Power distribution network reconstruction method and system based on large-scale parallel computing architecture

Abstract

The invention provides a power distribution network reconstruction method and system based on a large-scale parallel computing architecture, and the method comprises the steps: combining a pre-constructed power system network reconstruction model based on power system network parameters, employing a binary particle swarm optimization algorithm for solving, and obtaining a power system network reconstruction optimal solution; based on the communication system parameters in combination with a pre-constructed communication system fault reconstruction model, solving by adopting a mode of combining a single-source shortest path with a dynamic routing method to obtain a communication system network reconstruction optimal solution; and forming a power distribution network reconstruction scheme by the power system network reconstruction optimal solution and the communication system network reconstruction optimal solution. According to the method, faults, risks and optimization states of the power distribution network and the communication network are fully considered, network reconstruction strategy generation under mutual influence of the communication network and the power network is realized, and the method has relatively high applicability and feasibility for an actual field.

Description

Technical Field

[0001] The present invention relates to the field of distribution network operation analysis and control, and in particular to a distribution network network reconstruction method and system based on a large-scale parallel computing architecture. Background Art

[0002] With the rapid development of the economy and society, the scale of urban and rural distribution systems has grown rapidly. The scale of county distribution networks in economically developed areas has grown to 300-500 feeders, and the scale of super-large urban distribution networks has reached 6,000-7,000 feeders, 20,000-30,000 substations. The distribution network forms are also becoming increasingly diverse, such as the rapid development of double-petal distribution network racks, diamond-shaped distribution network racks, and snowflake-shaped distribution network racks. The distribution network has completely evolved into a large-scale complex system. In this case, when encountering distribution network failures caused by natural disasters, equipment failures, and human factors, network reconstruction and fault recovery will face major challenges, specifically the following problems:

[0003] (1) Large-scale complex distribution networks are highly nonlinear and multidimensional, which means that a large number of variables and constraints need to be processed during fault reconstruction. These variables and constraints are intertwined, making the solution process extremely complicated. Therefore, how to avoid the "curse of dimensionality" while obtaining the global optimal solution has become an important problem in large-scale distribution network fault reconstruction;

[0004] (2) As the last link of the power system, the distribution network is directly connected to the electricity users. Therefore, its status and data are constantly changing over time, which will affect the effect of reconstruction. That is, large-scale distribution network fault reconstruction is a dynamic optimization problem. This dynamic variability increases the difficulty and complexity of reconstruction.

[0005] (3) When solving the problem of large-scale distribution network fault reconstruction, a large number of local extreme points are often encountered. These local extreme points will interfere with the process of solving the global optimal solution, making the algorithm easily fall into the local optimal solution and unable to find the global optimal solution. Therefore, how to effectively avoid the influence of local extreme points has become another important problem in large-scale distribution network fault reconstruction;

[0006] (4) During the reconstruction process, it is generally necessary to perform power flow calculations multiple times to verify the feasibility of the reconstruction scheme. However, the power flow calculations of large-scale distribution networks are often very complex and require a lot of computing resources and time. Therefore, how to reduce the time of power flow calculations and improve computing efficiency has become an important aspect in the optimization process of large-scale distribution networks;

[0007] (5) With the rapid development of renewable energy, more and more distributed power sources are connected to the distribution network. The access of these distributed power sources makes the structure and characteristics of the distribution network more complex, bringing new challenges to fault reconstruction. For example, the access of distributed power sources may lead to changes in the direction of the power flow and increased voltage fluctuations, which need to be fully considered and handled during the reconstruction process;

[0008] (6) Since there are many participants in the current distribution system, such as distributed power aggregators, large-load users, charging station (pile) aggregators, etc., they are all deeply involved in the operation of the distribution system. In the event of a failure, the connectivity and real-time performance of the communication system are particularly important. When considering the availability of network reconstruction, the connectivity and real-time performance of the communication system should be further considered. This problem becomes a complex problem of coupling the power system and the communication system. Summary of the invention

[0009] In order to solve the problem that existing methods are insufficient in research on large-scale distribution network fault reconstruction, the present invention proposes a distribution network reconstruction method based on a large-scale parallel computing architecture, including:

[0010] Based on the power system network parameters and the pre-built power system network reconstruction model, the binary particle swarm optimization algorithm is used to solve the optimal solution for the power system network reconstruction;

[0011] Based on the communication system parameters and the pre-built communication system fault reconstruction model, the single-source shortest path and dynamic routing methods are combined to obtain the optimal solution for the communication system network reconstruction.

[0012] The distribution network reconstruction plan is composed of the power system network reconstruction optimal solution and the communication system network reconstruction optimal solution.

[0013] Optionally, the construction of the power system network reconstruction model includes:

[0014] The first reconstruction optimization objective is established with minimum line loss;

[0015] The second reconstruction optimization objective is to minimize the number of switch actions and transformer voltage regulation times;

[0016] Setting a first constraint condition for the first reconstruction optimization target and the second reconstruction optimization target;

[0017] Constructing a power system network reconstruction model based on the first reconstruction optimization objective and the second reconstruction optimization objective and constraint conditions;

[0018] The first constraint condition includes: power flow constraint, capacity constraint, connectivity constraint and ring network identification constraint.

[0019] Optionally, the first reconstruction optimization objective is as follows:

[0020]

[0021] Where U i is the node voltage; P i and Q i is the active and reactive outflow power of node i; P i+1 and Q i+1 is the active and reactive outflow power of node i+1; P L,i+1 and Q L,i+1 Inject active and reactive power into node i+1; P loss,i and Q loss,i is the active and reactive power loss before node i; P tloss represents the total network loss; N represents the number of nodes; P loss,i～(i+1) is the line loss from node i to node i+1; R i is the resistance of node i.

[0022] Optionally, the second reconstruction optimization objective is as follows:

[0023]

[0024] Where F is the second reconstruction optimization target, x1 is the influence weight of the number of switch actions, x2 is the influence weight of the number of transformer voltage regulation, and Y k and Z j are the states of the disconnecting switch and the tie switch after reconstruction, k and j are the numbers of the section switches and tie switches in the network, m and n are the numbers of the section switches and tie switches in the network, T a is the transformer voltage regulating volume number, a is the transformer serial number, and nb is the total number of transformers.

