A deep learning-based optimization scheduling method for gate pump groups in urban water supply systems
By combining hypergraph convolutional neural network and Borg algorithm, the problems of low computational efficiency and subjectivity of decision-making variable selection in the optimization scheduling model of gate pump group in urban water supply system are solved, and rapid calculation and efficient gate pump group optimization scheduling are achieved, which is significantly Shorten the model solution time.
Patent Information
- Application Number
- CN202111127085.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-09-26
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2041-09-26
AI Technical Summary
The calculation efficiency of the gate pump group optimization scheduling model of the existing urban water supply system is low, the model debugging time is long, the selection of decision variables is subjective and the calculation is large, which makes it take a long time.
Using a deep learning-based method, the hypergraph convolutional neural network is combined with the urban water supply system optimization scheduling model. By constructing a multi-objective gate pump group optimization scheduling model and water quality and water volume coupling simulation model, the Borg algorithm and genetic operator are used for solution. Combined with the hypergraph convolutional neural network model, the objective function value is quickly calculated, revealing the changes in hydrodynamic conditions and gate pump group scheduling rules.
It significantly shortens the solution time of the multi-objective optimization scheduling model of urban water supply system, simplifies the calculation process, provides a reference for the selection of decision variables, improves the calculation efficiency, and reduces the model debugging time.
Smart Images

Figure CN113850692B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a hypergraph convolutional neural network and computer simulation technology in the field of emergency disaster prevention, and in particular to a deep learning-based optimization scheduling method for a gate pump group of an urban water supply system. Background Art
[0002] The urban water supply system is a huge infrastructure in the world. From the perspective of society, industry and environment, it is vital to the survival of contemporary human beings. The optimization dispatching of the sluice pump group of the urban water supply system is the basic technical means for the dispatching of the water supply system, and it is also one of the important non-engineering measures to improve the efficiency of water supply. It has certain practical significance for ensuring the safety of water use for the people.
[0003] At present, when calculating the optimal dispatch of the gate pump group in the urban water supply system, the selection of decision variables will have a great impact on the calculation results. The selection of decision variables is often based on the personal experience of researchers, which is very subjective. In addition, the existing multi-objective water supply system gate pump group optimal dispatch model calculation requires thousands of iterations to solve the partial differential equation, which has the disadvantages of large amount of calculation and long time consumption. Summary of the invention
[0004] In view of the problems of low computational efficiency, long model debugging time and decision variable selection deviating from reality in the existing optimization scheduling model of sluice pump groups in urban water supply systems, the present invention proposes an optimization scheduling method for sluice pump groups in urban water supply systems based on deep learning, which combines a hypergraph convolutional neural network with the optimization scheduling model of the water supply system to realize the optimization scheduling of sluice pump groups.
[0005] The present invention provides a method for optimizing the scheduling of a gate pump group in an urban water supply system based on deep learning, the method comprising the following steps:
[0006] Step 1: Construct a multi-objective optimization scheduling model for gate and pump groups in the urban water supply system. The purpose of the optimization model is to determine the optimal gate and pump station operation. Therefore, the decision variables of the model are the gate design flow and the pump station flow capacity. The model consists of an objective function and constraints.
[0007] (1) Objective function
[0008] The model takes the lowest energy consumption of pumping station operation, social benefits of water supply and water pollution index as its objective function.
[0009] ① Pump operation energy consumption is expressed as water pump electricity fee
[0010] Objective function expression:
[0011]
[0012] Where: W t is the sum of the electricity costs of the water pumps at time t, wt is the unit electricity cost at time t, N p is the total number of pumps, E jt is the energy consumption rate of pump j at time t, X jt is the speed ratio of pump j at time t.
[0013] ② The social benefits of water supply are expressed as the difference between water supply and demand
[0014] Objective function expression:
[0015]
[0016] Among them: F2(x) is the social benefit value, β j is the water shortage weight coefficient, X ij The water demand of the water plant.
[0017] ③ The water pollution index is expressed by the concentration of pollutants such as BOD5, NH3-N, and dimethyl isoborneol in the water supply system.
[0018] Objective function expression:
[0019]
[0020] Where: C i is the water quality concentration of the water supply system, mg / L; S i is the standard value of the water quality index, mg / L; is the standard value of the water quality index, mg / L;
[0021] (2) Constraints
[0022] ① Constraints on water supply capacity
[0023]
[0024] in: is the minimum water supply of water source i; is the maximum water supply of water source i.
