown.

[0157] Step 6, set the algorithm population size, maximum running algebra and other parameter settings, run the algorithm program, complete the search when the running algebra reaches the set maximum value, and output the optimal objective function value and the optimal scheduling plan for the urban rainwater storage tank.

[0158] The specific construction steps of the rainwater storage tank emergency dispatch model in step 1 are as follows:

[0159] Step 1.1, construct the objective function of the emergency dispatch model of the regulating reservoir. The opening of the operating gate of the regulating reservoir is determined by the water level of the control point. In addition to considering the automatic closing of the gate when the regulating reservoir is full, the closing of the gate should also consider that when the rainwater flow has passed the peak but the regulating reservoir is not full, the control point will not overflow at this time, and the gate can also close automatically according to the water level of the control point. The main goal of the emergency dispatch of the regulating reservoir is to minimize the flood risk within its storage range, and in the actual operation process, its operating cost must also be considered, and the water storage capacity should be reduced as much as possible to reduce the operating cost. Taking both factors into consideration, the water level of the gate operation control node of each regulating reservoir is used as the decision variable, and the primary goal is to minimize the risk of water accumulation within the regulating reservoir. On the premise of minimizing the water accumulation, the secondary goal is to minimize the operating cost of the regulating reservoir. By introducing a sequential optimization strategy, the multi-objective problem is transformed into a single-objective problem, and the objectives are solved in sequence according to the priority.

[0160] The goals are solved in sequence according to the priority, specifically:

[0161] Combined with the distribution of flood-prone points within the regulation and storage area, the flood-prone area is divided according to the terrain grid units;

[0162] Taking into account the depth and duration of waterlogging, the waterlogging risk is divided into four levels: IV, III, II, and I in the form of a piecewise function; the maximum risk level of waterlogging in each grid unit is calculated;

[0163]

[0164] Among them, V max B is the risk of water accumulation at the flood-prone point at the time of maximum flooding, Ⅳ , B Ⅲ , B Ⅱ , B Ⅰ are the number of grid cells of each risk level, k Ⅳ , k Ⅲ , k Ⅱ , k Ⅰ are the weight functions of their respective risk levels, and k is taken according to the regional waterlogging situation. Ⅳ , k Ⅲ , k Ⅱ , k Ⅰ are taken as 0.1, 0.2, 0.3 and 0.4 respectively, z is the unit grid cell length, h max,b is the maximum water depth of the b-th grid at the moment, t s is the duration of water accumulation; integrating the risk water accumulation at flood-prone points within the storage range, the optimal solution for the primary goal is:

[0165]

[0166] Among them, F1 is the primary target, I is the number of storage tanks; V max,i is the risk water volume at the maximum flooding time within the storage range of the i-th storage tank; after obtaining the optimal value of the primary goal, the optimal value of the primary goal is used as the constraint condition to solve the secondary goal:

[0167] F2=min{C} (2)

[0168] Among them, F2 is the secondary target, C is the operation cost function of the storage tank, and the operation cost function of the storage tank is shown as follows:

[0169]

[0170] Among them, the operation cost function C of the storage tank is the sum of the power consumption of the drainage pump and the flushing system; W m,t is the operating power consumption of the mth pump in the tth period; T m is the total operation time of the pump, which is mainly determined by the amount of water in the pool S i And pump flow Q p,m Calculated; P n is the power consumption of the nth flushing device; T n The total operation time of the flushing equipment;

[0171] Step 1.2, construct the constraint function of the emergency dispatch model of the storage reservoir. According to the sequential optimization strategy, after solving the minimum value of F1, the operating cost of F2 is restricted by F1, including the constraints of the maximum water volume at the storage range and the total water volume at each time:

[0172] V con ≤min{F1} (4)

[0173]

[0174] Among them, V con V is the risk water accumulation within the storage range of all storage tanks. This value is less than or equal to the maximum risk water accumulation at the time of flooding obtained by the first-level objective function; total,i V is the total amount of risk water accumulation at each moment in the entire rainfall drainage process within the storage range of the i-th storage tank. Since the storage tank is closed in advance after the rain peak, the flow of some drainage pipes will increase again, which may cause overflow. Therefore, the total amount of surface water accumulation within the storage range during the entire rainfall drainage process is used as a constraint; V min is the minimum value of surface water volume within all storage areas. Considering that the machine learning model has certain errors, the predicted V min The constraints are met but violated when the flood numerical model is used for recalculation. Therefore, the slack variable δ is subtracted from this term.

