An optimal dispatching method for power-transportation network based on deep learning decoupling

Abstract

The present invention discloses an optimal dispatching method for an electric power-transportation network based on deep learning decoupling. By decoupling the electric power-transportation network model through deep learning, the optimal dispatching of the electric power-transportation network is achieved on the basis of ensuring the information privacy of the electric power grid and the transportation network. Taking into account the operation constraints of the power grid and the transportation network, a coordinated dispatching model of the electric power-transportation network taking into account electric vehicles is constructed; deep learning is used to learn the charging power of electric vehicles between the power grid and the transportation network, and the coordinated dispatching model of the electric power-transportation network is realized. The two networks operate independently, and the optimal dispatching model of the electric power-transportation network based on deep learning decoupling is constructed. The present invention takes into account the reality that there are information barriers between the current power grid and the transportation network. Through deep learning decoupling, the optimal dispatching of the electric power-transportation network is achieved on the basis of ensuring the independence of the operation of the power grid and the transportation network, and the solution efficiency of the model is effectively improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of electricity-transportation network collaborative optimization, and specifically relates to an optimal scheduling method for electricity-transportation networks based on deep learning decoupling. Background Art

[0002] The power-transportation network is a fusion of the power system and the transportation system, and is an important part of the future intelligent development of cities. The power-transportation network needs to consider not only the stability and reliability of the power system, but also the efficiency and convenience of the transportation system. Therefore, the optimal scheduling of the power-transportation network is an extremely challenging task. The traditional physical model collaborative scheduling not only has the problem of difficulty in crossing the information barrier between the power grid and the transportation network, but also has a slow solution efficiency and is difficult to meet the requirements of real-time scheduling. Based on the above description, the optimal scheduling method for the power-transportation network based on deep learning decoupling proposed in the present invention mainly considers the following two aspects: First, the optimal scheduling model of the power-transportation network based on deep learning decoupling is established, which not only guarantees the advantages of coordinated scheduling on the basis of decoupling the power grid and the transportation network, but also protects the information privacy of the two networks; second, through the deep learning model, the solution efficiency of the model is greatly improved. Summary of the invention

[0003] Purpose of the invention: The technical problem to be solved by the present invention is to provide an optimal scheduling method for power-transportation network based on deep learning decoupling in view of the deficiencies in the prior art. The present invention takes into account the coordination of the power system and the transportation network system, introduces a deep learning model to learn the charging power of electric vehicles between the power grid and the transportation network, and realizes the independent operation of the two networks on the basis of ensuring coordinated operation. The present invention can not only realize the optimal scheduling of the operation of the power-transportation network, but also ensure the information privacy of the power grid and the transportation network on this basis, and improve the efficiency of solving the model.

[0004] Technical solution: In order to solve the above technical problems, the present invention provides an optimal scheduling method for power-transportation network based on deep learning decoupling, comprising the following steps:

[0005] Step 1, obtaining the network coefficients and operation coefficients of the power grid and traffic network models, wherein the network coefficients include the resistance and impedance of the power grid lines, the free travel time of the traffic network sections and the section capacity, and the section and path coupling coefficients, and the operation coefficients include the power generation coefficient of the generator set and the unit travel time cost;

[0006] Step 2: Obtain grid load demand, traffic demand and photovoltaic output scenario data;

[0007] Step 3: Based on the network coefficients, operation coefficients and scenario data of the power grid and transportation network models, a coordinated dispatching model of the power grid and transportation network taking electric vehicles into account is established with the minimum energy consumption of the system operation as the objective function and the power grid operation constraints and transportation network operation constraints as the constraints, and the charging power of electric vehicles in the coordinated dispatching model of the power grid and transportation network taking electric vehicles into account is solved;

[0008] Step 4: Based on the electric vehicle charging power of the electricity-transportation network coordinated scheduling model taking electric vehicles into account in step 3, with grid load demand and traffic demand as input and electric vehicle charging power as output, a power-transportation network optimal scheduling model based on deep learning decoupling is established through deep learning model learning;

[0009] Step 5: Based on the optimal dispatching model of the power-transportation network based on deep learning decoupling in step 4, a nonlinear optimization solver is used to solve the model, and the power grid and transportation network are optimally dispatched to obtain the optimal dispatching scheme of the power-transportation network based on deep learning decoupling.

