A power distribution network topology reconfiguration method and system

Abstract

The application provides a power distribution network topology reconstruction method and system, the method comprising: constructing a topology reconstruction environment of a power distribution network according to node data and branch data of the power distribution network, wherein the reward of the topology reconstruction environment is the maximum load margin of the power distribution network; solving the topology reconstruction environment based on a deep reinforcement learning algorithm to obtain the power distribution network topology corresponding to the maximum load margin. The application reconstructs the current topology structure of the power distribution network according to the environmental state data of the power distribution network based on the deep reinforcement learning algorithm, thereby better improving the reliability of the power distribution network, reducing line loss and balancing the load.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power distribution network reconstruction, and particularly relates to a power distribution network topology reconstruction method and system. BACKGROUND

[0002] Power distribution network reconstruction is to determine the optimal operation mode of the power distribution network in terms of line loss, load balance and power supply quality, under the premise of meeting the requirements of radial distribution network, feeder hot melting, node voltage deviation and transformer capacity. Since there are a large number of sectional switches and tie switches in the power distribution network, the power distribution network reconstruction is a multi-objective nonlinear mixed optimization problem. The power distribution network reconstruction technology is to improve the reliability, reduce the line loss, balance the load and improve the power supply voltage quality by changing the topology structure of the power distribution network.

[0003] At present, the algorithms for power distribution network reconstruction can be roughly divided into traditional mathematical methods, heuristic methods and artificial intelligence algorithms. The traditional mathematical methods mainly include linear or nonlinear programming methods such as simplex method and branch and bound method. The method is to describe the power distribution network reconstruction problem by using a mathematical model, and then solve it by a certain algorithm to obtain the optimization result. The traditional mathematical optimization method is relatively mature, and can obtain a global optimal solution independent of the initial structure of the power distribution network. However, it has a long calculation time and a serious "dimension disaster", and cannot be used for processing complex large-scale power systems. When solving the optimization problem of the power system by using the traditional mathematical method, simplification and approximation are often required. The heuristic algorithm is an algorithm based on intuitive analysis. It usually iterates gradually according to certain principles until a satisfactory optimization result is obtained. In the current power distribution network reconstruction, the commonly used heuristic algorithms mainly include optimal flow mode algorithm and branch exchange method. However, the convergence of the heuristic algorithm for power distribution network reconstruction has no theoretical guarantee, and the actual effect is unstable. The artificial intelligence algorithm based on randomization technology has been widely applied in the optimization field in recent years. The theoretical analysis of the algorithm has developed rapidly and gradually has theoretical support. However, most artificial intelligence algorithms are still at the theoretical level and have not been well applied to the problem of power distribution network reconstruction.

[0004] Therefore, there is an urgent need for a power distribution network topology reconstruction method and system to solve the above problems. SUMMARY

[0005] In view of the problems existing in the prior art, the present application provides a power distribution network topology reconstruction method and system.

[0006] The present application provides a power distribution network topology reconstruction method, comprising:

[0007] According to the node data and branch data of the power distribution network, a topology reconstruction environment of the power distribution network is constructed, wherein the reward of the topology reconstruction environment is the maximum load margin of the power distribution network;

[0008] Solving the topology reconfiguration environment based on a deep reinforcement learning algorithm to obtain a power distribution network topology corresponding to the maximum load margin.

[0009] According to the power distribution network topology reconfiguration method provided by the application, the node data includes the number of load nodes, the load node injection power and the load node load margin; the branch data includes the number of power distribution network branches, the power distribution network branch opening state and the power distribution network branch power flow data.

[0010] According to the power distribution network topology reconfiguration method provided by the application, the topology reconfiguration environment of the power distribution network is constructed according to the node data and the branch data of the power distribution network, including:

[0011] According to the number of load nodes and the number of branches, an action space is constructed;

[0012] According to the load node injection power, the power distribution network branch opening state, the power distribution network branch power flow data and the load node load margin, a state space is constructed;

[0013] According to the action space and the state space, the maximum load margin of the power distribution network is taken as the reward, and the topology reconfiguration environment of the power distribution network is constructed.

[0014] According to the power distribution network topology reconfiguration method provided by the application, the action space is constructed according to the number of load nodes and the number of branches, including:

[0015] According to the number of load nodes and the number of branches, the maximum number of branch opening combinations in the power distribution network is obtained, and the formula is:

[0016] M×(M-1)+1;

[0017] Wherein, M represents the number of branches;

[0018] According to the maximum number of opening combinations, the action space is constructed.