[0025] Optionally, the method of solving the power system network reconstruction optimal solution based on the power system network parameters in combination with a pre-built power system network reconstruction model by using a binary particle swarm optimization algorithm includes:

[0026] ① Describe the on / off state of the switch as an array A;

[0027] ②Start from the power supply point to search for the ring network formed by the closing of the tie switch, and record the switch numbers included in the array {b i};

[0028] ③ Disconnect {b i}, making the ring network an open loop; when all the ring networks are open loops, as a particle x i, the particle displacement vector represents the switch state, and the particle velocity vector is the probability of displacement change; and detect the particle x i feasibility;

[0029] ④ Repeat step ③ until the particle size meets the set size requirements;

[0030] ⑤ Start the assisted search between particles, and perform topological verification on the searched optimization results, calculate the fitness value of the optimization results after topological verification, and save the fitness value; if it is better than the historical optimal solution, the optimal solution is updated by the fitness value of the optimization result after topological verification;

[0031] ⑥ Repeat step ⑤ until the maximum number of iterations t is reached max Or achieve the required accuracy, and then output the optimal solution and the topological map corresponding to the optimal solution.

[0032] Optionally, the construction of the communication system fault reconstruction model includes:

[0033] The communication system fault reconstruction objective function is constructed with the goal of minimizing the sum of the on / off, delay, and error time of the communication between the physical station and the master station;

[0034] Setting a second constraint condition for the communication system fault reconstruction objective function;

[0035] Constructing a communication system fault reconstruction model based on the communication system fault reconstruction objective function and the second constraint condition;

[0036] The second constraint conditions include: communication current carrying capacity constraint and communication terminal power supply constraint.

[0037] Optionally, the communication system fault reconstruction objective function is as follows:

[0038]

[0039] In the formula, f d (c) is the sum of interruption time, delay time and error time, c is the communication link number, R is the number of routing nodes in the communication layer; B(c) is the on / off status of the c-th link; T(c) is the communication delay of the c-th link; P(c) is the data upload error rate of the c-th link; H(c) is the error rate of the control command issued by the c-th link; α, β, γ are weight coefficients.

[0040] Optionally, the method of solving the problem based on the communication system parameters in combination with a pre-built communication system fault reconstruction model by combining a single-source shortest path with a dynamic routing method to obtain an optimal solution for the communication system network reconstruction includes:

[0041] Abstract the communication network into a graph, and divide all the vertices in the graph into two groups, one group is the vertices of the known shortest path, and the other group is the vertices of the unknown shortest path. Initially, all vertices belong to the vertices of the unknown shortest path, and mark the path length of each vertex as infinite, and set the path length of the starting point to 0;

[0042] In each stage, the vertex with the shortest path length is selected from all the vertices of the unknown shortest path, and the vertex is marked as the vertex of the known shortest path;

[0043] Then, the path length from the vertex of the known shortest path to the vertex of its adjacent vertex belonging to the unknown shortest path is calculated. If the path length is shorter than the original path length, the path length of the adjacent vertex is updated, and the preceding node of the adjacent vertex is set to the selected vertex; the loop operation is continuously performed until the number of vertices of the unknown shortest path is 0; and the communication routing table is output as the optimal solution for the communication system network reconstruction of the communication layer model.

[0044] Optionally, the method further includes: verifying the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction using a pre-built optimization model.

[0045] Optionally, the verifying the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction using a pre-built optimization model includes:

[0046] Perform preliminary partitioning of the physical network based on the topology of the information network;

[0047] Based on the preliminary zoning, physical network division is carried out;

[0048] After the physical network is divided, the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction are substituted into a pre-built optimization model to obtain the CPU calculation balance under consideration of electrical connection;

[0049] The CPU calculates the degree of balance to verify the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction.

[0050] Optionally, the construction of the optimization model includes:

[0051] Construct the verification objective function with the goal of minimizing CPU computing balance;

[0052] Setting verification constraints for the verification objective function;

[0053] The verification constraints include: the first feeder number calculated by the next CPU minus the last feeder number calculated by the previous CPU, the actual load number calculated by the i-th CPU cannot exceed the upper limit of the ideal load number, and the first feeder calculated by the first CPU must be the feeder numbered 1 and belongs to the first substation and partition variable constraints.

[0054] Optionally, the verification objective function is as shown below:

[0055]

[0056] Where: It measures the normalized computing power of the ith CPU; (m i,end -m i+1,st ) measures the feeder cross-substation situation calculated by the i-th CPU after normalization; ω1 is the weight coefficient of the first-level optimization target; ω2 is the weight coefficient of the second-level optimization target, ω1>ω2>0; is the equivalent load number of the ideal calculation of the i-th CPU; n cal,i is the equivalent load number actually calculated by the i-th CPU; m i,st is the substation number of the first feeder calculated by the i-th CPU; m i+1,st The substation number of the first feeder calculated by the i+1th CPU; m i,end is the substation number of the last feeder calculated by the i-th CPU; Fun is the CPU calculation balance, N CPU Indicates the CPU number.

[0057] Preferably, the preliminary partitioning of the physical network based on the topological structure of the information network includes:

[0058] ① Obtain the equivalent electrical distance between nodes as the node feature;

[0059] ② Classify each node in the power system into a separate category and calculate the distance between categories;

[0060] ③ Initialize the clustering acceleration factor δ;

[0061] ④Find all class pairs whose inter-class distance is less than δ. If there are none, select the class pair with the smallest inter-class distance;

[0062] ⑤ Check whether each class pair is electrically connected in turn. If there is no connecting line, set the electrical distance between the class pairs to be infinite. Otherwise, merge the class pairs into a new class.

[0063] ⑥Update the inter-class distance and δ value;

[0064] ⑦ Repeat steps ④, ⑤, and ⑥ until all nodes are merged into one class;

[0065] ⑧Analyze the statistic F, determine the target number of partitions, and obtain the partition results.