[0010] Furthermore, in step 3, the objective function is to minimize the energy consumption of the system operation:

[0011]

[0012] In the formula, F eco Energy consumption for system operation, Energy consumption for grid operation, Energy consumption for transportation network;

[0013]

[0014] In the formula, E N represents the set of power grid nodes; π(0) represents the set of nodes connected to the root node of the power grid; rs represents the set of starting and ending points of the traffic demand; b i represents the cost coefficient of distributed generation at node i; and represent the active power of distributed generation at node j; v and They represent the electricity price and electricity quantity of the upper main grid of the distribution network respectively; represents the minimum toll for an electric vehicle with a starting point r and a destination s; q rs It represents the number of vehicles with a starting point r and a destination s.

[0015] Furthermore, in step 3, the power grid operation constraints and transportation network operation constraints are:

[0016] (1) Grid operation constraints:

[0017]

[0018]

[0019]

[0020]

[0021]

[0022]

[0023]

[0024]

[0025] In the formula, and Respectively represent the active power and reactive power of line ij; represents the reactive power of distributed generation at node j; They represent the resistance, reactance and impedance of line ij respectively; represents the current of circuit ij; π(j) represents the set of child nodes h connected to node j; and They represent the active power and reactive power of line jh respectively; and Respectively represent the active and reactive loads of node j; U j represents the square of the voltage amplitude at node j; U i represents the square of the voltage amplitude at node i; Represents the upper limit of the current of circuit ij; and They represent the lower and upper limits of the square of the voltage amplitude at node i, respectively. and They represent the lower and upper limits of the active output of the distributed generation at node i respectively; and They represent the lower and upper limits of the reactive power output of distributed generation at node i respectively; represents the conventional power load at node j; represents the charging power of the electric vehicle at node j; C e (j) represents the set of charging sections connected to node j; represents the traffic flow of charging section a; E e Indicates the charging demand of electric vehicles;

[0026] (2) Traffic network operation constraints:

[0027]

[0028]

[0029]

[0030]

[0031]

[0032]

[0033]

[0034] In the formula, represents the set of paths k of electric vehicles with a starting point of r and a destination of s; represents the traffic flow of electric vehicles that choose path k; Respectively represent the conventional road section and the charging road section set; Respectively represent the traffic flow of the regular section and the charging section; Respectively represent the number of vehicles that can be accommodated in the regular section and the charging section; They represent the driving time on regular roads and the charging time respectively; They represent the free travel time of the conventional section and the charging section under zero traffic flow respectively; J is the charging time model parameter; represents the toll that an electric vehicle with a starting point r and a destination s needs to pay when choosing path k; ω, Respectively represent the unit driving time cost and charging price; Toll a , They represent congestion pricing and charging service fee respectively; A binary constant representing whether the road section a where the electric vehicle is located belongs to the path k. If so, it takes 1, otherwise it takes 0.

[0035] Furthermore, in step 4, the deep learning model is:

[0036] y=π(Wx+b) (A-18)

[0037] Where x∈R n represents the input vector, i.e. the number of vehicles in traffic demand q rs and conventional power load y∈R m Represents the predicted output vector, i.e., the charging power of the electric vehicle W∈R m×n Represents the weight matrix obtained after deep learning model learning; b∈R m is the bias vector obtained after deep learning model learning; π(•) represents the nonlinear activation function ReLU; R n , R m, R m×n They represent the n-dimensional real number set, the m-dimensional real number set, and the m×n-dimensional real number set respectively.

[0038] Further, in step 4, the electric vehicle charging power in formula (A-9) is Replaced with electric vehicle charging power predicted by deep learning The optimal dispatching model of the power grid based on deep learning decoupling is:

[0039]

[0040] Predicted electric vehicle charging power Reflected in the distribution of electric vehicles in the transportation network, The traffic flow of the charging section in the traffic network is obtained. It represents the traffic flow of the charging section predicted by deep learning. The optimal scheduling model of the traffic network based on deep learning decoupling is:

[0041]

[0042] Furthermore, in step 5, a nonlinear optimization solver is used to solve the optimal scheduling model of the power-transportation network based on deep learning decoupling, and the power grid and transportation network are optimized to obtain the optimal scheduling plan of the power-transportation network based on deep learning decoupling.

