A multi-scenario distribution network reconstruction method and device

Abstract

The present invention discloses a multi-scenario distribution network reconstruction method and device, including: initializing distribution network parameters, scenario parameters, and particle swarm algorithm parameters; the particle swarm algorithm parameters are binary-coded particle swarm algorithm parameters with probability jumps; according to the initialized distribution network parameters, obtaining the distribution function and probability density function of each distributed power output; generating a reconstruction solution for all distribution networks and screening out an effective reconstruction solution; according to the distribution function and probability density function of each distributed power output, as well as the initialized scenario parameters, particle swarm algorithm parameters, and the effective reconstruction solution, using the binary-coded particle swarm algorithm with probability jumps to obtain the optimal reconstruction solution, thereby obtaining a multi-scenario distribution network reconstruction method. The present invention solves the technical problems in the prior art of relatively rough distributed power modeling, simple and difficult-to-switch reconstruction targets, and low solution speed.

Description

Technical Field

[0001] The present invention relates to the technical field of power system distribution network planning and operation, and in particular to a multi-scenario distribution network reconstruction method and device. Background Art

[0002] The distribution network, located at the end of the power system, is a crucial link between the power grid and users. In recent years, with the rapid economic and social development and the rapid expansion of the power system, the distribution network has, on the one hand, carried an increasing amount of load, making its role in the power system increasingly critical; on the other hand, the structure of the distribution network has also become more complex, making its dispatch and control increasingly difficult. Distribution networks are typically designed as closed-loop systems and operated in open-loop configurations, containing a large number of sectionalizing switches and interconnecting switches. By changing the open and closed states of these switches, the topology of the distribution network can be altered, enabling reconstruction of the distribution network and changing the power flow distribution within it. This, in turn, reduces operating costs and network losses, while improving power quality and power supply reliability.

[0003] With the rapid development of new power systems, a large number of distributed generation (DG) sources, primarily wind power and photovoltaics, have been integrated into distribution networks, impacting them in a variety of ways. First, the integration of DG transforms the power flow in distribution networks from unidirectional to bidirectional, changing the magnitude of the flow. This impacts network losses and power quality, which are primarily determined by the flow. Second, DG can directly supply power to some loads in the distribution network, alleviating the pressure on the network. Furthermore, DG can provide some support for voltage regulation, thereby improving the reliability of the distribution network. Third, the output of renewable energy sources is highly random, and their integration may increase the frequency of voltage flickers and surges, leading to the failure or malfunction of protective devices.

[0004] Existing research on distribution network reconfiguration focuses on the following: First, most existing studies only model DGs as generator nodes or constant-power nodes using their average output, failing to consider the randomness of DG output, resulting in relatively crude DG modeling. Second, most existing studies focus solely on reducing network losses and improving power quality as reconfiguration objectives, failing to consider factors such as switching costs and DG configuration costs. Furthermore, these studies lack the flexibility to adapt to different scenarios, resulting in a simplistic and difficult-to-switch reconfiguration objective. Third, most existing studies fail to consider the reproducibility and independence of the computational processes involved in obtaining DG output distribution, determining the validity of the reconfigured solution, determining the validity of DG output samples, calculating the solution's objective function value, updating particle positions, and calculating particle transition probabilities. These studies rely solely on CPU serial implementation, resulting in a relatively slow solution speed. Consequently, the results of reconfiguring new distribution networks containing DGs using existing technologies are less than ideal, making it difficult to obtain effective reconfiguration solutions in a short period of time. Summary of the Invention

[0005] The present invention provides a multi-scenario distribution network reconstruction method and device thereof to solve the technical problems in the prior art such as relatively rough DG modeling, simple and difficult-to-switch reconstruction targets, and low solution speed.

[0006] In order to solve the above technical problems, an embodiment of the present invention provides a multi-scenario distribution network reconstruction method, including:

[0007] Initialize distribution network parameters, scenario parameters and particle swarm algorithm parameters; the particle swarm algorithm parameters are binary coded particle swarm algorithm parameters with probability jumps;

[0008] According to the initialized distribution network parameters, the distribution function and probability density function of each distributed power output are obtained;

[0009] Generate reconstruction solutions for the entire distribution network and select effective reconstruction solutions;

[0010] According to the distribution function and probability density function of each distributed power output, as well as the initialized scenario parameters, particle swarm algorithm parameters and the effective reconstruction solution, a binary coded particle swarm algorithm with probability jump is used to obtain the optimal reconstruction solution, and then a multi-scenario distribution network reconstruction method is obtained.

[0011] As a preferred solution, the initialization of distribution network parameters, scenario parameters and particle swarm algorithm parameters is specifically as follows:

[0012] Initialize the input distribution network parameters; the distribution network parameters include: the number, rated voltage and shunt admittance of each node, the number, connected nodes and series impedance of each branch, the branch location and status of each switch, the branch location, active power and reactive power of each load, the type, device parameters, node location and historical operation data set of each distributed power source;

[0013] Initializing the input scene parameters; the scene parameters include weight parameters and cost parameters;

[0014] Initialize the input particle swarm algorithm parameters; the particle swarm algorithm parameters include population size, maximum number of iterations and learning factor.

[0015] As a preferred solution, the distribution function and probability density function of each distributed power output are obtained according to the initialized distribution network parameters, specifically:

[0016] Based on the initialized distribution network parameters, a wind speed distribution model is formed. Combined with the constructed piecewise cubic model of the wind farm, the Monte Carlo method is used for random sampling to fit the sampling results and obtain the distribution function of the wind farm output. The numerical differentiation method is then used to obtain the probability density function of the wind farm output.

[0017] According to the initialized distribution network parameters, a light intensity distribution model is formed. Combined with the constructed piecewise quadratic model of the photovoltaic power station, the Monte Carlo method is used for random sampling to fit the sampling results and obtain the distribution function of the photovoltaic power station output. The numerical differentiation method is used to obtain the probability density function of the photovoltaic power station output.

[0018] As a preferred solution, the generation of reconstruction solutions for all distribution networks and screening out effective reconstruction solutions are specifically as follows:

[0019] According to the closed states of all section switches and tie switches of the distribution network, a distribution network topology with all switches closed and containing loops is obtained, and according to the distribution network topology, all meshes and branches under the distribution network topology are obtained;

[0020] Disconnect one branch in each mesh, so that after each branch is disconnected, a reconstruction solution of the distribution network is generated, and then all reconstruction solutions are generated;

[0021] The effectiveness of each reconstruction solution is judged through GPU parallelization, and all valid reconstruction solutions are screened out.