[0175] In addition, the constraints also include the infrastructure and operation rules of the reservoir itself:

[0176] S i ≤S allow,i (6)

[0177]

[0178]

[0179] Among them, the water volume of the regulating reservoir S i Need to be less than the maximum allowable volume S of the storage tank allow，i ;H on,j and H off,j are the gate opening and closing liquid level thresholds at the jth control point, and the value is the variable ΔH in the allowable range of each control point (H j,min , H j,max Considering that the judgment condition for closing the regulating reservoir is based on the saturation of the regulating reservoir volume, the closing liquid level threshold H is added. off,j When solving F2, when the water volume S i Less than the maximum allowable volume S allow，i The closure of the regulating reservoir is subject to the closing liquid level threshold Hoff,j Control, when the water volume S i Equal to the maximum allowable volume S allow，i The storage tank is automatically closed because the maximum volume constraint is met. At this time, the closing liquid level threshold H off,j Take the lower limit of liquid level H j,min .

[0180] In step 2, the specific construction steps of the particle swarm algorithm based on the mutant whale hunting operator are as follows:

[0181] Step 2.1, Particle Swarm Algorithm Construction. The known particle swarm algorithm speed and position are replaced with new formulas. In the tth iteration, the speed and position of the i-th particle in the d-dimensional search space are updated:

[0182]

[0183]

[0184] Among them, ω is the inertia weight; c 1 and c 2 are the individual learning and social learning factors of particles; r 1 and r 2 A random number between 0 and 1; pbest i d is the individual optimal solution of the i-th particle; gbest d is the global optimal solution for all particles in the d-th dimension space; x and V are the position and velocity of the particle respectively;

[0185] Step 2.2, construct a mutant whale hunting operator to ensure that the particle swarm algorithm has better global search capabilities in the early stage and local search capabilities in the later stage, and can jump out of the local optimal trap by mutation.

[0186] The current individual whale randomly selects another individual whale as a target and moves closer to its position, forcing the whale to deviate from its original position to find a more suitable prey, thereby increasing the randomness of the algorithm for global search, hoping to find a better prey. After finding the prey, the whale will surround the prey. The expression for searching and encircling is as follows:

[0187] X(t+1)=X rand (t)-M|C·X rand (t)-X(t)| (11)

[0188] X(t+1)=X best (t)-M|C·X best (t)-X(t)| (12)

[0189] Among them, X rand(t) is the position of a randomly selected individual whale in the population at time t; X bset (t) is the position of the best individual in the population at time t; X(t) is the position of the current whale individual at time t; C is the swing factor between [0,2]; M is the variant whale roundup operator constructed by the present invention, which is calculated by referring to the following formula:

[0190] M=m(2r-1) (13)

[0191]

[0192] Where r is a random number in (0,1); T max is the maximum number of iterations; m is the mutation control factor, and P is the mutation probability. In the early stage of algorithm iteration, the m value is large and decays slowly, and the algorithm has better global search ability; in the late stage of algorithm iteration, the m value decays rapidly, and the algorithm has better local search ability. At the same time, due to the existence of mutation, the m value will mutate in the late stage of the algorithm to prevent the algorithm from falling into the local optimal trap.

[0193] When whales hunt their prey, they use a bubble net hunting method, which mainly consists of two mechanisms, namely, encirclement and contraction and spiral bubble hunting. The mathematical model formula for this stage is as follows:

[0194] X(t+1)=|X best (t)-X(t)|·e lb ·cos(2πl)+X best (t) (15)

[0195] Among them, X bset (t) is the position of the optimal individual in the population at time t; b is the spiral constant; l is a random number between -1 and 1.