[0043] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0044] Compared with the basic scheme of traditional power-transportation network operation and dispatch, the technical scheme of the present invention introduces a deep learning model to learn the coupling relationship between the power grid and the transportation network, thereby realizing the decoupling of the power grid and the transportation network, so that each can be optimized separately. The results of the example test show that the method proposed by the present invention can greatly improve the efficiency of model solving compared with the existing method, and realize the decoupling optimization of the two networks on the basis of maintaining the information privacy of the power grid and the transportation network, while maintaining the operability of coordinated dispatch. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 is a flow chart of the method of the present invention;

[0046] Figure 2 This is an example diagram of the electricity-transportation network coordination model. DETAILED DESCRIPTION

[0047] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent modifications of the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0048] like Figure 1 As shown, the present invention provides an optimal scheduling method for power-transportation network based on deep learning decoupling, comprising the following steps:

[0049] Step 1, obtaining the network coefficients and operation coefficients of the power grid and traffic network models, wherein the network coefficients include the resistance and impedance of the power grid lines, the free travel time of the traffic network sections and the section capacity, and the section and path coupling coefficients, and the operation coefficients include the power generation coefficient of the generator set and the unit travel time cost;

[0050] Step 2: Obtain grid load demand, traffic demand and photovoltaic output scenario data;

[0051] Step 3: Based on the network coefficients, operation coefficients and scenario data of the power grid and transportation network models, a coordinated dispatching model of the power grid and transportation network taking electric vehicles into account is established with the minimum energy consumption of the system operation as the objective function and the power grid operation constraints and transportation network operation constraints as the constraints, and the charging power of electric vehicles in the coordinated dispatching model of the power grid and transportation network taking electric vehicles into account is solved;

[0052] Step 4: Based on the electric vehicle charging power of the electricity-transportation network coordinated scheduling model taking electric vehicles into account in step 3, with grid load demand and traffic demand as input and electric vehicle charging power as output, a power-transportation network optimal scheduling model based on deep learning decoupling is established through deep learning model learning;

[0053] Step 5: Based on the optimal dispatching model of the power-transportation network based on deep learning decoupling in step 4, a nonlinear optimization solver is used to solve the model, and the power grid and transportation network are optimally dispatched to obtain the optimal dispatching scheme of the power-transportation network based on deep learning decoupling.

[0054] Furthermore, in step 3, the objective function is to minimize the energy consumption of the system operation:

[0055]

[0056] In the formula, F eco Energy consumption for system operation, Energy consumption for grid operation, Energy consumption for transportation network;

[0057]

[0058] In the formula, E N represents the set of power grid nodes; π(0) represents the set of nodes connected to the root node of the power grid; rs represents the set of starting and ending points of the traffic demand; b i represents the cost coefficient of distributed generation at node i; and represent the active power of distributed generation at node j; ν and They represent the electricity price and electricity quantity of the upper main grid of the distribution network respectively; represents the minimum toll for an electric vehicle with a starting point r and a destination s; q rs It represents the number of vehicles with a starting point r and a destination s.

[0059] Furthermore, in step 3, the power grid operation constraints and transportation network operation constraints are:

[0060] (1) Grid operation constraints:

[0061]

[0062]

[0063]

[0064]

[0065]

[0066]

[0067]

[0068]

[0069] In the formula, and Respectively represent the active power and reactive power of line ij; represents the reactive power of distributed generation at node j; They represent the resistance, reactance and impedance of line ij respectively; represents the current of circuit ij; π(j) represents the set of child nodes h connected to node j; and They represent the active power and reactive power of line jh respectively; and Respectively represent the active and reactive loads of node j; U j represents the square of the voltage amplitude at node j; U i represents the square of the voltage amplitude at node i; Represents the upper limit of the current of circuit ij; and They represent the lower and upper limits of the square voltage amplitude at node i, respectively. and They represent the lower and upper limits of the active output of the distributed generation at node i respectively; and They represent the lower and upper limits of the reactive power output of distributed generation at node i respectively; represents the conventional power load at node j; represents the charging power of the electric vehicle at node j; C e (j) represents the set of charging sections connected to node j; represents the traffic flow of charging section a; E e Indicates the charging demand of electric vehicles;

[0070] (2) Traffic network operation constraints:

[0071]

[0072]

[0073]

[0074]

[0075]

[0076]

[0077]

[0078] In the formula, represents the set of paths k of electric vehicles with a starting point of r and a destination of s; represents the traffic flow of electric vehicles that choose path k; Respectively represent the conventional road section and the charging road section set; Respectively represent the traffic flow of the regular section and the charging section; Respectively represent the number of vehicles that can be accommodated in the regular section and the charging section; They represent the driving time on regular roads and the charging time respectively; They represent the free travel time of the conventional section and the charging section under zero traffic flow respectively; J is the charging time model parameter; represents the toll that an electric vehicle with a starting point r and a destination s needs to pay when choosing path k; ω, Respectively represent the unit driving time cost and charging price; Toll a , They represent congestion pricing and charging service fee respectively; A binary constant representing whether the road section a where the electric vehicle is located belongs to the path k. If so, it takes 1, otherwise it takes 0.