[0022] As a preferred solution, the validity of each reconstruction solution is judged by GPU parallelization to screen out all valid reconstruction solutions, specifically:

[0023] The validity of each reconstruction solution is judged through GPU parallelization. If the reconstruction solution is a radial topology, the number of lines put into operation is equal to the difference between the number of nodes and the number of connected subgraphs. If the reconstruction solution does not contain islands, all connected subgraphs contain power supplies. In this way, each time the reconstruction solution is judged by the GPU, the root node is searched downward to the end node, and from the end node, the first-level nodes are returned upward one by one to determine whether there is an unvisited path back to the first-level node. If so, the unvisited path is followed to the end node of the unvisited path. If not, the unvisited path is returned to the previous-level node to rewrite the unvisited path until the root node is returned, thereby screening out all valid reconstruction solutions.

[0024] As a preferred solution, according to the distribution function and probability density function of each distributed power output, as well as the initialized scene parameters, particle swarm algorithm parameters and the effective reconstruction solution, a binary coded particle swarm algorithm with probability jump is used to obtain the optimal reconstruction solution, specifically:

[0025] According to the status of each switch in the distribution network, each valid reconstruction solution is binary-encoded to form the corresponding binary position vector and binary velocity vector;

[0026] According to the binary position vector and binary velocity vector of the effective reconstruction solution, the initial position of each particle in the particle swarm is determined, and the initial velocity of each particle in the particle swarm is randomly generated;

[0027] Iteration is started through GPU parallel computing, the objective function value of the solution corresponding to the position of each particle in the particle swarm during the iteration is calculated, and the historical optimal position of each particle in the particle swarm is updated; wherein the optimality condition is that the objective function value of the solution corresponding to the position of the particle is minimized;

[0028] The probability of each particle in the particle swarm jumping to its historical optimal position in each iteration is calculated in parallel by GPU;

[0029] The roulette wheel method is used to make each particle in the particle swarm in the current iteration jump to its own historical optimal position according to the jump probability, until the first particle that successfully jumps is generated, and its position after the successful jump is recorded;

[0030] Update the position of each particle in the current iteration in parallel through the GPU;

[0031] If the number of iterations reaches the preset maximum value, the objective function value of the solution corresponding to the position of each particle in the particle swarm in the iteration is calculated in parallel by GPU, and the solution with the lowest objective function value is taken as the global optimal solution, thus stopping the iteration and outputting the reconstruction result; otherwise, another iterative calculation is performed.

[0032] As a preferred solution, the objective function value of the solution corresponding to the position of each particle in the particle swarm during iteration is calculated as follows:

[0033] The node admittance matrix of the solution is established by the branch addition method;

[0034] Based on the distribution function and probability density function of each distributed power output, several distributed power output samples are obtained;

[0035] Screen out effective distributed power output samples;

[0036] Based on the topology and node admittance matrix of the solution, the distribution network flow under each distributed power output sample is solved, and the total cost function under each distributed power output sample is calculated. Then, the average value of the total cost function under each distributed power output sample is calculated, and the average value is used as the objective function value of the solution.

[0037] As a preferred solution, the total cost function under each distributed power output sample includes: network loss cost, voltage offset cost, switch fee and photovoltaic power station configuration cost.

[0038] As a preferred solution, the position of each particle of the particle swarm in the current iteration is updated in parallel by the GPU, specifically:

[0039] Through GPU parallel calculation, the position of each particle in the particle swarm in the kth iteration is updated. The particle update formula is:

[0040]

[0041] Among them, round means that each component is rounded to the nearest integer; X i (k), X i (k+1) are the positions of particle i before and after the kth iteration, V i (k), V i (k+1) are the velocities of particle i before and after the kth iteration respectively; P i (k) is the historical optimal position of particle i at the kth iteration, X g (k) is the position of the optimal particle in the particle swarm at the kth iteration; c1 and c2 are learning factors, and 0 <c1<c2<2、 l=c1+c2; r1 and r2 are random numbers between 0 and 1.

[0042] Accordingly, the present invention also provides a multi-scenario distribution network reconstruction device, comprising: an initialization module, a function module, an effective reconstruction module and an optimal reconstruction module;

[0043] The initialization module is used to initialize the distribution network parameters, scene parameters and particle swarm algorithm parameters; the particle swarm algorithm parameters are binary coded particle swarm algorithm parameters with probability jumps;

[0044] The function module is used to obtain the distribution function and probability density function of each distributed power output according to the initialized distribution network parameters;

[0045] The effective reconstruction module is used to generate reconstruction solutions for the entire distribution network and screen out effective reconstruction solutions;

[0046] The optimal reconstruction module is used to obtain the optimal reconstruction solution using a binary coded particle swarm algorithm with probability jumps based on the distribution function and probability density function of the output of each distributed power source, as well as the initialized scenario parameters, particle swarm algorithm parameters and the effective reconstruction solution, thereby obtaining a multi-scenario distribution network reconstruction method.

[0047] As a preferred solution, the initialization of distribution network parameters, scenario parameters and particle swarm algorithm parameters is specifically as follows:

[0048] Initialize the input distribution network parameters; the distribution network parameters include: the number, rated voltage and shunt admittance of each node, the number, connected nodes and series impedance of each branch, the branch location and status of each switch, the branch location, active power and reactive power of each load, the type, device parameters, node location and historical operation data set of each distributed power source;

[0049] Initializing the input scene parameters; the scene parameters include weight parameters and cost parameters;

[0050] Initialize the input particle swarm algorithm parameters; the particle swarm algorithm parameters include population size, maximum number of iterations and learning factor.

[0051] As a preferred solution, the distribution function and probability density function of each distributed power output are obtained according to the initialized distribution network parameters, specifically:

[0052] Based on the initialized distribution network parameters, a wind speed distribution model is formed. Combined with the constructed piecewise cubic model of the wind farm, the Monte Carlo method is used for random sampling to fit the sampling results and obtain the distribution function of the wind farm output. The numerical differentiation method is then used to obtain the probability density function of the wind farm output.

[0053] According to the initialized distribution network parameters, a light intensity distribution model is formed. Combined with the constructed piecewise quadratic model of the photovoltaic power station, the Monte Carlo method is used for random sampling to fit the sampling results and obtain the distribution function of the photovoltaic power station output. The numerical differentiation method is used to obtain the probability density function of the photovoltaic power station output.

[0054] As a preferred solution, the generation of reconstruction solutions for all distribution networks and screening out effective reconstruction solutions are specifically as follows:

[0055] According to the closed states of all section switches and tie switches of the distribution network, a distribution network topology with all switches closed and containing loops is obtained, and according to the distribution network topology, all meshes and branches under the distribution network topology are obtained;

[0056] Disconnect one branch in each mesh, so that after each branch is disconnected, a reconstruction solution of the distribution network is generated, and then all reconstruction solutions are generated;

[0057] The effectiveness of each reconstruction solution is judged through GPU parallelization, and all valid reconstruction solutions are screened out.