[0196] Step 2.3, the mutant whale hunting operator is coupled into the particle swarm algorithm speed update formula, and the original known formula (9) is updated to:

[0197]

[0198] The particle swarm algorithm coupled with whale hunting has greatly solved the problem that the known particle swarm algorithm has poor balance between global search and local search and is prone to falling into local optimality.

[0199] In step 4, the specific steps for building an adaptive machine learning model suitable for emergency dispatch of the storage reservoir are as follows:

[0200] Step 4.1, urban rainwater storage tank dataset construction. The training dataset required for the machine learning model is divided into an input set and an output set. The input set includes rainfall characteristic parameters and storage tank operation parameters, and the output set includes parameters that can reflect the objective function and constraints described in step 1, such as the surface water volume, total water volume and storage tank water volume at the maximum water accumulation time within the storage range, which are calculated by the hydrological and hydrodynamic model described in step 3.

[0201] The input is concentrated, and there are many rainfall characteristic parameters, including rainfall recurrence period, rainfall peak, maximum one-hour rainfall, average rainfall, rain peak arrival time, etc. When training the model, it is necessary to select those that have a greater impact on the result prediction. The Pearson coefficient is used to analyze the correlation between rainfall characteristic parameters and output. The calculation formula is as follows:

[0202]

[0203] cov(x,y)=E(xy)-E(x)E(y) (18)

[0204]

[0205] Among them, ρ xy is the Pearson correlation coefficient; cov(x,y) is the covariance of parameters x and y; σ x and σ y Represent the variance of x and y respectively; E(x), E(y) and E(xy) are the mathematical expectations of x, y and xy respectively.

[0206] The operation parameters of the reservoir are the control variables that control the opening and closing of the reservoir gates. The control variables are the decision variables in step 1.1, and the water levels of the operation control nodes of each reservoir gate. The operation parameters of the reservoir and the rainfall characteristic parameters together constitute the training data set of the machine learning model. In order to avoid the differences in the numerical ranges of various types of data, the training data needs to be normalized before training:

[0207]

[0208] Where: x *i is the normalized value of a certain type of impact factor; x i is the original value; x i min and x i max are the minimum and maximum values ​​of this type of influencing factors respectively.

[0209] Step 4.2, machine learning model training and adaptive comparison strategy. There are many commonly used machine learning models, but there is no specification or criterion to indicate that a certain machine learning model is suitable for emergency dispatch of urban rainwater storage tanks. Different models have different sensitivities to the data set in step 4.1, and thus the resulting model fitting accuracy is also different. The adaptive selection strategy of the present invention can select the machine learning model that is suitable for emergency dispatch of urban rainwater storage tanks and has the highest accuracy.

[0210] The mapping relationship between input data set and output data set is established by using three different types of models, namely K-nearest neighbor model (KNN), back propagation neural network model (BP-ANN), and convolutional neural network model (CNN). The model training results are statistically analyzed with MAPE value and R 2 The values ​​are used as the criteria for model prediction accuracy and fit, respectively.

[0211] If all three models meet the conditions, that is, the MAPE value is less than 5% and R 2 The value is greater than 0.9, indicating that the three models have good adaptability. The model with the smallest MAPE value and R 2 The model with the largest value is used for model integration and error correction:

[0212]

[0213] Among them, P C is the prediction result of the modified model, P 1 To select the prediction result of the first model, M 1 and R 1 The MAPE value and R of the first model are 2 Value, P 2 To select the prediction results of the second model, M 2 and R 2 The MAPE and R values ​​of the second model are 2 value.

[0214] If two models meet the conditions and one does not, it means that there are two models with better adaptability. These two models are selected and formula 20 is used for model integration and error correction.

[0215] If only one model meets the conditions, the result of this model will be directly used as the final result.