[0019] According to the power distribution network topology reconfiguration method provided by the application, the topology reconfiguration environment is solved based on a deep reinforcement learning algorithm to obtain a power distribution network topology corresponding to the maximum load margin, including:

[0020] According to the deep reinforcement learning algorithm, each branch opening combination in the action space and the corresponding state space are iteratively solved to obtain the node injection power in each branch opening combination and the node voltage value after power injection;

[0021] According to the node original voltage value, the node injection power and the node voltage value after power injection, the load node load margin in the current iteration process is obtained.

[0022] Based on the load margin of the load nodes, the load margin of the distribution network is obtained. If the preset conditions are met, the iteration stops and the load margin of the distribution network obtained in the current iteration process is taken as the maximum load margin. The distribution network topology is then reconstructed according to the branch disconnection combination corresponding to the maximum load margin.

[0023] According to a distribution network topology reconfiguration method provided by the present invention, obtaining the load margin of load nodes in the current iteration process based on the original node voltage value, the injected power of the node, and the node voltage value after power injection includes:

[0024] Obtain the raw voltage value U of the load node origin ;

[0025] Inject power Δ into all load nodes in the current branch disconnection combination, and obtain the node voltage value U after power injection. Step ;

[0026] According to the original voltage value U origin and the node voltage value U after power injection Step The load margin D of the load node is calculated using the following formula:

[0027] D=(U Step -U origin ) / Δ.

[0028] According to the present invention, a distribution network topology reconfiguration method is provided, wherein the deep reinforcement learning algorithm is the Dueling-DQN algorithm.

[0029] The present invention also provides a distribution network topology reconfiguration system, comprising:

[0030] The topology environment construction module is used to construct the topology reconfiguration environment of the distribution network based on the node data and branch data of the distribution network, wherein the reward of the topology reconfiguration environment is the maximum load margin of the distribution network.

[0031] The reconfiguration module is used to solve the topology reconfiguration environment based on a deep reinforcement learning algorithm to obtain the distribution network topology corresponding to the maximum load margin.

[0032] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the above-described power distribution network topology reconfiguration methods.

[0033] The application further provides a non-transitory computer-readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the power distribution network topology reconfiguration method.

[0034] The application provides a power distribution network topology reconfiguration method and system, which is based on a deep reinforcement learning algorithm, reconfigures the current topology structure of the power distribution network according to the environmental state data of the power distribution network, and thus better improves the reliability of the power distribution network, reduces line loss, and balances the load. BRIEF DESCRIPTION OF DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative effort on the basis of these drawings.

[0036] Figure 1 The flowchart of the power distribution network topology reconfiguration method provided by the application is shown in the figure.

[0037] Figure 2 The power distribution network topology diagram provided by the application is shown in the figure.

[0038] Figure 3 The structure diagram of the power distribution network topology reconfiguration system provided by the application is shown in the figure.

[0039] Figure 4 The structure diagram of the electronic device provided by the application is shown in the figure. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solutions and advantages of the application clearer, the technical solutions in the application will be described clearly and completely in combination with the drawings in the application. Obviously, the described embodiments are some of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the application.

[0041] Based on the good convergence of the artificial intelligence algorithm, the power distribution network topology is reconstructed by using the deep reinforcement learning algorithm, and the artificial intelligence algorithm is Model free, is driven by data, and does not depend on the establishment of a model, and is very suitable for the problem of power distribution network topology reconstruction. The load margin of the power distribution network is considered, the topology structure of the power distribution network is reconstructed, the load margin refers to the load, the power generation mode, the load growth power that can be borne by the system from the current operating point to the collapse point, generally refers to the active power load margin, can be used as a measurement index of the power system, and the physical meaning is obvious and the linearity is good.

[0042] Figure 1 The flowchart of the power distribution network topology reconstruction method provided by the present application is shown in Figure 1 The present application provides a power distribution network topology reconstruction method, comprising:

[0043] Step 101, according to the node data and branch data of the power distribution network, the topology reconstruction environment of the power distribution network is constructed, wherein the reward of the topology reconstruction environment is the maximum load margin of the power distribution network.