[0058] As a preferred solution, the validity of each reconstruction solution is judged by GPU parallelization to screen out all valid reconstruction solutions, specifically:

[0059] The validity of each reconstruction solution is judged through GPU parallelization. If the reconstruction solution is a radial topology, the number of lines put into operation is equal to the difference between the number of nodes and the number of connected subgraphs. If the reconstruction solution does not contain islands, all connected subgraphs contain power supplies. In this way, each time the reconstruction solution is judged by the GPU, the root node is searched downward to the end node, and from the end node, the first-level nodes are returned upward one by one to determine whether there is an unvisited path back to the first-level node. If so, the unvisited path is followed to the end node of the unvisited path. If not, the unvisited path is returned to the previous-level node to rewrite the unvisited path until the root node is returned, thereby screening out all valid reconstruction solutions.

[0060] As a preferred solution, according to the distribution function and probability density function of each distributed power output, as well as the initialized scene parameters, particle swarm algorithm parameters and the effective reconstruction solution, a binary coded particle swarm algorithm with probability jump is used to obtain the optimal reconstruction solution, specifically:

[0061] According to the status of each switch in the distribution network, each valid reconstruction solution is binary-encoded to form the corresponding binary position vector and binary velocity vector;

[0062] According to the binary position vector and binary velocity vector of the effective reconstruction solution, the initial position of each particle in the particle swarm is determined, and the initial velocity of each particle in the particle swarm is randomly generated;

[0063] Iteration is started through GPU parallel computing, the objective function value of the solution corresponding to the position of each particle in the particle swarm during the iteration is calculated, and the historical optimal position of each particle in the particle swarm is updated; wherein the optimality condition is that the objective function value of the solution corresponding to the position of the particle is minimized;

[0064] The probability of each particle in the particle swarm jumping to its historical optimal position in each iteration is calculated in parallel by GPU;

[0065] The roulette wheel method is used to make each particle in the particle swarm in the current iteration jump to its own historical optimal position according to the jump probability, until the first particle that successfully jumps is generated, and its position after the successful jump is recorded;

[0066] Update the position of each particle in the current iteration in parallel through the GPU;

[0067] If the number of iterations reaches the preset maximum value, the objective function value of the solution corresponding to the position of each particle in the particle swarm in the iteration is calculated in parallel by GPU, and the solution with the lowest objective function value is taken as the global optimal solution, thus stopping the iteration and outputting the reconstruction result; otherwise, another iterative calculation is performed.

[0068] As a preferred solution, the objective function value of the solution corresponding to the position of each particle in the particle swarm during iteration is calculated as follows:

[0069] The node admittance matrix of the solution is established by the branch addition method;

[0070] Based on the distribution function and probability density function of each distributed power output, several distributed power output samples are obtained;

[0071] Screen out effective distributed power output samples;

[0072] Based on the topology and node admittance matrix of the solution, the distribution network flow under each distributed power output sample is solved, and the total cost function under each distributed power output sample is calculated. Then, the average value of the total cost function under each distributed power output sample is calculated, and the average value is used as the objective function value of the solution.

[0073] As a preferred solution, the total cost function under each distributed power output sample includes: network loss cost, voltage offset cost, switch fee and photovoltaic power station configuration cost.

[0074] As a preferred solution, the position of each particle of the particle swarm in the current iteration is updated in parallel by the GPU, specifically:

[0075] Through GPU parallel calculation, the position of each particle in the particle swarm in the kth iteration is updated. The particle update formula is:

[0076]

[0077] Among them, round means that each component is rounded to the nearest integer; X i (k), X i (k+1) are the positions of particle i before and after the kth iteration, V i (k), V i (k+1) are the velocities of particle i before and after the kth iteration respectively; P i (k) is the historical optimal position of particle i at the kth iteration, X g (k) is the position of the optimal particle in the particle swarm at the kth iteration; c1 and c2 are learning factors, and 0 <c1<c2<2、 l=c1+c2; r1 and r2 are random numbers between 0 and 1.

[0078] Accordingly, the present invention also provides a terminal device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the multi-scenario distribution network reconstruction method as described in any one of the above items when executing the computer program.

[0079] Accordingly, the present invention also provides a computer-readable storage medium, which includes a stored computer program; wherein, when the computer program is running, it controls the device where the computer-readable storage medium is located to execute the multi-scenario distribution network reconstruction method as described in any one of the above items.

[0080] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0081] The technical solution of the present invention obtains the distribution function and probability density function of each distributed power output by initializing the distribution network parameters, scenario parameters and particle swarm algorithm parameters, thereby taking into account the optimization requirements under multiple scenarios and avoiding the problems of relatively rough modeling, simple reconstruction targets and difficulty in switching in the existing technology. The effective reconstruction solution is screened out by generating the reconstruction solution of the entire distribution network, and then the binary coded particle swarm algorithm with probability jump is used to obtain the optimal reconstruction solution to obtain and complete the reconstruction of the distribution network under multiple scenarios, thereby improving the solution efficiency and fully considering the repeatability of the steps of obtaining DG output distribution, screening effective reconstruction solutions, screening effective distributed power output samples, calculating the fitness function of the effective solution, updating the position of particles, and calculating the jump probability of particles. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 : A flowchart of the steps of a multi-scenario distribution network reconstruction method provided by an embodiment of the present invention;

[0083] Figure 2 : A structural diagram of a multi-scenario distribution network reconstruction device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0084] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0085] Example 1

[0086] Please refer to Figure 1 , an embodiment of the present invention provides a multi-scenario distribution network reconstruction method, including the following steps S101-S104:

[0087] Step S101: Initialize distribution network parameters, scenario parameters and particle swarm algorithm parameters; the particle swarm algorithm parameters are binary-coded particle swarm algorithm parameters with probability jumps.

[0088] As a preferred solution of this embodiment, the initialization of distribution network parameters, scenario parameters and particle swarm algorithm parameters is specifically as follows:

[0089] Initialize the input distribution network parameters; the distribution network parameters include: the number, rated voltage and parallel admittance of each node, the number, connected nodes and series impedance of each branch, the branch location and status of each switch, the branch location, active power and reactive power of each load, the type, device parameters, node location and historical operation data set of each distributed power source; initialize the input scenario parameters; the scenario parameters include weight parameters and cost parameters; initialize the input particle swarm algorithm parameters; the particle swarm algorithm parameters include population size, maximum number of iterations and learning factor.

[0090] It should be noted that the input distribution network parameters include: the number, rated voltage, and parallel admittance of each node; the number, connected node, and series impedance of each branch; the branch and status of each switch; the node, active power, and reactive power of each load; and the type, equipment parameters, node, and historical operation data set of each DG.