[0079] Furthermore, in step 4, the deep learning model is:

[0080] y=π(Wx+b) (A-18)

[0081] Where x∈R n represents the input vector, i.e. the number of vehicles in traffic demand q rs and conventional power load y∈R m Represents the predicted output vector, i.e., the charging power of the electric vehicle W∈R m×n Represents the weight matrix obtained after deep learning model learning; b∈R m is the bias vector obtained after deep learning model learning; π(·) represents the nonlinear activation function ReLU; R n , R m , R m×n They represent the n-dimensional real number set, the m-dimensional real number set, and the m×n-dimensional real number set respectively.

[0082] Further, in step 4, the electric vehicle charging power in formula (A-9) is Replaced with electric vehicle charging power predicted by deep learning The optimal dispatching model of the power grid based on deep learning decoupling is:

[0083]

[0084] Predicted electric vehicle charging power Reflected in the distribution of electric vehicles in the transportation network, The traffic flow of the charging section in the traffic network is obtained. It represents the traffic flow of the charging section predicted by deep learning. The optimal scheduling model of the traffic network based on deep learning decoupling is:

[0085]

[0086] Furthermore, in step 5, a nonlinear optimization solver is used to solve the optimal scheduling model of the power-transportation network based on deep learning decoupling, and the power grid and transportation network are optimized to obtain the optimal scheduling plan of the power-transportation network based on deep learning decoupling.

[0087] Case Analysis

[0088] The following example illustrates the superiority of the optimal dispatching of the power-transportation network based on deep learning decoupling described in the present invention. Figure 2The IEEE 33-node power system and the 13-node Nguyen-Dupuis transportation network system are shown. In order to compare the superiority of the method proposed in the present invention, the power-transportation network coordinated scheduling method driven by the traditional physical model and the power-transportation network optimal scheduling method based on deep learning decoupling proposed in the present invention are respectively used for power-transportation network coordinated scheduling. The present invention is implemented through the GAMS optimization platform and uses the IPOPT solver to solve the nonlinear programming problem.

[0089] Based on this example, the changes in learning effects before and after the Bayesian tuning of the deep learning model hyperparameters (see Table 1 for results) and the comparison of solution efficiency and operating costs between the traditional physical model-driven electricity-transportation network coordinated scheduling method and the electricity-transportation network optimal scheduling method based on deep learning decoupling proposed in the present invention are shown (see Table 2 and Table 3 for results). It shows that after the Bayesian hyperparameters are tuned, the deep learning model can fit the power-transportation network well, ensuring the accuracy of the model solution results (Table 1). Compared with the traditional physical model-driven solution, deep learning decoupling greatly improves the solution efficiency, and the calculation time of the power grid and transportation network is shortened by more than 98% (Table 2), and the operating costs of the power grid and transportation network are almost the same, indicating that deep learning decoupling maintains the operability of collaborative optimization while improving the calculation efficiency.

[0090] Table 1 R before and after Bayesian hyperparameter tuning 2 contrast

[0091] Hyperparameter Type The first hidden layer nodes The second hidden layer nodes The third hidden layer nodes Learning Rate batch <![CDATA[R 2 value<!-- 7 --> ]]> Initial hyperparameters 150 600 200 0.05 32 0.74 Hyperparameters after Bayesian tuning 140 566 252 0.00546 64 0.91

[0092] Table 2 Comparison of solution efficiency of different optimization methods

[0093] Table 3 Comparison of operating costs of different optimization methods

[0094]

[0095] The above description is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a technician familiar with the technical field within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention.