[0044] In the present application, the topology of the power distribution network system is first constructed, which is used for setting the action space, state space and reward function of the subsequent reinforcement learning algorithm, and plays a crucial role in the final effect of the algorithm, therefore, the present application needs to construct the topology reconstruction environment of the power distribution network based on the current power distribution network environment state data, including node data and branch data (also can be the power distribution network environment state data in a preset time period), and can also obtain the original topology of the power distribution network, in the present application, the node data includes the number of load nodes, the load node injection power and the load node load margin; the branch data includes the number of power distribution network branches, the power distribution network branch opening and closing state, and the power distribution network branch power flow data. Figure 2 The power distribution network topology diagram provided by the present application can refer to Figure 2 In the present application, there is only one slack node in the power distribution network topology, that is, the power supply; at the same time, in the power distribution network topology, there are N load nodes and M branches, and the radial topology structure must be ensured in the power distribution network topology to avoid the appearance of loop and cannot have isolated nodes. In addition, the injection power of the load node is given by the topology reconstruction environment, and only the opening and closing state of the branch can be changed based on the deep reinforcement learning algorithm, such as connecting and opening some branches.

[0045] Step 102, based on the deep reinforcement learning algorithm, the topology reconstruction environment is solved, and the power distribution network topology corresponding to the maximum load margin is obtained.

[0046] In the present application, the deep reinforcement learning algorithm is a Dueling-DQN algorithm. The Dueling-DQN algorithm is a commonly used algorithm for solving reinforcement learning problems in discrete action space based on value functions, which is a variant of the DQN algorithm. By decomposing the Q value function into potential energy function V and advantage function A, the information between different action data is shared, which greatly improves the sampling efficiency and training convergence speed. In the present application, the steps for solving the topology reconfiguration environment based on the Dueling-DQN algorithm are as follows:

[0047] Step S1, randomly initialize the current Dueling Q network and the target Dueling Q' network, set the training step number epoch = 1, the maximum training step number E max , the parameter update frequency C, the experience replay number m, and the control sequence length N C ;

[0048] Step S2, obtain a random initial state S0 from the topology reconfiguration environment, and set the control step counter step to 1;

[0049] Step S3, according to the current state S, obtain the Q value through the current Q network, and determine the action A by using the ε-greedy method, and execute the action A (i.e. adjust the branch break of the power distribution network topology), obtain the next state S', the reward R, store (S, S', A, R) in the cache, and step = step + 1;

[0050] Step S4, take out m samples from the cache, calculate the loss function, and update the current Dueling Q network;

[0051] Step S5, if Step % C = 0, update the target Dueling Q' network;

[0052] Step S6, if step = N C , or the control state cannot continue, end the current training step, epoch = epoch + 1, if epoch = E max , the training is completed, and the target Dueling Q' network is used as the final solution for online power distribution network reconfiguration control; otherwise, go to step S2.

[0053] Through the above steps, the algorithm iterates about 6000 steps and basically converges. Compared with the original topology structure, the load margin score based on the deep reinforcement learning algorithm is improved by about 33.6%. This score has no dimension, and can be multiplied by a coefficient to obtain the score, which is scaled to about 10, which is convenient for algorithm convergence.

[0054] In the application, according to branch data and node data of the power distribution network, it is known that there are various branch breaking combinations in the power distribution network environment, specifically, one branch is connected and another branch is disconnected from M branches, or no action, a total of Mx(M-1)+1 different actions; then, the deep reinforcement learning algorithm is used to iteratively solve the topology reconstruction environment data, the data under the current branch breaking combination is taken as the initial value of the branch breaking combination in the next iteration, and the reward function is defined as the sum of the load margin of each load node, when the preset iteration number is met, or the load margin of the power distribution network cannot be increased by replacing the branch breaking combination, the maximum load margin of the power distribution network is obtained, and finally the original topology of the power distribution network is reconstructed according to the branch breaking combination corresponding to the maximum load margin.

[0055] The power distribution network topology reconstruction method provided by the application is based on a deep reinforcement learning algorithm, reconstructs the current topology structure of the power distribution network according to the environment state data of the power distribution network, and thus the reliability of the power distribution network is better improved, line loss is reduced, and load is balanced.

[0056] On the basis of the above-mentioned embodiments, the topology reconstruction environment of the power distribution network is constructed according to the node data and the branch data of the power distribution network, including:

[0057] An action space is constructed according to the number of load nodes and the number of branches.

[0058] A state space is constructed according to the injected power of the load node, the branch breaking state of the power distribution network, the branch power flow data of the power distribution network and the load margin of the load node.

[0059] The maximum load margin of the power distribution network is taken as a reward to construct a topology reconstruction environment of the power distribution network according to the action space and the state space.