[0066] In another aspect, the present invention further provides a distribution network reconstruction system based on a large-scale parallel computing architecture, comprising:

[0067] The power system reconstruction module is used to obtain the optimal solution for power system network reconstruction based on power system network parameters and a pre-built power system network reconstruction model using a binary particle swarm optimization algorithm;

[0068] The network system reconstruction module is used to solve the communication system network reconstruction optimal solution based on the communication system parameters and the pre-built communication system fault reconstruction model by combining the single-source shortest path with the dynamic routing method;

[0069] The reconstruction scheme module is used to form a distribution network reconstruction scheme based on the optimal solution for the reconstruction of the power system network and the optimal solution for the reconstruction of the communication system network.

[0070] Optionally, a power system reconstruction module is further included, for:

[0071] The first reconstruction optimization objective is established with minimum line loss;

[0072] The second reconstruction optimization objective is to minimize the number of switch actions and transformer voltage regulation times;

[0073] Setting a first constraint condition for the first reconstruction optimization target and the second reconstruction optimization target;

[0074] Constructing a power system network reconstruction model based on the first reconstruction optimization objective and the second reconstruction optimization objective and constraint conditions;

[0075] The first constraint condition includes: power flow constraint, capacity constraint, connectivity constraint and ring network identification constraint.

[0076] Optionally, the first reconstruction optimization objective is as follows:

[0077]

[0078] Where U i is the node voltage; P i and Q i is the active and reactive outflow power of node i; P L,i+1 and Q L,i+1 Inject active and reactive power into node i+1; P loss,i and Q loss,i is the active and reactive power loss before node i; P tloss represents the total network loss; N represents the number of nodes; P loss,i～(i+1)is the line loss from node i to node i+1.

[0079] Optionally, the second reconstruction optimization objective is as follows:

[0080]

[0081] Where F is the second reconstruction optimization target, x1 is the influence weight of the number of switch actions, x2 is the influence weight of the number of transformer voltage regulation, and Y k and Z j are the states of the disconnecting switch and the tie switch after reconstruction, m and n are the numbers of the sectionalizing switch and the tie switch in the network, T a is the number of transformer voltage regulators, a is the number of transformers, and nb is the total number of transformers.

[0082] Optionally, the power system reconstruction module is specifically used to:

[0083] ① Describe the on / off state of the switch as an array A;

[0084] ②Start from the power supply point to search for the ring network formed by the closing of the tie switch, and record the switch numbers included in the array {b i};

[0085] ③ Disconnect {b i}, making the ring network an open loop; when all the ring networks are open loops, as a particle x i , the particle displacement vector represents the switch state, and the particle velocity vector is the probability of displacement change; and detect the particle x i feasibility;

[0086] ④ Repeat step ③ until the particle size meets the set size requirements;

[0087] ⑤ Start the assisted search between particles, and perform topological verification on the searched optimization results, calculate the fitness value of the optimization results after topological verification, and save the fitness value; if it is better than the historical optimal solution, the optimal solution is updated by the fitness value of the optimization result after topological verification;

[0088] ⑥ Repeat step ⑤ until the maximum number of iterations t is reached max Or achieve the required accuracy, and then output the optimal solution and the topological map corresponding to the optimal solution.

[0089] Optionally, a network model building module is also included, for:

[0090] The communication system fault reconstruction objective function is constructed with the goal of minimizing the sum of the on / off, delay, and error time of the communication between the physical station and the master station;

[0091] Setting a second constraint condition for the communication system fault reconstruction objective function;

[0092] Constructing a communication system fault reconstruction model based on the communication system fault reconstruction objective function and the second constraint condition;

[0093] The second constraint conditions include: communication current carrying capacity constraint and communication terminal power supply constraint.

[0094] Optionally, the communication system fault reconstruction objective function is as follows:

[0095]

[0096] In the formula, f d (c) is the sum of interruption time, delay time and error time, c is the communication link number, R is the number of routing nodes in the communication layer; B(c) is the on / off status of the c-th link; T(c) is the communication delay of the c-th link; P(c) is the data upload error rate of the c-th link; H(c) is the error rate of the control command issued by the c-th link; α, β, γ are weight coefficients.

[0097] Optionally, the network system reconstruction module is specifically used to:

[0098] Abstract the communication network into a graph, and divide all the vertices in the graph into two groups, one group is the vertices of the known shortest path, and the other group is the vertices of the unknown shortest path. Initially, all vertices belong to the vertices of the unknown shortest path, and mark the path length of each vertex as infinite, and set the path length of the starting point to 0;

[0099] In each stage, the vertex with the shortest path length is selected from all the vertices of the unknown shortest path, and the vertex is marked as the vertex of the known shortest path;

[0100] Then, the path length from the vertex of the known shortest path to the vertex of its adjacent vertex belonging to the unknown shortest path is calculated. If the obtained path length is shorter than the original length, the path length of the adjacent vertex is updated, and the preceding node of the adjacent vertex is set to the selected vertex; the loop operation is continuously performed until the number of vertices of the unknown shortest path is 0; and the communication routing table is output as the optimal solution for the communication system network reconstruction of the communication layer model.

[0101] Optionally, it also includes: a verification module, which is used to verify the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction using a pre-built decoupling model.

[0102] Optionally, the verification module is specifically used to:

[0103] Perform preliminary partitioning of the physical network based on the topology of the information network;

[0104] Based on the preliminary zoning, physical network division is carried out;

[0105] After the physical network is divided, the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction are substituted into a pre-built optimization model to obtain the CPU calculation balance under consideration of electrical connection;

[0106] The CPU calculates the degree of balance to verify the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction.

[0107] Optionally, an optimization model building module is also included, for:

[0108] Construct the verification objective function with the goal of minimizing CPU computing balance;

[0109] Setting verification constraints for the verification objective function;

[0110] The verification constraints include: the first feeder number calculated by the next CPU minus the last feeder number calculated by the previous CPU, the actual load number calculated by the i-th CPU cannot exceed the upper limit of the ideal load number, and the first feeder calculated by the first CPU must be the feeder numbered 1 and belongs to the first substation and partition variable constraints.