[0025] ② Constraints on water purification capacity of water plants during different time periods:
[0026] p j,t , ,P j,t (5)
[0027] Where: p j,t is the water supply of water plant j at time t; P j,t is the net water output of water plant j at time t.
[0028] ③ Pump station flow capacity constraints:
[0029] 0≤Q p (t)≤Q p,max(6)
[0030]
[0031] Where: Q p (t) is the pumping flow rate of the pth pumping station at time t, m 3 / s;Q p,max is the maximum design flow rate of p pumping stations, m 3 / s;D p (t) is the inflow of the pump station at time t, m 3 / s;D pb is the pumping capacity of the pumping station, m 3 / s; Z p (t-1) is the water level of the forebay of the pumping station at the beginning of period t, m; Z yp is the pre-drainage level of the pumping station, m.
[0032] ④ Gate design flow constraints:
[0033] 0≤Q n (t)≤Q n,max (8)
[0034] Where: Q n (t) is the flow rate of the nth gate at time t, m 3 / s;Q n,max is the design flow rate of the nth gate, m 3 / s.
[0035] Step 2: Construct a water quality and water quantity coupling simulation model for the urban water supply system to simulate the actual hydraulic state of the water supply system. The model considers the impact of different gate openings, pump station speed ratios, and operating numbers on water quality, water supply and demand differences, and pump energy consumption, and is constructed using the measured operating data of the water supply system. The numerical model can be obtained by jointly solving the pipe network continuity equation, energy equation, and head loss equation.
[0036] Step 3: Use the water quality and water quantity coupling simulation model of the urban water supply system constructed in step 2 to simulate the operation of the water supply system under various gate and pump station conditions, build a water quality and water quantity prediction model, and realize the prediction of water quantity, water quality, and pump energy consumption changes under different hydraulic conditions.
[0037] Step 4: Scale and reverse the various water supply system operation data based on the gate and pump station conditions in step 3 and the water quantity, water quality, and pump operation monitoring data of each monitoring point, and take values according to a fixed time step.
[0038] Step 5. Substitute the data processed in step 4 into the hypergraph convolutional neural network model for iterative training to reveal the complex nonlinear mapping relationship between the changes in hydrodynamic conditions and sluice pump group scheduling rules and the water quality, water quantity, and pump operation energy consumption at the nodes of the water supply system.
[0039] Step 6: Use the Borg algorithm to solve the multi-objective optimization scheduling model of the water supply system gate pump group built in step 2 to obtain the optimal water supply system gate pump optimization scheduling result.
[0040] The solution process described in step 6 is as follows:
[0041] ① Normalize the water quality of the water supply system, the energy consumption of the pump operation, and the social benefits of water supply, and set the pollution index of the water supply node, the energy consumption of the pump operation, the water treatment cost, and the water supply and demand difference as the corresponding solution of the objective function X=(n1,n2,···,n i ,α,D x ,ω f ) T , the ε-dominance method is used to sort the solution set, where ε represents the minimum resolution of the solution set, which affects the solution accuracy and evolution speed of the solution set.
[0042] ② Use the Borg algorithm to generate the initial population and establish an external solution set archive to save the current optimal non-dominated solution.
[0043] ③ In this solution process, there are six types of genetic operators, namely, simulated binary crossover operator + polynomial mutation, differential evolution operator + polynomial mutation, uniform mutation operator, parent-centered crossover operator + polynomial mutation, peak normal crossover operator + polynomial mutation, simplex crossover operator + polynomial mutation. Each time an operator is selected for genetic operation, the operator selection rule is: by establishing a feedback mechanism, select the operator that can produce higher quality offspring.
[0044] ④ Randomly and evenly select 1 and k-1 parent individuals from the archive and the population respectively, and then the operator containing k parent individuals will generate a child individual, which will be passed into the multi-objective optimization scheduling model of the water supply system pump group as a decision variable.
[0045] ⑤ Combined with the hypergraph convolutional neural network model constructed in step five, the three objective function values in step one corresponding to the current offspring individual can be quickly calculated.