[0091] Furthermore, the scene parameters are input, including the weight parameter w LOSS 、w ΔU 、w switch 、w DG and cost parameter c loss 、c ΔU 、c switch 、c WT1 、c WT2 、c PV1 、c PV2 ; Input binary coded particle swarm algorithm parameters with probability jump, including: population size N0, maximum number of iterations k max , learning factors c1, c2(0 <c1<c2<2)。

[0092] Step S102: Obtain the distribution function and probability density function of each distributed power source output according to the initialized distribution network parameters.

[0093] As a preferred solution of this embodiment, the distribution function and probability density function of each distributed power output are obtained according to the initialized distribution network parameters, specifically:

[0094] According to the initialized distribution network parameters, a wind speed distribution model is formed, and in conjunction with the constructed piecewise cubic model of the wind farm, the Monte Carlo method is used for random sampling, so as to fit the sampling results to obtain the distribution function of the wind farm output, and the numerical differentiation method is used to obtain the wind farm output probability density function; according to the initialized distribution network parameters, a light intensity distribution model is formed, and in conjunction with the constructed piecewise quadratic model of the photovoltaic power station, the Monte Carlo method is used for random sampling, so as to fit the sampling results to obtain the distribution function of the photovoltaic power station output, and the numerical differentiation method is used to obtain the probability density function of the photovoltaic power station output.

[0095] It should be noted that, in this embodiment, for each wind farm, the mean wind speed v is first calculated based on the historical operation data set. u and standard deviation v σ , forming the Gumbel distribution model of wind speed:

[0096]

[0097] Where v is the actual wind speed and f(v) is the probability density function of wind speed.

[0098] Combined with the piecewise cubic model of the wind farm:

[0099]

[0100] Where, v ci is the cut-in wind speed, v N is the rated wind speed, v co is the cut-out wind speed, P WTN is the rated output of the wind farm, which is an equipment parameter; P W The output of the wind farm.

[0101] The Monte Carlo method is used to perform random sampling and fit the sampling results to obtain the distribution function F(P WT ), the probability density function f(P WT ).

[0102] Furthermore, in this embodiment, for each photovoltaic power station, the average value of the light intensity θ is first calculated based on the historical operation data set. u and standard deviation θ σ , forming a Beta distribution model of light intensity:

[0103]

[0104] Where θ is the light intensity, f(θ) is the probability density function of light intensity.

[0105] Combined with the segmented quadratic model of the photovoltaic power station:

[0106]

[0107] Where θ GR is the local average light intensity, obtained from the historical operation data set; θ s is the turning point of light intensity, P PVN is the rated output of the photovoltaic power station, which is an equipment parameter; P PV The output of the photovoltaic power station.

[0108] The Monte Carlo method is used to perform random sampling and fit the sampling results to obtain the distribution function F(P PV ), the probability density function f(P PV ).

[0109] Step S103: Generate reconstruction solutions for all distribution networks and select valid reconstruction solutions.

[0110] As a preferred solution of this embodiment, the generation of reconstruction solutions for all distribution networks and screening out effective reconstruction solutions are specifically as follows:

[0111] Based on the closed states of all sectional switches and tie switches in the distribution network, a distribution network topology with all switches closed and containing loops is obtained. Based on the distribution network topology, all meshes and branches under the distribution network topology are obtained. One branch in each mesh is disconnected, so that after each branch is disconnected, a reconstruction solution of the distribution network is generated, and then all reconstruction solutions are generated. The validity of each reconstruction solution is judged through GPU parallelization, and all valid reconstruction solutions are screened out.

[0112] In this embodiment, all sectional switches and tie switches of the distribution network are closed to form a distribution network topology with all switches closed and containing loops. All meshes and their constituent branches in this topology are then identified. One branch in each mesh is then disconnected to generate a distribution network reconstruction solution. The disconnected branches are then modified one by one to generate a complete distribution network reconstruction solution. GPU parallelization is then used to determine the validity of each reconstruction solution.

[0113] As a preferred solution of this embodiment, the validity of each reconstruction solution is judged by GPU parallelization to screen out all valid reconstruction solutions, specifically:

[0114] The validity of each reconstruction solution is judged through GPU parallelization. If the reconstruction solution is a radial topology, the number of lines put into operation is equal to the difference between the number of nodes and the number of connected subgraphs. If the reconstruction solution does not contain islands, all connected subgraphs contain power supplies. In this way, each time the reconstruction solution is judged by the GPU, the root node is searched downward to the end node, and from the end node, the first-level nodes are returned upward one by one to determine whether there is an unvisited path back to the first-level node. If so, the unvisited path is followed to the end node of the unvisited path. If not, the unvisited path is returned to the previous-level node to rewrite the unvisited path until the root node is returned, thereby screening out all valid reconstruction solutions.

[0115] In this embodiment, GPU parallelization is used to determine the validity of each reconstruction solution. The judgment criteria are: the reconstruction solution is a radial topology, that is, the number of lines put into operation is equal to the difference between the number of nodes and the number of connected subgraphs; the reconstruction solution does not contain isolated islands, that is, all connected subgraphs contain power supplies.

[0116] During the judgment process, the GPU is used to perform parallel judgment on each reconstruction solution. Each judgment uses a depth-first search algorithm to traverse the topology. The specific process is as follows:

[0117] Step 1: Starting from the root node, search downward along a path until you reach the end node where you cannot go any further.

[0118] Step 2: Return to the parent node and determine whether there is an unvisited path to the parent node. If so, follow another unvisited path until reaching the end node. If not, return to the parent node and repeat Step 2 until returning to the root node.

[0119] At this point, all valid reconstruction solutions can be screened out.

[0120] Step S104: Based on the distribution function and probability density function of the output of each distributed power source, as well as the initialized scenario parameters, particle swarm algorithm parameters and the effective reconstruction solution, a binary coded particle swarm algorithm with probability jump is used to obtain the optimal reconstruction solution, thereby obtaining a multi-scenario distribution network reconstruction method.