[0111] Optionally, the verification objective function is as shown below:

[0112]

[0113] Where: It measures the normalized computing power of the ith CPU; (m i,end -m i+1,st ) measures the feeder cross-substation situation calculated by the i-th CPU after normalization; ω1 is the weight coefficient of the first-level optimization target; ω2 is the weight coefficient of the second-level optimization target, ω1>ω2>0; is the equivalent load number of the ideal calculation of the i-th CPU; n cal,i is the equivalent load number actually calculated by the i-th CPU; m i,st is the substation number of the first feeder calculated by the i-th CPU; m i+1,st The substation number of the first feeder calculated by the i+1th CPU; m i,end is the substation number of the last feeder calculated by the i-th CPU; M is the total number of substations, Fun is the CPU calculation balance, N CPU Indicates the CPU number.

[0114] Preferably, the verification module performs preliminary partitioning of the physical network based on the topological structure of the information network, and the specific implementation steps include:

[0115] ① Obtain the equivalent electrical distance between nodes as the node feature;

[0116] ② Classify each node in the power system into a separate category and calculate the distance between categories;

[0117] ③ Initialize the clustering acceleration factor δ;

[0118] ④Find all class pairs whose inter-class distance is less than δ. If there are none, select the class pair with the smallest inter-class distance;

[0119] ⑤ Check whether each class pair is electrically connected in turn. If there is no connecting line, set the electrical distance between the class pairs to be infinite. Otherwise, merge the class pairs into a new class.

[0120] ⑥Update the inter-class distance and δ value;

[0121] ⑦ Repeat steps ④, ⑤, and ⑥ until all nodes are merged into one class;

[0122] ⑧Analyze the statistic F, determine the target number of partitions, and obtain the partition results.

[0123] Optionally, after the physical network is divided, the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction are substituted into a pre-built optimization model to obtain a CPU calculation balance considering electrical connection, including:

[0124] Initialize the MPI environment and treat the verification of each physical sub-network as a subtask, with one subtask corresponding to one instruction;

[0125] Map images distribute instructions to multiple workers. The task granularity of workers during execution generally requires that the small data block is less than or equal to the size of a block in HDFS, and the number of Maps and Reduces is greater than the number of workers.

[0126] The reduce specification combines the results calculated by the Map workers;

[0127] The Master periodically detects the Worker and re-executes the Map if a failure occurs. If the Reduce is not completed, it is re-executed.

[0128] In another aspect, the present application further provides an electronic device, comprising: at least one processor and a memory; the memory and the processor are connected via a bus;

[0129] The memory is used to store one or more programs;

[0130] When the one or more programs are executed by the at least one processor, the distribution network reconstruction method based on a large-scale parallel computing architecture as described above is implemented.

[0131] On the other hand, the present application also provides a readable storage medium having an execution program stored thereon, and when the execution program is executed, the distribution network reconstruction method based on a large-scale parallel computing architecture as described above is implemented.

[0132] Compared with the prior art, the present invention has the following beneficial effects:

[0133] The present invention provides a distribution network reconstruction method based on a large-scale parallel computing architecture, including: based on the power system network parameters combined with a pre-constructed power system network reconstruction model, a binary particle swarm optimization algorithm is used to solve and obtain the optimal solution for the power system network reconstruction; based on the communication system parameters combined with a pre-constructed communication system fault reconstruction model, a single-source shortest path is used in combination with a dynamic routing method to solve and obtain the optimal solution for the communication system network reconstruction; the power system network reconstruction optimal solution and the communication system network reconstruction optimal solution form a distribution network network reconstruction plan. The present invention fully considers the faults, risks, and optimization states of the distribution network and the communication network, realizes the generation of network reconstruction strategies under the mutual influence of the communication network and the power network, and has strong applicability and feasibility for actual sites. BRIEF DESCRIPTION OF THE DRAWINGS

[0134] Figure 1 It is a flow chart of the distribution network reconstruction method based on large-scale parallel computing architecture of the present invention;

[0135] Figure 2 It is an overall flow chart of the distribution network reconstruction method based on large-scale parallel computing architecture of the present invention;

[0136] Figure 3 A flow chart for solving the power distribution network reconstruction model of the present invention;

[0137] Figure 4 The flow chart of solving the distribution network communication network reconstruction model of the present invention is

[0138] Figure 5 Flowchart for realizing the method for decoupling power distribution network partitions of the present invention

[0139] Figure 6 Design flow chart of MPI parallel programming of the present invention;

[0140] Figure 7 It is a schematic diagram of the MPI calculation execution process of the present invention;

[0141] Figure 8 The figure is a schematic diagram of the structure of an electronic device of the present invention. DETAILED DESCRIPTION

[0142] The present invention proposes a distribution network reconstruction method based on a large-scale parallel computing architecture, which simultaneously considers the deep coupling and mutual influence relationship between the communication system and the power system, realizes the generation of large-scale distribution network fault reconstruction strategy and the testing and verification of its rationality and feasibility, effectively improves the safety and reliability of the distribution system, and provides accurate and real-time auxiliary decision support for distribution operation and maintenance personnel.

[0143] In order to better understand the present invention, the content of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0144] Embodiment 1:

[0145] Distribution network reconstruction methods based on large-scale parallel computing architecture, such as Figure 1 As shown, including:

[0146] Step S1: Based on the power system network parameters and the pre-built power system network reconstruction model, a binary particle swarm optimization algorithm is used to solve and obtain the optimal solution for the power system network reconstruction;

[0147] Step S2: Based on the communication system parameters and the pre-built communication system fault reconstruction model, a single-source shortest path combined with a dynamic routing method is used to solve the problem, and the optimal solution for the communication system network reconstruction is obtained;

[0148] Step S3: A distribution network reconstruction plan is formed by the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction.