[0216] If all three models do not meet the requirements, it means that the data set is insufficient and the training should be repeated after supplementing the data.

[0217] The present invention has shown good results in the actual application of urban storage tank operation. Figure 5The convergence curve of the dispatching rules obtained by the present invention is shown. The dispatching rules obtained have achieved beneficial effects in both surface waterlogging and operating costs. The present invention can achieve good results under heavy rains with different return periods, such as Figure 6 As shown. The optimized dispatching scheme obtained in the case of the present invention is obtained under the calculation of the hydrological and hydrodynamic numerical model, thus ensuring the calculation accuracy. At the same time, through the method proposed by the present invention, the original optimization calculation time of hundreds of hours is shortened to less than 4 hours, achieving a rapid response to emergency dispatch problems and providing emergency dispatch management personnel with sufficient emergency response lead time.

[0218] A terminal device is provided in one embodiment of the present invention. The terminal device of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps in the above-mentioned method embodiments are implemented. Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are implemented.

[0219] The computer program may be divided into one or more modules / units, and the one or more modules / units are stored in the memory and executed by the processor to accomplish the present invention.

[0220] The terminal device may be a computing device such as a desktop computer, a notebook, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0221] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0222] The memory may be used to store the computer programs and / or modules, and the processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory.

[0223] If the module / unit integrated in the terminal device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0224] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A rapid response method for emergency dispatch of urban rainwater storage tanks, characterized in that: include: Based on the operation modes of water inflow, emptying and flushing of the regulating reservoir, the operation scheduling model of the rainwater regulating reservoir is established by comprehensively considering the waterlogging risk level in the regulating range, the regulating operation efficiency of the regulating reservoir, the operation cost and the model constraint function. Construct a particle swarm algorithm based on the mutant whale hunting operator; Based on the case study hydrological and meteorological data, high-precision terrain image data, land use type and infiltration data, as well as pipe network and storage tank data, a hydrological and hydrodynamic model of the coupled storage tank is established; The rainfall characteristic parameters and storage tank operation parameters are used as the model training input data set, and the objective function and constraint conditions of the rainwater storage tank operation scheduling model are used as the model training target data set. Based on the model training input data set, model training target data set and adaptive machine learning model, a machine learning model suitable for emergency scheduling of urban rainwater storage tanks with the highest accuracy is obtained. The adaptive machine learning model is used to replace the hydrological and hydrodynamic model as the solver of the particle swarm algorithm of the mutant whale hunting operator, and the population size and maximum running algebraic parameter setting of the particle swarm algorithm of the mutant whale hunting operator are set to output the optimal objective function value and the optimal scheduling plan of the urban rainwater storage tank. The operation parameters based on the regulating reservoir are considered to consider the peak flood regulation operation efficiency, operation cost and model constraint condition function of the regulating reservoir, specifically: based on the water inflow regulation, emptying and flushing operation modes of the regulating reservoir, the water level of each regulating reservoir gate operation control node is selected as the decision variable, and the primary goal is to minimize the risk of water accumulation within the regulating range of the regulating reservoir. On the premise of meeting the requirements of waterlogging prevention and control, the minimization of the operating cost of the regulating reservoir is taken as the secondary goal, and the goals are solved in sequence according to the priority; the model constraint condition is: according to the sequential optimization strategy, after the optimal value of the primary goal is obtained, the optimal solution of the secondary goal is restricted by the primary goal; The objectives are solved in sequence according to the priority, specifically: Combined with the distribution of flood-prone points within the regulation and storage area, the flood-prone area is divided according to the terrain grid units; In the flood-prone area, the depth and duration of waterlogging are comprehensively considered, and the waterlogging risk is divided into four levels: IV, III, II, and I in the form of a piecewise function; Calculate the maximum risk level of water accumulation in each grid cell; Among them, V max B is the risk of water accumulation at the flood-prone point at the time of maximum flooding, Ⅳ , B Ⅲ , B Ⅱ , B Ⅰ are the number of grid cells of each risk level, k Ⅳ , k Ⅲ , k Ⅱ , k Ⅰ are the weight functions of their respective risk levels, and k is taken according to the regional waterlogging situation. Ⅳ , k Ⅲ , k Ⅱ , k Ⅰ are taken as 0.1, 0.2, 0.3 and 0.4 respectively, z is the unit grid cell length, h max,b is the maximum water depth of the bth grid at the moment, t s The duration of water accumulation; Integrating the risk of water accumulation at flood-prone points within the regulation and storage range, the optimal solution for the primary goal is: Among them, F1 is the primary target, I is the number of storage tanks; V max,i is the risk water volume at the maximum flooding moment within the storage range of the i-th storage pond; After obtaining the optimal value of the primary objective, the optimal value of the primary objective is used as the constraint condition to solve the secondary objective: F2=min{C} (2) Among them, F2 is the secondary target, C is the operation cost function of the storage tank, and the operation cost function of the storage tank is shown in formula (3): Among them, the operation cost function C of the storage tank is the sum of the power consumption of the drainage pump and the flushing system; W m,t is the operating power consumption of the mth pump in the tth period; T m is the total operation time of the pump, which is determined by the amount of water in the pool S i And pump flow Q p,m Calculated; P n is the power consumption of the nth flushing device; T n The total time the flushing equipment is running.