[0121] As a preferred solution of this embodiment, according to the distribution function and probability density function of each distributed power output, as well as the initialized scene parameters, particle swarm algorithm parameters and the effective reconstruction solution, a binary coded particle swarm algorithm with probability jump is used to obtain the optimal reconstruction solution, specifically:

[0122] According to the status of each switch in the distribution network, each effective reconstruction solution is binary-encoded to form a corresponding binary position vector and binary velocity vector; based on the binary position vector and binary velocity vector of the effective reconstruction solution, the initial position of each particle in the particle swarm is determined, and the initial velocity of each particle in the particle swarm is randomly generated; the iteration is started through GPU parallel computing, the objective function value of the solution corresponding to the position of each particle in the particle swarm in the iteration is calculated, and the historical optimal position of each particle in the particle swarm is updated; wherein the optimality condition is that the objective function value of the solution corresponding to the position of the particle is the minimum; each iteration is calculated through GPU parallel computing The transition probability of each particle in the particle swarm in the current iteration to its historical optimal position is calculated; the roulette method is used to make each particle in the particle swarm in the current iteration transition to its respective historical optimal position according to the transition probability, until the first particle that successfully transitions is generated, and its position after the successful transition is recorded; the position of each particle in the particle swarm in the current iteration is updated in parallel through the GPU; if the number of iterations reaches the preset maximum value, the objective function value of the solution corresponding to the position of each particle in the particle swarm in the iteration is calculated in parallel through the GPU, and the solution with the lowest objective function value is taken as the global optimal solution, thereby stopping the iteration and outputting the reconstruction result; otherwise, another iterative calculation is performed.

[0123] In this embodiment, each valid reconstructed solution is binary-coded according to the state of each switch therein to form its binary position vector. The binary position vector of solution S is denoted as X, and the binary velocity vector is denoted as V. Thus, the initial position of each particle in the particle swarm {X1(1),X2(1),…,XN0(1)} is determined, and the initial velocity of each particle in the particle swarm is randomly generated. Set the number of iterations k = 1. Use GPUs to parallelly calculate the objective function value of the solution corresponding to the position of each particle in the particle swarm in the kth iteration; use GPUs to parallelly update the historical optimal position of each particle in the particle swarm. The optimality condition is that the objective function value of the solution corresponding to this position is the minimum. Let the historical optimal position of particle i in the particle swarm in the kth iteration be P i (k).

[0124] Select the optimal particle of the particle swarm in the kth iteration. The optimality condition is that the objective function value of the solution corresponding to the position of the particle is the smallest. Let the position X of the optimal particle g(k) of the particle swarm in the kth iteration be g (k) The corresponding solution is S g (k), which is the local optimal solution of the kth iteration. GPU is used to parallelly calculate the transition probability of each particle in the particle swarm in the kth iteration to its historical optimal position. The transition probability of particle i is:

[0125]

[0126] The roulette method is used to make each particle in the particle swarm in the kth iteration jump to its own historical optimal position according to the jump probability, until the first particle that successfully jumps is generated, and its position after the jump is recorded as X g '(k).

[0127] As a preferred solution of this embodiment, the position of each particle of the particle swarm in the current iteration is updated in parallel by the GPU, specifically:

[0128] Through GPU parallel calculation, the position of each particle in the particle swarm in the kth iteration is updated. The particle update formula is:

[0129]

[0130] Among them, round means that each component is rounded to the nearest integer; X i (k), X i (k+1) are the positions of particle i before and after the kth iteration, V i (k), V i (k+1) are the velocities of particle i before and after the kth iteration respectively; P i (k) is the historical optimal position of particle i at the kth iteration, X g (k) is the position of the optimal particle in the particle swarm at the kth iteration; c1 and c2 are learning factors, and 0 <c1<c2<2、 l=c1+c2; r1 and r2 are random numbers between 0 and 1.

[0131] If k = k max By using GPU parallel computing, the current k=k max The objective function value of the solution corresponding to the position of each particle in the particle swarm in the iteration is used to calculate the objective function value of the solution corresponding to the position of each particle in the particle swarm at this time. The solution with the lowest objective function value is taken as the global optimal solution, and the iteration is stopped and the reconstruction result is output; otherwise, let k = k + 1, and use the GPU to parallelly calculate the objective function value of the solution corresponding to the position of each particle in the particle swarm in the current k = k + 1 iteration, and continue the iteration.

[0132] As a preferred solution of this embodiment, the objective function value of the solution corresponding to the position of each particle in the particle swarm during iteration is calculated as follows:

[0133] The node admittance matrix of the solution is established through the branch addition method; based on the distribution function and probability density function of each distributed power output, several distributed power output samples are obtained; effective distributed power output samples are screened out; based on the topology and node admittance matrix of the solution, the distribution network flow under each distributed power output sample is solved, and the total cost function under each distributed power output sample is calculated, and then the average value of the total cost function under each distributed power output sample is calculated, and the average value is used as the objective function value of the solution.

[0134] It should be noted that, in this embodiment, the position X of particle i i (k) The corresponding solution S i The objective function value C(S i The calculation steps of (k)) are as follows:

[0135] Use the branch addition method to establish the solution S i (k) node admittance matrix; Based on the distribution function and probability density function of DG output, use Monte Carlo sampling to obtain several DG output samples; Use GPU parallel screening of valid DG output samples. If there are samples whose DG output exceeds the allowable range, these samples are eliminated, and the process returns to the step of obtaining several DG output samples using Monte Carlo sampling based on the distribution function and probability density function of DG output again, until the number of valid samples is not less than the set value; Based on the solution S i (k) topology and node admittance matrix, and use GPU parallelism to solve the distribution network power flow under each DG output sample. Considering that the solution has a radial topology, the forward-backward substitution method is used for power flow calculation; the total cost function under each DG output sample is calculated using GPU parallelism, and the calculation formula is:

[0136] C=w loss C loss +w ΔU C ΔU +w switch C switch +w DG (C WT +C PV )

[0137] Where w loss 、w ΔU 、w switch 、w DG are the weight parameters of network loss cost, voltage offset cost, switch cost, and DG configuration cost, respectively. By adjusting the above weight parameters, the objective function of distribution network reconstruction can be changed to meet the reconstruction requirements in different scenarios. loss 、C ΔU 、C switch 、C WT 、C PVThey are network loss cost, voltage deviation cost, switch fee cost, wind farm configuration cost, and photovoltaic power station configuration cost.

[0138] As a preferred solution of this embodiment, the total cost function under each distributed power output sample includes: network loss cost, voltage offset cost, switch fee and photovoltaic power station configuration cost.

[0139] It should be noted that the network loss cost C loss :

[0140]

[0141] Where c loss is the unit electricity price of the distribution network, T max is the operating time of the distribution network, is the total network loss power, where r i 、P i , Q i 、U i are the resistance of branch i, the active power injected at the head end, the reactive power injected at the head end, the voltage at the head end, and k Li is the state 0-1 variable of branch i, 0 is open and 1 is closed.

[0142] Voltage offset cost C ΔU :

[0143]

[0144] Where c ΔU is the equivalent electricity price of voltage deviation, is the sum of the squares of the node voltage offset rates, where U i 、U iN are the actual voltage and rated voltage of node i respectively.

[0145] Switching fee C switch :

[0146]

[0147] Where c switch is the cost of a single switch action, is the number of switch operations, where k Soi 、k Sni are the state 0-1 variables of switch i before and after reconstruction, 0 means open and 1 means closed.