[0149] The present invention proposes a distribution network reconstruction method based on a large-scale parallel computing architecture for large-scale distribution systems containing distributed resources. Firstly, a large-scale distribution network reconstruction algorithm covering multi-dimensional objectives and multi-dimensional constraints such as line loss, voltage safety, line current carrying capacity, upper and lower limits of distributed power supply capacity, and number of switch operations is proposed. Secondly, on this basis, communication side constraints are added, and a distribution network reconstruction algorithm under communication system status constraints is proposed. Finally, a distribution network simulation method based on a large-scale parallel computing architecture is proposed to verify the rationality and feasibility of the network reconstruction strategy.

[0150] The present invention proposes a distribution network reconstruction method and system based on a large-scale parallel computing architecture. The overall framework is as shown in the attached Figure 2 As shown, the specific steps are as follows.

[0151] Before step S1, the method also includes: constructing a power system network reconstruction model.

[0152] The construction of the power system network reconstruction model includes:

[0153] The first reconstruction optimization objective is established with minimum line loss;

[0154] The second reconstruction optimization objective is to minimize the number of switch actions and transformer voltage regulation times;

[0155] Setting a first constraint condition for the first reconstruction optimization target and the second reconstruction optimization target;

[0156] Constructing a power system network reconstruction model based on the first reconstruction optimization objective and the second reconstruction optimization objective and constraint conditions;

[0157] The first constraint condition includes: power flow constraint, capacity constraint, connectivity constraint and ring network identification constraint.

[0158] Furthermore, the first reconstruction optimization objective is as follows:

[0159]

[0160] Where U i is the node voltage; P i and Q i is the active and reactive outflow power of node i; P i+1 and Q i+1 is the active and reactive outflow power of node i+1; P L,i+1 and Q L,i+1 Inject active and reactive power into node i+1; P loss,i and Q loss,i is the active and reactive power loss before node i; P tloss represents the total network loss; N represents the number of nodes; P loss,i～(i+1) is the line loss from node i to node i+1; R i is the resistance of node i.

[0161] Furthermore, the second reconstruction optimization objective is as follows:

[0162]

[0163] Where F is the second reconstruction optimization target, x1 is the influence weight of the number of switch actions, x2 is the influence weight of the number of transformer voltage regulation, and Y k and Z j are the states of the disconnecting switch and the tie switch after reconstruction, m and n are the numbers of the sectionalizing switch and the tie switch in the network, T a is the transformer voltage regulating volume number, a is the transformer serial number, and nb is the number of transformers.

[0164] The following is a detailed description of the construction process:

[0165] Step 1: Establish a power system network reconstruction model

[0166] Step 1: Establish constraints

[0167] (1) Constraint 1: Power flow constraint

[0168]

[0169] Where P ac,i and Q ac,i They represent the active power and reactive power injected into the AC system node i respectively; G ac,ij , B ij ,θ ij Respectively represent the conductance, susceptance and voltage phase difference between nodes i and j in the AC system; U ac,i and U ac,j represents the voltage amplitude of node i, j; P dc,i represents the injected active power of DC system node i; U dc,i and U dc,j represents the voltage amplitude of the DC system node i, j; G dc,ij represents the conductance between nodes i and j in the DC system, i is node i, j is node j, N dc is the number of DC nodes in the system, N ac is the number of nodes in the AC system.

[0170] (2) Constraint 2: Equipment, line, and distributed resource capacity constraints

[0171] U min ≤U i ≤U max (2)

[0172] I min ≤I ij ≤I max (3)

[0173] P min ≤P i ≤P max (4)

[0174] Q min ≤Q i ≤Q max (5)

[0175] Where U min is the lower limit of system voltage, U max is the upper limit of system voltage, U i is the voltage at node i, I min is the lower limit of the line current carrying capacity, I ij is the current flowing through branch ij, I maxP is the upper limit of the line current carrying capacity, min is the lower limit of the equipment power capacity of access node i, P i is the device accessing node i, P max is the upper limit of the device power capacity of access node i, Q min is the lower limit of reactive capacity of the equipment connected to node i, Q i is the reactive power demand of the equipment connected to node i, Q max is the upper limit of reactive power capacity of the equipment connected to node i.

[0176] (3) Constraint 3: Connectivity constraint and ring network discrimination constraint

[0177]

[0178] ∑ j:i→j S ij +F i =∑ k:k→i S ki ,i,j,k∈N / r (7)

[0179] S ij ≤(1-x ij )M (8)

[0180] Formula (6) is the tree-like decoupling constraint of the distribution network topology, which ensures the connectivity of the network after restoration and reconstruction. In it, N is the total number of nodes, i and j are the i-th and j-th nodes in the distribution network, and x ij is the connectivity status of link ij, connected is 1, and disconnected is 0; Equation (7) is the virtual flow KCL constraint of the nodes except the root node r, S ij is the apparent power flowing out of the i-th node, F i is the virtual flow demand of node i, S ki is the virtual flow flowing into node i, k, i, j are all node numbers, r is the root node number. M is a large real number, and formula (8) is a large M constraint.

[0181] Step 2: Establish reconstruction optimization goals

[0182] (1) Reconstruction optimization goal 1: Minimum line loss

[0183] The line loss equations are shown in equations (9) and (10):

[0184]

[0185] Where U i is the voltage at node i; P i and Q i is the active and reactive outflow power of node i; P L,i+1 and Q L,i+1Inject active and reactive power into node i+1; P loss,i and Q loss,i is the active and reactive power loss before node i; P tloss represents the total network loss; N represents the number of nodes; R i is the impedance of the downstream branch.

[0186] (2) Reconstruction optimization goal 2: Minimum number of switch actions and transformer voltage regulation times

[0187]

[0188] Where F is the reconstruction optimization target 2, x1 is the influence weight of the number of switch actions, x2 is the influence weight of the number of transformer voltage regulation, and Y k and Z j are the states of the disconnecting switch and the tie switch after reconstruction, m and n are the numbers of the sectionalizing switch and the tie switch in the network, T a The number of transformer voltage regulators.