[0046] ⑥ Check whether the objective function value satisfies the corresponding constraints. If not, readjust the evolution strategy, adjust the solution set, and then change the values of the three objective functions. Repeat the calculation until a solution set that meets the requirements is found.
[0047] Compared with the prior art, the present invention combines the hypergraph convolutional neural network with the water supply system optimization scheduling model, which not only reveals the complex nonlinear mapping relationship between the hydrodynamic conditions and the changes in the scheduling rules of the gate pump group and the water quality, water quantity, and pump operation energy consumption at the nodes of the water supply system, providing a reference for the selection of decision variables; but also effectively solves the problem of solving the complex multi-objective optimization scheduling model of the urban water supply system, significantly shortens the model solution time compared with the conventional solution method, and simplifies the calculation process. The present invention can be applied to some multi-objective urban water supply system gate pump group optimization scheduling problems. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] Figure 1 This is an overall flow chart of a method for optimizing the scheduling of a gate pump group in a city water supply system based on deep learning according to the present invention. DETAILED DESCRIPTION
[0049] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings.
[0050] The hypergraph convolutional neural network model is a model developed in recent years by deep learning. The trained hypergraph convolutional neural network model is used to solve the multi-objective urban water supply system pump group optimization scheduling model, which can greatly improve the computational efficiency. The Borg algorithm has been widely used in solving multi-objective optimization problems due to its good search ability.
[0051] like Figure 1 As shown in FIG. 1 , it is an overall flow chart of a multi-objective gate pump group optimization scheduling method for a water supply system based on a hypergraph convolutional neural network of the present invention. The process specifically includes the following steps:
[0052] Step 1: Construct a multi-objective optimization scheduling model for gate and pump groups in the urban water supply system. The purpose of the optimization model is to determine the optimal gate and pump station operation. Therefore, the decision variables of the model are the gate design flow and the pump station flow capacity. The model consists of an objective function and constraints.
[0053] (1) Objective function
[0054] The model takes the lowest energy consumption of pumping station operation, social benefits of water supply and water pollution index as its objective function.
[0055] ① Pump operation energy consumption is expressed as water pump electricity fee
[0056] Objective function expression:
[0057]
[0058] Where: W t is the sum of the electricity costs of the water pumps at time t, w t is the unit electricity cost at time t, N p is the total number of pumps, E jtis the energy consumption rate of pump j at time t, X jt is the speed ratio of pump j at time t.
[0059] ② The social benefits of water supply are expressed as the difference between water supply and demand
[0060] Objective function expression:
[0061]
[0062] Among them: F2(x) is the social benefit value, β j is the water shortage weight coefficient, X ij The water demand of the water plant.
[0063] ③ The water pollution index is expressed by the concentration of pollutants such as BOD5, NH3-N, and dimethyl isoborneol in the water supply system.
[0064] Objective function expression:
[0065]
[0066] Where: C i is the water quality concentration of the water supply system, mg / L; S i is the standard value of the water quality index, mg / L; is the standard value of the water quality index, mg / L;
[0067] (2) Constraints
[0068] ① Constraints on water supply capacity
[0069]
[0070] in: is the minimum water supply of water source i; is the maximum water supply of water source i.
[0071] ② Constraints on water purification capacity of water plants during different time periods:
[0072] p j,t , ,P j,t (5)
[0073] Where: p j,t is the water supply of water plant j at time t; P j,t is the net water output of water plant j at time t.
[0074] ③ Pump station flow capacity constraints:
[0075] 0≤Q p (t)≤Q p,max (6)
[0076]
[0077] Where: Q p (t) is the pumping flow rate of the pth pumping station at time t, m 3 / s;Q p,max is the maximum design flow rate of p pumping stations, m 3 / s;D p (t) is the inflow of the pump station at time t, m 3 / s;D pb is the pumping capacity of the pumping station, m 3 / s; Z p (t-1) is the water level of the forebay of the pumping station at the beginning of period t, m; Z yp is the pre-drainage level of the pumping station, m.
[0078] ④ Gate design flow constraints:
[0079] 0≤Q n (t)≤Q n,max (8)
[0080] Where: Q n (t) is the flow rate of the nth gate at time t, m 3 / s;Q n,max is the design flow rate of the nth gate, m 3 / s.