2. The rapid response method for emergency dispatch of urban rainwater storage tanks according to claim 1 is characterized in that: After obtaining the optimal value of the primary objective according to the sequential optimization strategy, the optimal solution of the secondary objective is restricted by the primary objective, specifically: the maximum water accumulation at any time and the total water accumulation at each time within the regulation range of the secondary objective are restricted by the primary objective; V con ≤min{F1} (4) Among them, V con V is the risk water accumulation within the storage range of all storage tanks. This value is less than or equal to the maximum risk water accumulation at the time of flooding obtained by the first-level objective function; total,i V is the total amount of risk water accumulation at each moment in the entire rainfall drainage process within the storage range of the i-th storage tank; min is the minimum value of surface water volume within all storage ranges, and δ is the relaxation variable; The constraints also include the infrastructure and operation rules of the reservoir itself: S i ≤S allow,i (6) Among them, the water volume of the regulating reservoir S i Need to be less than the maximum allowable volume S of the storage tank allow，i ;H on,j and H off,j are the gate opening and closing liquid level thresholds at the jth control point, and the value is the variable ΔH in the allowable range of each control point (H j,min , H j,max ) to change; considering that the judgment condition for closing the regulating reservoir is based on the saturation of the regulating reservoir volume, the closing liquid level threshold H is added off,j , when solving the optimal solution for the secondary objective, when the water volume S i Less than the maximum allowable volume S allow，i The closure of the storage tank is subject to the closure level threshold H off,j Control, when the water volume S i Equal to the maximum allowable volume S allow，i The storage tank is automatically closed because the maximum volume constraint is met. At this time, the closing liquid level threshold H off,j Take the lower limit of liquid level H j,min .