Claims

1. A method for optimal dispatching of power-transportation networks based on deep learning decoupling, characterized in that: The following steps are involved: Step 1, obtaining the network coefficients and operation coefficients of the power grid and traffic network models, wherein the network coefficients include the resistance and impedance of the power grid lines, the free travel time of the traffic network sections and the section capacity, and the section and path coupling coefficients, and the operation coefficients include the power generation coefficient of the generator set and the unit travel time cost; Step 2: Obtain grid load demand, traffic demand and photovoltaic output scenario data; Step 3: Based on the network coefficients, operation coefficients and scenario data of the power grid and transportation network models, a coordinated dispatching model of the power grid and transportation network taking electric vehicles into account is established with the minimum energy consumption of the system operation as the objective function and the power grid operation constraints and transportation network operation constraints as the constraints, and the charging power of electric vehicles in the coordinated dispatching model of the power grid and transportation network taking electric vehicles into account is solved; Step 4: Based on the electric vehicle charging power of the electricity-transportation network coordinated scheduling model taking electric vehicles into account in step 3, with grid load demand and traffic demand as input and electric vehicle charging power as output, a power-transportation network optimal scheduling model based on deep learning decoupling is established through deep learning model learning; Step 5: Based on the optimal dispatching model of the power-transportation network based on deep learning decoupling in step 4, a nonlinear optimization solver is used to solve the model, and the power grid and the transportation network are optimally dispatched to obtain the optimal dispatching scheme of the power-transportation network based on deep learning decoupling; In step 3, the objective function is to minimize the energy consumption of the system: In the formula, F eco Energy consumption for system operation, Energy consumption for grid operation, Energy consumption for transportation network; In the formula, E N represents the set of power grid nodes; π(0) represents the set of nodes connected to the root node of the power grid; rs represents the set of starting and ending points of the traffic demand; b i represents the cost coefficient of distributed generation at node i; represents the active power of the distributed generation at node i; ν and They represent the electricity price and electricity quantity of the upper main grid of the distribution network respectively; represents the minimum toll for an electric vehicle with a starting point r and a destination s; q rs It represents the number of vehicles with the starting point r and the destination s in the traffic demand; In step 3, the power grid operation constraints and transportation network operation constraints are: (1) Grid operation constraints: In the formula, and Respectively represent the active power and reactive power of line ij; represents the reactive power of distributed generation at node j, represents the active power of distributed generation at node j; They represent the resistance, reactance and impedance of line ij respectively; represents the current of circuit ij; π(j) represents the set of child nodes h connected to node j; and They represent the active power and reactive power of line jh respectively; and Respectively represent the active and reactive loads of node j; U j represents the square of the voltage amplitude at node j; U i represents the square of the voltage amplitude at node i; Represents the upper limit of the current of circuit ij; and They represent the lower and upper limits of the square voltage amplitude at node i, respectively. and They represent the lower and upper limits of the active output of the distributed generation at node i respectively; and They represent the lower and upper limits of the reactive power output of distributed generation at node i respectively; represents the conventional power load at node j; represents the charging power of the electric vehicle at node j; C e (j) represents the set of charging sections connected to node j; represents the traffic flow of charging section a; E e Indicates the charging demand of electric vehicles; (2) Traffic network operation constraints: In the formula, represents the set of paths k of electric vehicles with a starting point of r and a destination of s; represents the traffic flow of electric vehicles that choose path k; Respectively represent the conventional road section and the charging road section set; Respectively represent the traffic flow of the regular section and the charging section; Respectively represent the number of vehicles that can be accommodated in the regular section and the charging section; They represent the driving time on regular roads and the charging time respectively; They represent the free travel time of the conventional section and the charging section under zero traffic flow respectively; J is the charging time model parameter; represents the toll that an electric vehicle with a starting point r and a destination s needs to pay when choosing path k; ω, Respectively represent the unit driving time cost and charging price; Toll a , They represent congestion pricing and charging service fee respectively; A binary constant representing whether the road section a where the electric vehicle is located belongs to the path k. If so, it takes 1, otherwise it takes 0; In step 4, the deep learning model is: y=π(Wx+b)(A-18) Where x∈R n represents the input vector, i.e. the number of vehicles in traffic demand q rs and conventional power load y∈R m Represents the predicted output vector, i.e., the charging power of the electric vehicle W∈R m×n Represents the weight matrix obtained after deep learning model learning; b∈R m is the bias vector obtained after deep learning model learning; π(·) represents the nonlinear activation function ReLU; R n , R m , R m×n They represent n-dimensional real number set, m-dimensional real number set, and m×n-dimensional real number set respectively; In step 4, the electric vehicle charging power in equation (A-9) is Replaced with electric vehicle charging power predicted by deep learning The optimal dispatching model of the power grid based on deep learning decoupling is: Predicted electric vehicle charging power Reflected in the distribution of electric vehicles in the transportation network, The traffic flow of the charging section in the traffic network is obtained. It represents the traffic flow of the charging section predicted by deep learning. The optimal scheduling model of the traffic network based on deep learning decoupling is:

2. The method for optimal dispatching of power-transportation network based on deep learning decoupling according to claim 1, characterized in that: In step 5, a nonlinear optimization solver is used to solve the optimal scheduling model of the power-transportation network based on deep learning decoupling, and the power grid and transportation network are optimized to obtain the optimal scheduling plan of the power-transportation network based on deep learning decoupling.

Images