[0060] On the basis of the above-mentioned embodiments, the action space is constructed according to the number of load nodes and the number of branches, including:

[0061] The maximum breaking combination number of branches in the power distribution network is obtained according to the number of load nodes and the number of branches, and the formula is:

[0062] Mx(M-1)+1;

[0063] Wherein, M represents the number of branches.

[0064] The action space is constructed according to the maximum breaking combination number.

[0065] In the application, reference can be made to Figure 2As shown, the action space, state space, and setup of the topology reconfiguration environment need to be defined. Specifically, a radial structure with N+1 nodes (N load nodes and 1 relaxed node) must have exactly N connected branches. Therefore, disconnecting one branch necessarily requires connecting another branch, and vice versa, while maintaining the radial structure. Further, a maximum action space is defined, selecting one branch to connect from M branches and disconnecting another branch, for a total of M×(M-1)+1 different actions. The state space calculates the power flow of each branch and the load margin of each load node based on the current load status and topology connectivity of each load node. Therefore, to allow the algorithm to utilize as much useful information as possible, the state space defined in this invention includes: the injected power of each node, the on / off status and power flow values ​​of each branch, and the load margin information of each node. The reward function aims to enable the algorithm to learn to improve the load margin of the power grid. Therefore, by feeding the load margin back to the reward R, the algorithm can be guided to learn to improve the load margin of the power grid.

[0066] R = ∑ Load margin of each load node.

[0067] Based on the above embodiments, the step of solving the topology reconfiguration environment using a deep reinforcement learning algorithm to obtain the distribution network topology corresponding to the maximum load margin includes:

[0068] Based on a deep reinforcement learning algorithm, the topology reconstruction environment is iteratively solved according to each branch disconnection combination in the action space and the corresponding state space to obtain the node injection power and the node voltage value after power injection in each branch disconnection combination.

[0069] Based on the original node voltage value, the injected power of the node, and the node voltage value after power injection, the load margin of the load node in the current iteration process is obtained;

[0070] Based on the load margin of the load nodes, the load margin of the distribution network is obtained. If the preset conditions are met, the iteration stops and the load margin of the distribution network obtained in the current iteration process is taken as the maximum load margin. The distribution network topology is then reconstructed according to the branch disconnection combination corresponding to the maximum load margin.

[0071] In the application, according to the number M of branches of the original topology of the power distribution network and the nodes, there are Mx(M-1)+1 combinations of branch breaking combinations in the power distribution network, then, after setting the preset training condition, the deep reinforcement learning algorithm firstly randomly obtains the initial state S0 of the original topology of the power distribution network, and obtains the current state S after completing the current training step number; then, the e-greedy method is used to determine the action A of the next iteration process from the Mx(M-1)+1 combinations, and the action A is executed (that is, according to the branch breaking combination of the action A, the current branch breaking is adjusted), and the next state S' is obtained. Then, by calculating the voltage value of the next state and the original topology of the power distribution network, it is judged whether the load margin of the power distribution network is improved, if the load margin of the power distribution network in the next state is improved compared with the current state, the next iteration is carried out; if the load margin of the power distribution network does not continue to improve, the iteration is stopped (the maximum iteration number can also be set), the load margin of the power distribution network in the current state is taken as the maximum load margin, and the topology of the power distribution network is reconstructed according to the branch breaking combination at this time.

[0072] On the basis of the above embodiment, the load node load margin in the current iteration process is obtained according to the node original voltage value, the node injected power, and the node voltage value after the power injection, comprising:

[0073] obtaining the original voltage value U of the load node origin ;

[0074] injecting the node injected power Δ to all load nodes in the current branch breaking combination, and obtaining the node voltage value U after the power injection Step ;

[0075] calculating the load margin D of the load node according to the original voltage value U origin and the node voltage value U after the power injection Step , and the formula is:

[0076] D=(U Step -U origin ) / Δ.

[0077] In the application, since the calculation of the load margin needs multiple iterations, the calculation complexity is very high. Therefore, according to the property that the gradient of the load curve is highly related to the load margin, that is, when the slope of the load operating point is small, the load margin is large, and when the slope is large, the load margin is small, the gradient information is used to approximate the load margin information, which greatly reduces the complexity of the algorithm operation.

[0078] Figure 3 The structural schematic diagram of the power distribution network topology reconstruction system provided by the application is as follows, Figure 3As shown, the application provides a power distribution network topology reconfiguration system, comprising a topology environment construction module 301 and a reconfiguration module 302, wherein the topology environment construction module 301 is used to construct a topology reconfiguration environment of the power distribution network according to node data and branch data of the power distribution network, wherein the reward of the topology reconfiguration environment is the maximum load margin of the power distribution network; the reconfiguration module 302 is used to solve the topology reconfiguration environment based on a deep reinforcement learning algorithm, and obtain the power distribution network topology corresponding to the maximum load margin.