[0148] Wind farm configuration cost C WT :

[0149]

[0150] Where cWT1 、c WT2 are the grid connection cost and construction cost per unit capacity of wind farm, P WTi 、P WTiN are the actual output and rated output of wind farm i respectively.

[0151] Photovoltaic power station configuration cost C PV :

[0152]

[0153] Where c PV1 、c PV2 are the grid connection cost and construction cost per unit capacity of photovoltaic power station, P PVi 、P PViN are the actual output and rated output of photovoltaic power station i respectively.

[0154] In this embodiment, based on the historical operating dataset of the DG location, the Gumbel wind speed distribution model and the wind turbine segmented cubic model are used to characterize the wind farm output, and the Beta light intensity distribution model and the photovoltaic generator segmented quadratic model are used to characterize the photovoltaic power station output. A DG model that considers output randomness is established. At the same time, by comprehensively considering traditional factors such as network loss and power quality and practical factors such as switching costs and DG configuration costs, the voltage offset indicator is used to measure power quality, and the switching cost is measured by the number of switching times. After normalization, a weighted comprehensive objective function is formed with weight coefficients that can be adjusted based on scenario requirements. Furthermore, the repeatability of the steps of obtaining the DG output distribution, screening effective reconstruction solutions, screening effective DG output samples, calculating the fitness function of effective solutions, updating particle positions, and calculating the particle transition probability is considered, and these steps are implemented in parallel using GPUs.

[0155] It can be understood that, in terms of DG modeling, this embodiment fully considers the randomness of DG output, uses statistical methods to establish a DG model that considers the randomness of output, and through the optimization of the objective function, fully considers the optimization requirements under multiple scenarios, and forms a weighted comprehensive objective function that can adjust the weight coefficient based on the scenario requirements. Secondly, in terms of algorithm implementation, it fully considers the repeatability of steps such as obtaining DG output distribution, screening effective reconstruction solutions, screening effective DG output samples, calculating the fitness function of the effective solution, updating the position of the particles, and calculating the transition probability of the particles. GPU parallelization is used to implement the solution steps in this embodiment, which effectively improves the solution speed of the algorithm.

[0156] The implementation of the above embodiment has the following effects:

[0157] Compared with the prior art, the embodiments of the present invention initialize the distribution network parameters, scenario parameters, and particle swarm algorithm parameters to obtain the distribution function and probability density function of each distributed power output, thereby taking into account the optimization requirements under multiple scenarios and avoiding the problems of relatively rough modeling, simple reconstruction targets, and difficulty in switching in the prior art. Effective reconstruction solutions are screened out by generating reconstruction solutions of the entire distribution network, and then a binary coded particle swarm algorithm with probability jumps is used to obtain the optimal reconstruction solution to obtain and complete the reconstruction of the distribution network under multiple scenarios, thereby improving the solution efficiency and fully considering the repeatability of steps such as obtaining DG output distribution, screening effective reconstruction solutions, screening effective distributed power output samples, calculating the fitness function of the effective solution, updating the position of particles, and calculating the jump probability of particles.

[0158] Example 2

[0159] See also Figure 2 , which is a multi-scenario distribution network reconstruction device provided by the present invention, including: an initialization module 201, a function module 202, an effective reconstruction module 203 and an optimal reconstruction module 204.

[0160] The initialization module 201 is used to initialize the distribution network parameters, scenario parameters and particle swarm algorithm parameters; the particle swarm algorithm parameters are binary-coded particle swarm algorithm parameters with probability jumps.

[0161] The function module 202 is used to obtain the distribution function and probability density function of each distributed power source output according to the initialized distribution network parameters.

[0162] The effective reconstruction module 203 is used to generate reconstruction solutions for the entire distribution network and screen out effective reconstruction solutions.

[0163] The optimal reconstruction module 204 is used to obtain the optimal reconstruction solution using a binary coded particle swarm algorithm with probability jumps based on the distribution function and probability density function of each distributed power output, as well as the initialized scenario parameters, particle swarm algorithm parameters and the effective reconstruction solution, thereby obtaining a multi-scenario distribution network reconstruction method.

[0164] As a preferred solution of this embodiment, the initialization of distribution network parameters, scenario parameters and particle swarm algorithm parameters is specifically as follows:

[0165] Initialize the input distribution network parameters; the distribution network parameters include: the number, rated voltage and parallel admittance of each node, the number, connected nodes and series impedance of each branch, the branch location and status of each switch, the branch location, active power and reactive power of each load, the type, device parameters, node location and historical operation data set of each distributed power source; initialize the input scenario parameters; the scenario parameters include weight parameters and cost parameters; initialize the input particle swarm algorithm parameters; the particle swarm algorithm parameters include population size, maximum number of iterations and learning factor.

[0166] As a preferred solution of this embodiment, the distribution function and probability density function of each distributed power output are obtained according to the initialized distribution network parameters, specifically:

[0167] According to the initialized distribution network parameters, a wind speed distribution model is formed, and in conjunction with the constructed piecewise cubic model of the wind farm, the Monte Carlo method is used for random sampling, so as to fit the sampling results to obtain the distribution function of the wind farm output, and the numerical differentiation method is used to obtain the wind farm output probability density function; according to the initialized distribution network parameters, a light intensity distribution model is formed, and in conjunction with the constructed piecewise quadratic model of the photovoltaic power station, the Monte Carlo method is used for random sampling, so as to fit the sampling results to obtain the distribution function of the photovoltaic power station output, and the numerical differentiation method is used to obtain the probability density function of the photovoltaic power station output.

[0168] As a preferred solution of this embodiment, the generation of reconstruction solutions for all distribution networks and screening out effective reconstruction solutions are specifically as follows:

[0169] Based on the closed states of all sectional switches and tie switches in the distribution network, a distribution network topology with all switches closed and containing loops is obtained. Based on the distribution network topology, all meshes and branches under the distribution network topology are obtained. One branch in each mesh is disconnected, so that after each branch is disconnected, a reconstruction solution of the distribution network is generated, and then all reconstruction solutions are generated. The validity of each reconstruction solution is judged through GPU parallelization, and all valid reconstruction solutions are screened out.

[0170] As a preferred solution of this embodiment, the validity of each reconstruction solution is judged by GPU parallelization to screen out all valid reconstruction solutions, specifically:

[0171] The validity of each reconstruction solution is judged through GPU parallelization. If the reconstruction solution is a radial topology, the number of lines put into operation is equal to the difference between the number of nodes and the number of connected subgraphs. If the reconstruction solution does not contain islands, all connected subgraphs contain power supplies. In this way, each time the reconstruction solution is judged by the GPU, the root node is searched downward to the end node, and from the end node, the first-level nodes are returned upward one by one to determine whether there is an unvisited path back to the first-level node. If so, the unvisited path is followed to the end node of the unvisited path. If not, the unvisited path is returned to the previous-level node to rewrite the unvisited path until the root node is returned, thereby screening out all valid reconstruction solutions.