3. The rapid response method for emergency dispatch of urban rainwater storage tanks according to claim 2 is characterized in that: The particle swarm algorithm based on the mutant whale hunting operator is constructed as follows: The well-known particle swarm algorithm speed and position replacement formula is as follows: In the tth iteration, the speed and position of the i-th particle in the d-dimensional search space are updated: Among them, ω is the inertia weight; c1 and c2 are the individual learning and social learning factors of the particles respectively; r1 and r2 are random numbers between 0 and 1; pbest i d is the individual optimal solution of the i-th particle; gbest d is the global optimal solution for all particles in the d-th dimension space; x and V are the position and velocity of the particle respectively; Construct a mutant whale hunting operator. After finding prey, the whale will surround the prey. The expression when searching and hunting is as follows: X(t+1)=X rand (t)-M|C·X rand (t)-X(t)| (11) X(t+1)=X best (t)-M|C·X best (t)-X(t)| (12) Among them, X rand (t) is the position of a randomly selected individual whale in the population at time t; X bset (t) is the position of the best individual in the population at time t; X(t) is the position of the current whale individual at time t; C is the swing factor between [0,2]; M is the mutant whale hunting operator, expressed as: M=m(2r-1) (13) Where r is a random number in (0,1); T max is the maximum number of iterations; m is the mutation control factor, and P is the mutation probability; When hunting prey, whales use a bubble net hunting method, including: encirclement and contraction and spiral bubble hunting. The mathematical model formula for this stage is as follows: X(t+1)=|X best (t)-X(t)|·e lb ·cos(2πl)+X best (t) (15) Among them, X bset (t) is the position of the best individual in the population at time t; b is the spiral constant; l is a random number between -1 and 1; The mutant whale hunting operator is coupled into the particle swarm algorithm speed update formula, and formula (9) is updated as follows:

4. The rapid response method for emergency dispatch of urban rainwater storage tanks according to claim 1 is characterized in that: The construction of an adaptive machine learning model suitable for emergency dispatching of a storage tank is specifically as follows: using rainfall characteristic parameters and storage tank operation parameters as model training input data sets, using objective functions and constraints of a rainwater storage tank operation dispatching model as model training target data sets; normalizing the model training input data sets; Based on the K nearest neighbor model, back propagation neural network model and convolutional neural network model, the mapping relationship between the normalized model training input data set and the model training target data set is respectively obtained; the model training results are statistically analyzed and the optimal training model is selected.

5. The rapid response method for emergency dispatch of urban rainwater storage tanks according to claim 4 is characterized in that: The rainfall characteristic parameters and the operation parameters of the storage tank are used as the model training input data set, and the objective function and constraint conditions of the rainwater storage tank operation scheduling model are used as the model training target data set. Specifically, the correlation analysis between the rainfall characteristic parameters and the output is performed based on the Pearson term coefficient, and the calculation formula is as follows: cov(x,y)=E(xy)-E(x)E(y) (18) Among them, ρ xy is the Pearson correlation coefficient; cov(x,y) is the covariance of parameters x and y; σ x and σ y Represent the variance of x and y respectively; E(x), E(y) and E(xy) are the mathematical expectations of x, y and xy respectively; The normalization process of the model training input data set is specifically as follows: Before training the reservoir operation parameters and rainfall characteristic parameters, the training data needs to be normalized: Where: x *i is the normalized value of a certain type of impact factor; x i is the original value; x i min and x i max are the minimum and maximum values ​​of this type of influencing factors respectively.

6. The rapid response method for emergency dispatch of urban rainwater storage tanks according to claim 5 is characterized in that: The optimal training model is selected as follows: according to the mean absolute percentage error MAPE and the determination coefficient R 2 Determine the applicability of the K nearest neighbor model, the back propagation neural network model, and the convolutional neural network model; if the K nearest neighbor model, the back propagation neural network model, and the convolutional neural network model all meet the conditions, that is, the MAPE value is less than 5% and R 2 The value is greater than 0.9, indicating that the three models have good adaptability. The model with the smallest MAPE value and R 2 The model with the largest value is used for model integration and error correction: Among them, P C is the prediction result of the modified model, P1 is the prediction result of the first model, M1 and R1 are the MAPE value and R of the first model respectively. 2 value, P2 is the prediction result of the second model, M2 and R2 are the MAPE value and R of the second model respectively. 2 value; if two of the K nearest neighbor model, back propagation neural network model and convolutional neural network model meet the conditions and one does not, it means that there are two models with good adaptability. These two models are selected to use formula (21) for model integration and error correction; if only one model meets the conditions, the result of this model is directly used as the final result; if all three models do not meet the conditions, it means that the number of data sets is insufficient and retraining is performed after supplementing the data.