[0189] Step S1: Based on the power system network parameters and the pre-built power system network reconstruction model, a binary particle swarm optimization algorithm is used to solve and obtain the optimal solution for the power system network reconstruction, including:

[0190] ① Describe the on / off state of the switch as an array A;

[0191] ②Start from the power supply point to search for the ring network formed by the closing of the tie switch, and record the switch numbers included in the array {b i};

[0192] ③ Disconnect {b i}, making the ring network an open loop; when all the ring networks are open loops, as a particle x i , the particle displacement vector represents the switch state, and the particle velocity vector is the probability of displacement change; and detect the particle x i feasibility;

[0193] ④ Repeat step ③ until the particle size meets the set size requirements;

[0194] ⑤ Start the assisted search between particles, and perform topological verification on the searched optimization results, calculate the fitness value of the optimization results after topological verification, and save the fitness value; if it is better than the historical optimal solution, the optimal solution is updated by the fitness value of the optimization result after topological verification;

[0195] ⑥ Repeat step ⑤ until the maximum number of iterations t is reached max Or achieve the required accuracy, and then output the optimal solution and the topological map corresponding to the optimal solution.

[0196] Step S1 specifically includes:

[0197] Step 3: Model solution

[0198] BPSO converts the variables in the traditional PSO algorithm into binary codes. Compared with the traditional PSO, it is necessary to set a particle velocity threshold to prevent the algorithm from falling into the local optimal solution. In the present invention, BPSO is used to search for feasible networks in the reconstruction problem. The algorithm steps are described as follows, and the solution process is as follows Figure 3 As shown:

[0199] ① Describe the on / off state of the switch as an array A, 0 and 1, representing open and closed;

[0200] ② Generate the adjacent branch matrix and node association matrix of the original network. Start from the power point to search for the ring network formed by the closing of the tie switch, and record the switch numbers they contain into an array {b i};

[0201] ③ Generate particle population: disconnect {b i}, making it an open loop. When all the loops are open loops, the new particle x i The displacement and velocity of the particle are recorded as two vectors: the displacement vector represents the switch state, and the velocity vector affects the probability of displacement change, as shown in Equation (10).

[0202] {x i},{v i}(i=1,2,3...D)(12)

[0203] In the formula, v i is the particle velocity, and D is the particle dimension. During the open loop process, the relationship between the displacement and velocity of each particle is shown in the following formula:

[0204]

[0205] In the formula, x i (t+1) is the particle x at time t+1 i The displacement, v i (t) is the particle x at time t i speed, is a threshold value, usually 0.5 by default. i There are overlaps between the rings in the network, so rules must be formulated to prevent multiple rings from selecting the same switch to open. After the particles are generated, topological analysis is required to check their feasibility.

[0206] ④ Repeat step ③ until the particle size meets the size requirements.

[0207] ⑤ Start the assisted search between particles and perform topological verification on the searched optimization results. After verification, calculate the fitness value and save the evaluation index. If it is better than the historical optimal, update the optimal solution.

[0208] ⑥ Repeat step ⑤ until the maximum number of iterations t is reached max Or achieve the required accuracy, and then output the optimal solution and the corresponding topology map.

[0209] Before S2, it also includes: construction of a communication system fault reconstruction model.

[0210] Furthermore, the construction of the communication system fault reconstruction model includes:

[0211] The communication system fault reconstruction objective function is constructed with the goal of minimizing the sum of the on / off, delay, and error time of the communication between the physical station and the master station;

[0212] Setting a second constraint condition for the communication system fault reconstruction objective function;

[0213] Constructing a communication system fault reconstruction model based on the communication system fault reconstruction objective function and the second constraint condition;

[0214] The second constraint conditions include: communication current carrying capacity constraint and communication terminal power supply constraint.

[0215] Furthermore, the communication system fault reconstruction objective function is as follows:

[0216]

[0217] In the formula, f d (c) is the sum of interruption time, delay time and error time, c is the communication link number, R is the number of routing nodes in the communication layer; B(c) is the on / off status of the c-th link; T(c) is the communication delay of the c-th link; P(c) is the data upload error rate of the c-th link; H(c) is the error rate of the control command issued by the c-th link; α, β, γ are weight coefficients.

[0218] The following is a detailed introduction to the construction of the communication system fault reconstruction model.

[0219] Step 2: Establish the communication system fault reconstruction process and communication status constraints

[0220] Step 1: Communication system fault reconstruction constraints

[0221] (1) Constraint 1: Communication current carrying capacity constraint

[0222]

[0223] Where p(c) is the flow rate of the cth communication link; is the maximum flow of the cth communication link.

[0224] (2) Constraint 2: Communication terminal power supply constraint

[0225] The communication node relies on the physical layer node in that the corresponding communication terminal needs the physical layer node to provide a certain amount of power application. When the physical node connected to the communication node cannot provide enough power and the battery of the communication terminal cannot support it, the communication node will not work properly. This constraint can be described as:

[0226]

[0227] In the formula, τ i is the power supply validity flag of the communication terminal node i; L(i) is the flag bit indicating whether the physical node can supply power to the communication terminal node; soc(ES i ) is the energy storage charge state of communication terminal i.

[0228] Step 2: Communication system fault reconstruction objective function

[0229] In the communication layer of the distribution network, the main consideration is to minimize the on / off, delay, bit error and other aspects of the communication between the physical and master stations.

[0230]

[0231] In the formula, f d (c) is the sum of interruption time, delay time and error time, c is the communication link number, R is the number of routing nodes in the communication layer; B(c) is the on / off status of the c-th link; T(c) is the communication delay of the c-th link; P(c) is the data upload error rate of the c-th link; H(c) is the error rate of the control command issued by the c-th link; α, β, γ are weight coefficients.

[0232] Step S2: Based on the communication system parameters and the pre-built communication system fault reconstruction model, a single-source shortest path combined with a dynamic routing method is used to solve the problem, and the optimal solution for the communication system network reconstruction is obtained, including:

[0233] Abstract the communication network into a graph, and divide all the vertices in the graph into two groups, one group is the vertices of the known shortest path, and the other group is the vertices of the unknown shortest path. Initially, all vertices belong to the vertices of the unknown shortest path, and mark the path length of each vertex as infinite, and set the path length of the starting point to 0;

[0234] In each stage, the vertex with the shortest path length is selected from all the vertices of the unknown shortest path, and the vertex is marked as the vertex of the known shortest path;

[0235] Then, the path length from the vertex of the known shortest path to the vertex of its adjacent vertex belonging to the unknown shortest path is calculated. If the path length is shorter than the original path length, the path length of the adjacent vertex is updated, and the preceding node of the adjacent vertex is set to the selected vertex; the loop operation is continuously performed until the number of vertices of the unknown shortest path is 0; and the communication routing table is output as the optimal solution for the communication system network reconstruction of the communication layer model.