[0172] As a preferred solution of this embodiment, according to the distribution function and probability density function of each distributed power output, as well as the initialized scene parameters, particle swarm algorithm parameters and the effective reconstruction solution, a binary coded particle swarm algorithm with probability jump is used to obtain the optimal reconstruction solution, specifically:

[0173] According to the status of each switch in the distribution network, each effective reconstruction solution is binary-encoded to form a corresponding binary position vector and binary velocity vector; based on the binary position vector and binary velocity vector of the effective reconstruction solution, the initial position of each particle in the particle swarm is determined, and the initial velocity of each particle in the particle swarm is randomly generated; the iteration is started through GPU parallel computing, the objective function value of the solution corresponding to the position of each particle in the particle swarm in the iteration is calculated, and the historical optimal position of each particle in the particle swarm is updated; wherein the optimality condition is that the objective function value of the solution corresponding to the position of the particle is the minimum; each iteration is calculated through GPU parallel computing The transition probability of each particle in the particle swarm in the current iteration to its historical optimal position is calculated; the roulette method is used to make each particle in the particle swarm in the current iteration transition to its respective historical optimal position according to the transition probability, until the first particle that successfully transitions is generated, and its position after the successful transition is recorded; the position of each particle in the particle swarm in the current iteration is updated in parallel through the GPU; if the number of iterations reaches the preset maximum value, the objective function value of the solution corresponding to the position of each particle in the particle swarm in the iteration is calculated in parallel through the GPU, and the solution with the lowest objective function value is taken as the global optimal solution, thereby stopping the iteration and outputting the reconstruction result; otherwise, another iterative calculation is performed.

[0174] As a preferred solution of this embodiment, the objective function value of the solution corresponding to the position of each particle in the particle swarm during iteration is calculated as follows:

[0175] The node admittance matrix of the solution is established through the branch addition method; based on the distribution function and probability density function of each distributed power output, several distributed power output samples are obtained; effective distributed power output samples are screened out; based on the topology and node admittance matrix of the solution, the distribution network flow under each distributed power output sample is solved, and the total cost function under each distributed power output sample is calculated, and then the average value of the total cost function under each distributed power output sample is calculated, and the average value is used as the objective function value of the solution.

[0176] As a preferred solution of this embodiment, the total cost function under each distributed power output sample includes: network loss cost, voltage offset cost, switch fee and photovoltaic power station configuration cost.

[0177] As a preferred solution of this embodiment, the position of each particle of the particle swarm in the current iteration is updated in parallel by the GPU, specifically:

[0178] Through GPU parallel calculation, the position of each particle in the particle swarm in the kth iteration is updated. The particle update formula is:

[0179]

[0180] Among them, round means that each component is rounded to the nearest integer; X i (k), X i (k+1) are the positions of particle i before and after the kth iteration, V i (k), V i (k+1) are the velocities of particle i before and after the kth iteration respectively; P i (k) is the historical optimal position of particle i at the kth iteration, X g (k) is the position of the optimal particle in the particle swarm at the kth iteration; c1 and c2 are learning factors, and 0 <c1<c2<2、 l=c1+c2; r1 and r2 are random numbers between 0 and 1.

[0181] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0182] The implementation of the above embodiment has the following effects:

[0183] Compared with the prior art, the embodiments of the present invention initialize the distribution network parameters, scenario parameters, and particle swarm algorithm parameters to obtain the distribution function and probability density function of each distributed power output, thereby taking into account the optimization requirements under multiple scenarios and avoiding the problems of relatively rough modeling, simple reconstruction targets, and difficulty in switching in the prior art. Effective reconstruction solutions are screened out by generating reconstruction solutions of the entire distribution network, and then a binary coded particle swarm algorithm with probability jumps is used to obtain the optimal reconstruction solution to obtain and complete the reconstruction of the distribution network under multiple scenarios, thereby improving the solution efficiency and fully considering the repeatability of steps such as obtaining DG output distribution, screening effective reconstruction solutions, screening effective distributed power output samples, calculating the fitness function of the effective solution, updating the position of particles, and calculating the jump probability of particles.

[0184] Example 3

[0185] Accordingly, the present invention also provides a terminal device, comprising: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements the multi-scenario distribution network reconstruction method described in any one of the above embodiments.

[0186] The terminal device of this embodiment includes: a processor, a memory, and a computer program and computer instructions stored in the memory and capable of running on the processor. When the processor executes the computer program, each step in the above embodiment 1 is implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiment are realized, such as the effective reconstruction module 203 .

[0187] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the terminal device. For example, the effective reconstruction module 203 is configured to generate reconstruction solutions for the entire distribution network and select effective reconstruction solutions.

[0188] The terminal device may be a computing device such as a desktop computer, laptop, PDA, or cloud server. The terminal device may include, but is not limited to, a processor and memory. Those skilled in the art will appreciate that the schematic diagram is merely an example of a terminal device and does not limit the terminal device. The terminal device may include more or fewer components than shown, or a combination of certain components, or different components. For example, the terminal device may also include input / output devices, network access devices, buses, and the like.

[0189] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting various parts of the entire terminal device using various interfaces and lines.

[0190] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created based on the use of the mobile terminal, etc. In addition, the memory can include a high-speed random access memory and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0191] If the module / unit integrated into the terminal device is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention can implement all or part of the process of the above-mentioned method embodiment by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal, and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media does not include electric carrier signals and telecommunication signals.

[0192] Example 4

[0193] Accordingly, the present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the multi-scenario distribution network reconstruction method described in any one of the above embodiments.