[0079] The power distribution network topology reconfiguration system provided by the application is based on a deep reinforcement learning algorithm, and reconfigures the current topology structure of the power distribution network according to environmental state data of the power distribution network, thereby better improving the reliability of the power distribution network, reducing line loss, and balancing load.

[0080] On the basis of the above-mentioned embodiments, the topology environment construction module comprises:

[0081] An action space construction unit is configured to construct an action space according to the number of load nodes and the number of branches.

[0082] A state space construction unit is configured to construct a state space according to the load injection power of the load node, the branch opening state of the power distribution network, the branch power flow data of the power distribution network, and the load margin of the load node.

[0083] The maximum load margin of the power distribution network is taken as a reward to construct a topology reconfiguration environment of the power distribution network according to the action space and the state space.

[0084] On the basis of the above-mentioned embodiments, the action space construction unit comprises:

[0085] A branch opening combination acquisition subunit is configured to acquire the maximum number of branch opening combinations in the power distribution network according to the number of load nodes and the number of branches, and the formula is:

[0086] Mx(M-1)+1;

[0087] Wherein, M represents the number of branches.

[0088] A space construction subunit is configured to construct an action space according to the maximum number of branch opening combinations.

[0089] On the basis of the above-mentioned embodiments, the reconfiguration module comprises:

[0090] A processing unit is configured to iteratively solve the topology reconfiguration environment according to each branch opening combination in the action space and the corresponding state space based on a deep reinforcement learning algorithm, and acquire the node injection power in each branch opening combination and the node voltage value after power injection.

[0091] The load margin calculation unit is used to obtain the load margin of the load node in the current iteration process based on the original node voltage value, the node injected power, and the node voltage value after power injection.

[0092] The judgment unit is used to obtain the distribution network load margin based on the load margin of the load node. If the preset condition is met, the iteration is stopped, and the distribution network load margin obtained in the current iteration process is taken as the maximum load margin. The distribution network topology is then reconstructed according to the branch disconnection combination corresponding to the maximum load margin.

[0093] Based on the above embodiments, the load margin calculation unit includes:

[0094] The first calculation subunit is used to obtain the original voltage value U of the load node. origin ;

[0095] The second calculation subunit is used to inject node power Δ into all load nodes in the current branch disconnection combination and obtain the node voltage value U after power injection. Step ;

[0096] The third calculation subunit is used to calculate based on the original voltage value U. origin and the node voltage value U after power injection Step The load margin D of the load node is calculated using the following formula:

[0097] D=(U Step -U origin ) / Δ.

[0098] The system provided in this embodiment of the invention is used to execute the above-described method embodiments. For specific processes and details, please refer to the above embodiments, which will not be repeated here.

[0099] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include a processor 401, a communication interface 402, a memory 403, and a communication bus 404. The processor 401, communication interface 402, and memory 403 communicate with each other via the communication bus 404. The processor 401 can call logical instructions in the memory 403 to execute a distribution network topology reconfiguration method. This method includes: constructing a distribution network topology reconfiguration environment based on node and branch data of the distribution network, wherein the reward of the topology reconfiguration environment is the maximum load margin of the distribution network; and solving the topology reconfiguration environment based on a deep reinforcement learning algorithm to obtain the distribution network topology corresponding to the maximum load margin.

[0100] Further, the logic instructions in the memory 403 described above can be implemented in the form of software functional units and sold or used as independent products, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the parts that make contributions to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0101] In another aspect, the present application also provides a computer program product, which comprises a computer program stored on a non-transitory computer readable storage medium, and the computer program comprises program instructions, and when the program instructions are executed by a computer, the computer can execute the power distribution network topology reconfiguration method provided by the above method. The method comprises: constructing a topology reconfiguration environment of a power distribution network according to node data and branch data of the power distribution network, wherein the reward of the topology reconfiguration environment is the maximum load margin of the power distribution network; and solving the topology reconfiguration environment based on a deep reinforcement learning algorithm to obtain a power distribution network topology corresponding to the maximum load margin.

[0102] In another aspect, the present application also provides a non-transitory computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the power distribution network topology reconfiguration method provided by the above method. The method comprises: constructing a topology reconfiguration environment of a power distribution network according to node data and branch data of the power distribution network, wherein the reward of the topology reconfiguration environment is the maximum load margin of the power distribution network; and solving the topology reconfiguration environment based on a deep reinforcement learning algorithm to obtain a power distribution network topology corresponding to the maximum load margin.