[0236] Step S2 specifically includes:

[0237] Step 3: Model solution

[0238] The communication system is solved by combining the single-source shortest path with the dynamic routing method. The specific process is as follows: Figure 4 As shown:

[0239] (1) First, the communication network is abstracted into a graph. All vertices in the graph are divided into two groups: one group is the vertices of the known shortest path, and the other group is the vertices of the unknown shortest path. Initially, all vertices belong to the vertices of the unknown shortest path, and the path length of each vertex is marked as infinite, and the path length of the starting point is set to 0;

[0240] (2) In each stage, select the vertex with the shortest path length among all the vertices of the unknown shortest path and mark it as the vertex of the known shortest path;

[0241] (3) Then calculate the path length from the vertex to the vertex in its adjacent vertices that belongs to the unknown shortest path. If the obtained path length is shorter than the original length, update the path length of the adjacent vertex and set the preceding node of the adjacent vertex to the selected vertex; continue the loop operation until the number of vertices in the unknown shortest path is 0. Output the communication routing table as the solution of the communication layer model.

[0242] The distribution network reconstruction method based on a large-scale parallel computing architecture provided by the present invention also includes:

[0243] The optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction are verified using a pre-built optimization model.

[0244] Furthermore, the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction are verified using a pre-built optimization model, including:

[0245] Perform preliminary partitioning of the physical network based on the topology of the information network;

[0246] Based on the preliminary zoning, physical network division is carried out;

[0247] After the physical network is divided, the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction are substituted into a pre-built optimization model to obtain the CPU calculation balance under consideration of electrical connection;

[0248] The CPU calculates the degree of balance to verify the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction.

[0249] Furthermore, the preliminary partitioning of the physical network based on the topological structure of the information network includes:

[0250] ① Obtain the equivalent electrical distance between nodes as the node feature;

[0251] ② Classify each node in the power system into a separate category and calculate the distance between categories;

[0252] ③ Initialize the clustering acceleration factor δ;

[0253] ④Find all class pairs whose inter-class distance is less than δ. If there are none, select the class pair with the smallest inter-class distance;

[0254] ⑤ Check whether each class pair is electrically connected in turn. If there is no connecting line, set the electrical distance between the class pairs to be infinite. Otherwise, merge the class pairs into a new class.

[0255] ⑥Update the inter-class distance and δ value;

[0256] ⑦ Repeat steps ④, ⑤, and ⑥ until all nodes are merged into one class;

[0257] ⑧Analyze the statistic F, determine the target number of partitions, and obtain the partition results.

[0258] The following is a detailed description of the verification of the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction using a pre-built optimization model:

[0259] Step 3: Establish large-scale complex network parallel simulation calculation process and simulation verification

[0260] Step 1: Construct a large-scale distribution network decoupling method

[0261] The present invention adopts the agglomerative hierarchical clustering theory to construct a large-scale distribution network decoupling method, and the specific implementation steps are as follows: Figure 5 As shown:

[0262] ① Obtain the equivalent electrical distance between nodes as the node feature;

[0263] ②Divide each node in the power system into a class and calculate the distance between classes. The calculation method of the distance between classes is shown in formula (15);

[0264]

[0265] In the formula, D(K i ,K j ) is the corrected inter-class distance, dis(K i ,K j ) is the distance between classes defined by the centroid distance, N(i), N DG (i) and N(j), N DG (j) are sub-regions K i , K j The number of midpoints and DGs, α, β∈[1.0,1.1].

[0266] ③ Initialize δ, the clustering acceleration factor;

[0267] ④Find all class pairs whose inter-class distance is less than δ. If there is no class pair that meets this condition, select the class pair with the smallest inter-class distance;

[0268] ⑤ Check whether each class pair is electrically connected in turn. If there is no connecting line, set the electrical distance between the class pairs to infinity. If there is a connecting line, merge them into a new class.

[0269] ⑥Update the inter-class distance and δ value;

[0270] ⑦ Repeat steps ④, ⑤, and ⑥ until all nodes are merged into one class;

[0271] ⑧Analyze the F statistic, determine the target number of partitions, and obtain the corresponding partition results.

[0272] Step 2: Construct a large-scale distribution network partition subtask allocation process considering the communication system

[0273] (1) Optimization goal

[0274] 1) Balanced computing power. As an important indicator of CPU performance, the optimal computing power of a single core can ensure the balance between simulation time and computing power within the neighborhood of this optimal computing power. When the CPU single core load exceeds the optimal boundary value, the simulation time increases sharply with the increase in computing power, so the limitation of CPU computing power must be considered when partitioning large-scale distribution networks. Since the computing power of different components in transient simulation varies greatly, it is necessary to accurately measure the computing power differences of different types of components in the transient simulation process and allocate CPU cores.

[0275] 2) Consider electrical connections. In order to achieve load transfer under fault conditions, multiple outgoing lines in the same substation are often connected at the end through a connecting switch, resulting in strong electrical connections between feeders, which are difficult to decouple naturally. Networks that cannot be decoupled can only use serial mode during the simulation process, which greatly occupies computing resources. In order to facilitate the elimination of the coupling relationship between multi-core computing modules, this project reflects the feeder partitioning of the distribution network in the parallel computing core splitting technology, so that the outgoing lines with electrical connections are numbered adjacently, and an item is introduced in the optimization objective to reflect the difference in the feeder numbers of the same core calculation.