[0194] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A multi-scenario distribution network reconstruction method, characterized in that: include: Initialize distribution network parameters, scenario parameters and particle swarm algorithm parameters; the particle swarm algorithm parameters are binary coded particle swarm algorithm parameters with probability jumps; According to the initialized distribution network parameters, the distribution function and probability density function of each distributed power output are obtained; Generate reconstruction solutions for the entire distribution network and select effective reconstruction solutions; According to the distribution function and probability density function of each distributed power output, as well as the initialized scenario parameters, particle swarm algorithm parameters and the effective reconstruction solution, a binary coded particle swarm algorithm with probability jump is used to obtain the optimal reconstruction solution, thereby obtaining a multi-scenario distribution network reconstruction method; Generate the reconstruction solutions of the entire distribution network and select the effective reconstruction solutions, specifically: According to the closed states of all section switches and tie switches of the distribution network, a distribution network topology with all switches closed and containing loops is obtained, and according to the distribution network topology, all meshes and branches under the distribution network topology are obtained; Disconnect one branch in each mesh, so that after each branch is disconnected, a reconstruction solution of the distribution network is generated, and then all reconstruction solutions are generated; The effectiveness of each reconstruction solution is judged through GPU parallelization, and all valid reconstruction solutions are screened out; The effectiveness of each reconstruction solution is judged by GPU parallelization, and all valid reconstruction solutions are screened out. Specifically: The validity of each reconstruction solution is determined through GPU parallelization. If the reconstruction solution is a radial topology, the number of lines put into operation is equal to the difference between the number of nodes and the number of connected subgraphs. If the reconstruction solution does not contain isolated islands, all connected subgraphs contain power sources. In this way, each time the reconstruction solution is determined by the GPU, the search is performed downward from the root node to the end node, and from the end node, the search is returned to the first-level nodes one by one to determine whether there is an unvisited path back to the first-level node. If so, the unvisited path is followed to the end node of the unvisited path. If not, the search is returned to the previous-level node to re-determine the unvisited path until the root node is returned, thereby screening out all valid reconstruction solutions. According to the distribution function and probability density function of each distributed power output, as well as the initialized scene parameters, particle swarm algorithm parameters and the effective reconstruction solution, a binary coded particle swarm algorithm with probability jump is used to obtain the optimal reconstruction solution, specifically: According to the status of each switch in the distribution network, each valid reconstruction solution is binary-encoded to form the corresponding binary position vector and binary velocity vector; According to the binary position vector and binary velocity vector of the effective reconstruction solution, the initial position of each particle in the particle swarm is determined, and the initial velocity of each particle in the particle swarm is randomly generated; Iteration is started through GPU parallel computing, the objective function value of the solution corresponding to the position of each particle in the particle swarm during the iteration is calculated, and the historical optimal position of each particle in the particle swarm is updated; wherein the historical optimal position is the position of the particle corresponding to the minimum objective function value of the solution; The probability of each particle in the particle swarm jumping to its historical optimal position in each iteration is calculated in parallel by GPU; The roulette wheel method is used to make each particle in the particle swarm in the current iteration jump to its own historical optimal position according to the jump probability, until the first particle that successfully jumps is generated, and its position after the successful jump is recorded; Update the position of each particle in the current iteration in parallel through the GPU; If the number of iterations reaches the preset maximum value, the objective function value of the solution corresponding to the position of each particle in the particle swarm in the iteration is calculated in parallel by GPU, and the solution with the smallest objective function value is taken as the global optimal solution, thus stopping the iteration and outputting the reconstruction result; otherwise, another iterative calculation is performed.

2. A multi-scenario distribution network reconstruction method according to claim 1, characterized in that: The initialization of distribution network parameters, scenario parameters and particle swarm algorithm parameters is specifically as follows: Initialize the input distribution network parameters; the distribution network parameters include: the number, rated voltage and shunt admittance of each node, the number, connected nodes and series impedance of each branch, the branch location and status of each switch, the node location, active power and reactive power of each load, the type, device parameters, node location and historical operation data set of each distributed power source; Initializing the input scene parameters; the scene parameters include weight parameters and cost parameters; Initialize the input particle swarm algorithm parameters; the particle swarm algorithm parameters include population size, maximum number of iterations and learning factor.

3. A multi-scenario distribution network reconstruction method according to claim 1, characterized in that: The distribution function and probability density function of each distributed power output are obtained according to the initialized distribution network parameters, specifically: Based on the initialized distribution network parameters, a wind speed distribution model is formed. Combined with the constructed piecewise cubic model of the wind farm, the Monte Carlo method is used for random sampling to fit the sampling results and obtain the distribution function of the wind farm output. The numerical differentiation method is then used to obtain the probability density function of the wind farm output. According to the initialized distribution network parameters, a light intensity distribution model is formed. Combined with the constructed piecewise quadratic model of the photovoltaic power station, the Monte Carlo method is used for random sampling to fit the sampling results and obtain the distribution function of the photovoltaic power station output. The numerical differentiation method is used to obtain the probability density function of the photovoltaic power station output.

4. A multi-scenario distribution network reconstruction method according to claim 1, characterized in that: The objective function value of the solution corresponding to the position of each particle in the particle swarm during the iteration is calculated as follows: The node admittance matrix of the solution is established by the branch addition method; Based on the distribution function and probability density function of each distributed power output, several distributed power output samples are obtained; Screen out effective distributed power output samples; Based on the topology and node admittance matrix of the solution, the distribution network flow under each distributed power output sample is solved, and the total cost function under each distributed power output sample is calculated. Then, the average value of the total cost function under each distributed power output sample is calculated, and the average value is used as the objective function value of the solution.

5. A multi-scenario distribution network reconstruction method according to claim 4, characterized in that: The total cost function under each distributed power generation output sample includes: network loss cost, voltage offset cost, switching cost and photovoltaic power station configuration cost.

6. A multi-scenario distribution network reconstruction method according to claim 4, characterized in that: The position of each particle in the particle swarm in the current iteration is updated in parallel by the GPU, specifically: Through GPU parallel calculation, the position of each particle in the particle swarm in the kth iteration is updated. The particle update formula is: Among them, round means that each component is rounded to the nearest integer; X i (k), X i (k+1) are the positions of particle i at the kth iteration and the k+1th iteration, respectively. V i (k), V i (k+1) are the velocities of particle i at the kth iteration and the k+1th iteration respectively; P i (k) is the historical optimal position of particle i at the kth iteration, X g (k) is the position of the optimal particle in the particle swarm at the kth iteration; c1 and c2 are learning factors, and 0 <c1<c2<2、 l=c1+c2; r1 and r2 are random numbers between 0 and 1.

7. A multi-scenario distribution network reconstruction device, characterized in that: Used to implement the multi-scenario distribution network reconstruction method according to any one of claims 1 to 6, comprising: an initialization module, a function module, an effective reconstruction module and an optimal reconstruction module; The initialization module is used to initialize the distribution network parameters, scene parameters and particle swarm algorithm parameters; the particle swarm algorithm parameters are binary coded particle swarm algorithm parameters with probability jumps; The function module is used to obtain the distribution function and probability density function of each distributed power output according to the initialized distribution network parameters; The effective reconstruction module is used to generate reconstruction solutions for the entire distribution network and screen out effective reconstruction solutions; The optimal reconstruction module is used to obtain the optimal reconstruction solution using a binary coded particle swarm algorithm with probability jumps based on the distribution function and probability density function of the output of each distributed power source, as well as the initialized scenario parameters, particle swarm algorithm parameters and the effective reconstruction solution, thereby obtaining a multi-scenario distribution network reconstruction method.

Images