[0103] The device embodiments described above are only schematic, wherein the units illustrated as separate components can or can not be physically separate, and the components illustrated as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment. Those skilled in the art can understand and implement without creative labor.

[0104] Those skilled in the art can clearly understand the technical solutions of the various embodiments from the above description of the embodiments, and the various embodiments can be implemented by means of software with the necessary general hardware platforms, and of course, can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that makes a contribution, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, and the like, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0105] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features therein; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A power distribution network topology reconfiguration method, characterized by, The application relates to a power distribution network topology reconfiguration method and device. A topology reconfiguration environment of a power distribution network is constructed according to node data and branch data of the power distribution network, wherein the reward of the topology reconfiguration environment is the maximum load margin of the power distribution network; The topology reconfiguration environment is solved based on a deep reinforcement learning algorithm to obtain the power distribution network topology corresponding to the maximum load margin. The node data comprises the number of load nodes, the load node injection power and the load node load margin; the branch data comprises the number of power distribution network branches, the power distribution network branch opening state and the power distribution network branch power flow data; The topology reconfiguration environment of the power distribution network is constructed according to the node data and the branch data of the power distribution network, and the reward of the topology reconfiguration environment is the maximum load margin of the power distribution network. The action space is constructed according to the number of load nodes and the number of branches; The state space is constructed according to the load node injection power, the power distribution network branch opening state, the power distribution network branch power flow data and the load node load margin; The topology reconfiguration environment of the power distribution network is constructed according to the action space and the state space, and the maximum load margin of the power distribution network is taken as the reward.

2. The power distribution network topology reconfiguration method of claim 1, wherein, The action space is constructed according to the number of load nodes and the number of branches. The maximum opening combination number of branches in the power distribution network is obtained according to the number of load nodes and the number of branches, and the formula is: Mx (M-1) +1; Wherein, M represents the number of branches. The action space is constructed according to the maximum opening combination number.

3. The power distribution network topology reconfiguration method of claim 1, wherein, The topology reconfiguration environment is solved based on a deep reinforcement learning algorithm to obtain the power distribution network topology corresponding to the maximum load margin. The topology reconfiguration environment is iteratively solved based on the deep reinforcement learning algorithm according to each branch opening combination in the action space and the corresponding state space to obtain the node injection power in each branch opening combination and the node voltage value after power injection. The load node load margin in the current iteration process is obtained according to the original node voltage value, the node injection power and the node voltage value after power injection. The power distribution network load margin is obtained according to the load node load margin, and if a preset condition is met, the iteration is stopped, the power distribution network load margin obtained in the current iteration process is taken as the maximum load margin, and the power distribution network topology is reconfigured according to the branch opening combination corresponding to the maximum load margin.

4. The power distribution network topology reconfiguration method of claim 3, wherein, The deep reinforcement learning algorithm is a Dueling-DQN algorithm. Obtaining a raw voltage value of a load node ; Injecting node injection power Δ to all load nodes in the current branch breaking combination, and obtaining node voltage values after power injection ; According to the original voltage value and the node voltage value after the power injection , the load margin of the load node is calculated D , and the formula is: 。 5. The power distribution network topology reconfiguration method of claim 1, wherein, The application relates to a power distribution network topology reconfiguration method and device.

6. A power distribution network topology reconfiguration system characterized by, A topology reconfiguration environment of a power distribution network is constructed according to node data and branch data of the power distribution network, wherein the reward of the topology reconfiguration environment is the maximum load margin of the power distribution network; The topology reconfiguration environment is solved based on a deep reinforcement learning algorithm to obtain the power distribution network topology corresponding to the maximum load margin. The node data comprises the number of load nodes, the load node injection power and the load node load margin; the branch data comprises the number of power distribution network branches, the power distribution network branch opening state and the power distribution network branch power flow data; The topology environment construction module comprises: ​ An action space construction unit is configured to construct an action space according to the number of load nodes and the number of branches; A state space construction unit is configured to construct a state space according to the load node injection power, the power distribution network branch open state, the power distribution network branch power flow data, and the load node load margin; A topology reconfiguration environment of the power distribution network is constructed according to the action space and the state space, with the maximum load margin of the power distribution network as a reward.

7. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the steps of the power distribution network topology reconfiguration method according to any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the power distribution network topology reconfiguration method according to any one of claims 1 to 5.

Images