[0081] Step 2: Construct a water quality and water quantity coupling simulation model for the urban water supply system to simulate the actual hydraulic state of the water supply system. The model considers the impact of different gate openings, pump station speed ratios, and operating numbers on water quality, water supply and demand differences, and pump energy consumption, and is constructed using the measured operating data of the water supply system. The numerical model can be obtained by jointly solving the pipe network continuity equation, energy equation, and head loss equation.
[0082] Step 3: Use the water quality and water quantity coupling simulation model of the urban water supply system constructed in step 2 to simulate the operation of the water supply system under various gate and pump station conditions, build a water quality and water quantity prediction model, and realize the prediction of water quantity, water quality, and pump energy consumption changes under different hydraulic conditions.
[0083] Step 4: Scale and reverse the various water supply system operation data based on the gate and pump station conditions in step 3 and the water quantity, water quality, and pump operation monitoring data of each monitoring point, and take values according to a fixed time step.
[0084] Step 5. Substitute the data processed in step 4 into the hypergraph convolutional neural network model for iterative training to reveal the complex nonlinear mapping relationship between the changes in hydrodynamic conditions and sluice pump group scheduling rules and the water quality, water quantity, and pump operation energy consumption at the nodes of the water supply system.
[0085] Step 6: Use the Borg algorithm to solve the multi-objective optimization scheduling model of the water supply system gate pump group built in step 2 to obtain the optimal water supply system gate pump optimization scheduling result.
[0086] The solution process described in step 6 is as follows:
[0087] ① Normalize the water quality of the water supply system, the energy consumption of the pump operation, and the social benefits of water supply, and set the pollution index of the water supply node, the energy consumption of the pump operation, the water treatment cost, and the water supply and demand difference as the corresponding solution of the objective function X=(n1,n2,···,n i ,α,D x ,ω f ) T , the precision values of the three objective functions (water pump electricity fee, water supply and demand difference, water quality pollution index) are set to 0.001, 0.0001, and 0.01, the initial population size is 200, the evolutionary generation is 5000, and the ε-dominance method is used to sort the solution set.
[0088] ② Use the Borg algorithm to generate the initial population and establish an external solution archive to save the current optimal non-dominated solution.
[0089] ③ In this solution process, there are six types of genetic operators, namely, simulated binary crossover operator + polynomial mutation, differential evolution operator + polynomial mutation, uniform mutation operator, parent-centered crossover operator + polynomial mutation, peak normal crossover operator + polynomial mutation, simplex crossover operator + polynomial mutation. Each time an operator is selected for genetic operation, the operator selection rule is: by establishing a feedback mechanism, select the operator that can produce higher quality offspring.
[0090] ④ Randomly and evenly select 1 and k-1 parent individuals from the archive and the population respectively, and then the operator containing k parent individuals will generate a child individual, which will be passed into the multi-objective optimization scheduling model of the water supply system pump group as a decision variable.
[0091] ⑤ Combined with the hypergraph convolutional neural network model constructed in step five, the three objective function values in step one corresponding to the current offspring individual can be quickly calculated.
[0092] ⑥ Check whether the objective function value satisfies the corresponding constraints. If not, readjust the evolution strategy, adjust the solution set, and then change the values of the three objective functions. Repeat the calculation until a solution set that meets the requirements is found.