7. A rapid response system for emergency dispatch of urban rainwater storage tanks, characterized in that: include: The first building module establishes a rainwater storage tank operation scheduling model based on the water inflow, emptying and flushing operation modes of the storage tank, taking into account the waterlogging risk level in the storage range, the storage operation efficiency, the operation cost and the model constraint function of the storage tank; A second building block, wherein the second building block builds a particle swarm algorithm based on a mutated whale hunting operator; A third construction module, wherein the third construction module establishes a hydrological and hydrodynamic model of a coupled storage tank based on the case hydrological and meteorological data, high-precision terrain image data, land use type and infiltration data, and pipe network and storage tank data; An acquisition module, wherein the acquisition module uses rainfall characteristic parameters and storage tank operation parameters as model training input data sets, uses objective functions and constraints of the rainwater storage tank operation scheduling model as model training target data sets, and acquires a machine learning model suitable for emergency scheduling of urban rainwater storage tanks and with the highest accuracy based on the model training input data sets, the model training target data sets, and the adaptive machine learning model; An output module, wherein the output module uses an adaptive machine learning model to replace a hydrological and hydrodynamic model as a solver of a particle swarm algorithm of a mutant whale hunting operator, sets a population size of the particle swarm algorithm of the mutant whale hunting operator, sets a maximum running algebraic parameter setting, and outputs an optimal objective function value and an optimal scheduling plan for an urban rainwater storage tank; The operation parameters based on the regulating reservoir are considered to consider the peak flood regulation operation efficiency, operation cost and model constraint condition function of the regulating reservoir, specifically: based on the water inflow regulation, emptying and flushing operation modes of the regulating reservoir, the water level of each regulating reservoir gate operation control node is selected as the decision variable, and the primary goal is to minimize the risk of water accumulation within the regulating range of the regulating reservoir. On the premise of meeting the requirements of waterlogging prevention and control, the minimization of the operating cost of the regulating reservoir is taken as the secondary goal, and the goals are solved in sequence according to the priority; the model constraint condition is: according to the sequential optimization strategy, after the optimal value of the primary goal is obtained, the optimal solution of the secondary goal is restricted by the primary goal; The objectives are solved in sequence according to the priority, specifically: Combined with the distribution of flood-prone points within the regulation and storage area, the flood-prone area is divided according to the terrain grid units; In the flood-prone area, the depth and duration of waterlogging are comprehensively considered, and the waterlogging risk is divided into four levels: IV, III, II, and I in the form of a piecewise function; Calculate the maximum risk level of water accumulation in each grid cell; Among them, V max B is the risk of water accumulation at the flood-prone point at the time of maximum flooding, Ⅳ , B Ⅲ , B Ⅱ , B Ⅰ are the number of grid cells of each risk level, k Ⅳ , k Ⅲ , k Ⅱ , k Ⅰ are the weight functions of their respective risk levels, and k is taken according to the regional waterlogging situation. Ⅳ , k Ⅲ , k Ⅱ , k Ⅰ are taken as 0.1, 0.2, 0.3 and 0.4 respectively, z is the unit grid cell length, h max,b is the maximum water depth of the bth grid at the moment, t s The duration of water accumulation; Integrating the risk of water accumulation at flood-prone points within the regulation and storage range, the optimal solution for the primary goal is: Among them, F1 is the primary target, I is the number of storage tanks; V max,i is the risk water volume at the maximum flooding moment within the storage range of the i-th storage pond; After obtaining the optimal value of the primary objective, the optimal value of the primary objective is used as the constraint condition to solve the secondary objective: F2=min{C} (2) Among them, F2 is the secondary target, C is the operation cost function of the storage tank, and the operation cost function of the storage tank is shown in formula (3): Among them, the operation cost function C of the storage tank is the sum of the power consumption of the drainage pump and the flushing system; W m,t is the operating power consumption of the mth pump in the tth period; T m is the total operation time of the pump, which is determined by the amount of water in the pool S i And pump flow Q p,m Calculated; P n is the power consumption of the nth flushing device; T n The total time the flushing equipment is running.

Images