Claims
1. A method for optimizing the scheduling of a sluice pump group in an urban water supply system based on deep learning, characterized in that: The method is carried out according to the following steps: Step 1: construct a multi-objective gate-pump group optimization scheduling model for the urban water supply system to determine the optimal gate and pump station operation. The decision variables of the model are the gate design flow and the pump station flow capacity. The model consists of an objective function and constraint conditions. (1) Objective function The model takes the minimum energy consumption of pumping station operation, social benefits of water supply, and water pollution index as its objective function; ① Pump operation energy consumption is expressed as water pump electricity fee Objective function expression: Where: W t is the sum of the electricity costs of the water pumps at time t, w t is the unit electricity cost at time t, N p is the total number of pumps, E jt is the energy consumption rate of pump j at time t, X jt is the speed ratio of water pump j at time t; ② The social benefits of water supply are expressed as the difference between water supply and demand Objective function expression: Among them: F2(x) is the social benefit value, β j is the water shortage weight coefficient; ③ The water pollution index is expressed by the concentration of BOD5, NH3-N and dimethylisoborneol pollutants in the water supply system. Objective function expression: Where: C i is the water quality concentration of the water supply system, mg / L; S i is the standard value of the water quality index, mg / L; is the standard value of the water quality index, mg / L; (2) Constraints ① Constraints on water supply capacity in: is the minimum water supply of water source i; is the maximum water supply of water source i; ② Constraints on water purification capacity of water plants during different time periods: p j,i≥ P j,t (5) Where: p j,t is the water supply of water plant j at time t; P j,t is the net water output of water plant j at time t; ③ Pump station flow capacity constraints: 0≤Q p (t)≤Q p,max (6) Where: Q p (t) is the pumping flow rate of the pth pumping station at time t, m 3 / s;Q p,max is the maximum design flow rate of p pumping stations, m 3 / s;D p (t) is the inflow of the pump station at time t, m 3 / s;D pb is the pumping capacity of the pumping station, m 3 / s; Z p (t-1) is the water level of the forebay of the pumping station at the beginning of period t, m; Z yp is the pre-drainage level of the pumping station, m; ④ Gate design flow constraints: 0≤Q n (t)≤Q n,max (8) Where: Q n (t) is the flow rate of the nth gate at time t, m 3 / s;Q n,max is the design flow rate of the nth gate, m 3 / s; Step 2: Construct a water quality and water quantity coupling simulation model for the urban water supply system. The model considers the impact of different gate openings, pump station speed ratios, and operating numbers on water quality, water supply and demand difference, and pump energy consumption. The model is constructed using the measured operating data of the water supply system. The numerical model is obtained by jointly solving the pipe network continuity equation, energy equation, and head loss equation. Step 3: Through the water quality and water quantity coupling simulation model of the urban water supply system constructed in step 2, simulate the operation of the water supply system under various gate and pump station conditions, build a water quality and water quantity prediction model, and realize the prediction of water quantity, water quality, and pump energy consumption changes under different hydraulic conditions; Step 4: Scale and reverse the various water supply system operation data based on the gate and pump station conditions in step 3 and the water quantity, water quality, and pump operation monitoring data of each monitoring point, and take values according to a fixed time step; Step 5: Substitute the data processed in step 4 into the hypergraph convolutional neural network model for iterative training to reveal the complex nonlinear mapping relationship between the changes in hydrodynamic conditions and the scheduling rules of the sluice pump group and the water quality, water volume, and pump operation energy consumption at the nodes of the water supply system; Step 6: Use the Borg algorithm to solve the multi-objective optimization scheduling model of the water supply system sluice pump group built in step 2 to obtain the optimal water supply system sluice pump optimization scheduling result; The solution process described in step 6 is as follows: ① Normalize the water quality of the water supply system, the energy consumption of the pump operation, and the social benefits of water supply, and set the pollution index of the water supply node, the energy consumption of the pump operation, the water treatment cost, and the water supply and demand difference as the corresponding solution of the objective function X=(n1,n2,…,n i ,α,D x ,ω f ) T , the ε-dominance method is used to sort the solution set, where ε represents the minimum resolution of the solution set, which affects the solution accuracy and evolution speed of the solution set; ② Use the Borg algorithm to generate the initial population and establish an external solution set archive to save the current optimal non-dominated solution; ③ In this solution process, there are six types of genetic operators, namely, simulated binary crossover operator + polynomial mutation, differential evolution operator + polynomial mutation, uniform mutation operator, parent-centered crossover operator + polynomial mutation, peak normal crossover operator + polynomial mutation, simplex crossover operator + polynomial mutation; each time an operator is selected for genetic operation, and the operator selection rule is: by establishing a feedback mechanism, select the operator that can produce higher quality offspring; ④ Randomly and evenly select 1 and k-1 parent individuals from the archive and population respectively, and then the operator containing k parent individuals will generate a child individual, which will be passed into the multi-objective optimization scheduling model of the water supply system gate pump group as a decision variable; ⑤ Combined with the hypergraph convolutional neural network model constructed in step 5, the three objective function values in step 1 corresponding to the current offspring individual can be quickly calculated; ⑥ Check whether the objective function value satisfies the corresponding constraints. If not, readjust the evolution strategy, adjust the solution set, and then change the values of the three objective functions. Repeat the calculation until a solution set that meets the requirements is found.