[0276] (2) Optimization model

[0277] Based on the network division principle mentioned above, network division is required when optimizing subtask allocation, which is divided into two steps: 1) Preliminary partitioning of the physical network based on the topological structure of the information network; 2) Further physical network division based on the preliminary partitioning. Modeling is performed for the two optimization objectives mentioned above, the CPU computing balance and the electrical connection of the distribution network itself. The first priority is the utilization of the computing power of each CPU to ensure the computing balance of each CPU; the second priority is to divide the feeders of the same substation into the same CPU as much as possible for calculation. The objective function is as follows:

[0278]

[0279] Where: It measures the normalized computing power of the ith CPU; (m i,end -m i+1,st ) measures the feeder cross-substation situation calculated by the i-th CPU after normalization; ω1 is the weight coefficient of the first-level optimization target; ω2 is the weight coefficient of the second-level optimization target, ω1>ω2>0; is the equivalent load number of the ideal calculation of the i-th CPU; n cal,i is the equivalent load number actually calculated by the i-th CPU; m i,st is the substation number of the first feeder calculated by the i-th CPU; m i,end is the substation number of the last feeder calculated by the i-th CPU; M is the total number of substations.

[0280] When all CPUs have the same performance, the optimization results of the above model ensure that the number of nodes calculated by each CPU is the same or close to the number of nodes; when the CPU performance is different, the optimization results ensure that the number of nodes calculated by each CPU is the same or close to the ideal number of nodes. The constraints that must be met at the same time are:

[0281] 1) The first feeder number calculated by the next CPU minus the last feeder number calculated by the previous CPU

[0282] k i+1,st -k i,end =1(19)

[0283] And the sum of all feeder line numbers calculated by all CPUs is equal to the total feeder line number.

[0284] 2) The actual load calculated by the i-th CPU cannot exceed the upper limit of the ideal load

[0285]

[0286] 3) The first feeder calculated by the first CPU must be the feeder numbered 1 and belongs to the first substation

[0287] m 1,st =1(21)

[0288] k 1,st =1(22)

[0289] Nth CPU The last feeder calculated by a CPU must be numbered N. L The feeder belongs to the last substation

[0290]

[0291] 4) Partition constraint variable A j

[0292]

[0293] Variable Description: Indicates the Nth CPU The last feeder calculated by the CPU, Indicates the Nth L Feeder lines, Indicates the Nth CPU +1 CPU calculation for the first feeder, Indicates the Nth L+1 Feeder lines, Indicates the Nth CPU The last substation calculated by the CPU, Indicates the Nth CPU+1 The first substation calculated by the CPU, N C Indicates the number of preliminary partitions; L C,j represents the feeder set of the jth partition after preliminary partitioning; l k represents the kth feeder; N L is the total number of feeders; N CPU is the total number of CPUs; is the equivalent load number of the ideal calculation of the i-th CPU; n cal,iis the equivalent load number actually calculated by the i-th CPU; the load number calculated by the i-th CPU is equal to the constraint m i,st is the substation number of the first feeder calculated by the i-th CPU; m i,end is the substation number of the last feeder calculated by the i-th CPU; k i,st is the first feeder number calculated by the i-th CPU; k i,end is the number of the last feeder calculated by the i-th CPU; M is the total number of substations; n line,i is the number of feeder lines calculated by the i-th CPU; L CPU,i is the feeder set calculated by the i-th CPU; the sum of all elements of the feeder sets of all CPUs is N L ; is the load number of the kth feeder; is the number of distributed generation sources on the kth feeder; a is the ratio of the load to the upper limit of the computing capacity of the unit; A j is the partition constraint variable.

[0294] Step 3: Parallel computing framework

[0295] In the MPI system process, call the required functions and applications, and finally use MPI_Finalize() to end the process. The specific process is as follows: Figure 6 shown.

[0296] The user submits the program to be executed or the problem to be processed to the cloud computing platform Hadoop, and the subtasks are executed in parallel. Figure 7 As shown in the figure. In this model, Map is executed near the data input as much as possible to avoid large amounts of data movement. Map images distribute instructions to multiple workers, and reduce specifications merge the results calculated by Map workers. The Master periodically detects the workers and re-executes Map when a failure occurs. If Reduce is not completed, it is re-executed. During execution, the task granularity generally requires that the small data block is less than or equal to the size of a Block in HDFS, and that the number of Maps and Reduces is greater than the number of workers to facilitate load balancing and fault recovery.

[0297] Step 4: End.

[0298] The present invention constructs a reconstruction optimization system that covers multi-dimensional constraints such as voltage qualification rate, equipment / line / distributed adjustable resources, ring network constraints, and multi-dimensional optimization targets such as line loss, number of switch actions, and number of transformer voltage regulation. It also proposes a reconstruction model solution method based on an improved particle swarm algorithm, and constrains the formation of a ring network through the natural advantages of the algorithm. It also proposes a topology verification method, and fully considers the method of island power restoration when an island is formed.

[0299] A reconstruction algorithm for a large-scale distribution network corresponding to a communication network, the main steps of which are as follows:

[0300] (1) A communication network reconstruction optimization objective covering multi-dimensional optimization objectives such as communication delay, communication interruption, and communication error is proposed;

[0301] (2) Constraints on the power-on status and battery reserve status of communication equipment are proposed, and these are used as interactive data to achieve embedding and iteration with the power system power restoration algorithm.

[0302] (3) A distribution network partition decoupling method based on agglomerative hierarchical clustering theory is proposed. Furthermore, considering the communication network partition, a subtask decomposition method that takes into account both the communication network and the power grid is proposed.

[0303] (4) An MPI computing implementation method and system suitable for parallel computing of communication network and power grid subtasks are proposed.

[0304] The present invention proposes a distribution network reconstruction system based on a large-scale parallel computing architecture. This method fully considers the failures, risks, and optimization states of the distribution network and the communication network, and realizes the generation of network reconstruction strategies under the mutual influence of the communication network and the power network. It has strong applicability and feasibility for actual sites. In addition, the present invention also provides an algorithm and implementation process suitable for large-scale complex distribution network simulation calculations, supports online dynamic data access, meets the verification and deduction requirements of the reconstruction strategy, and well guarantees the accuracy and real-time performance of the reconstruction strategy.

[0305] Embodiment 2:

[0306] The present invention based on the same inventive concept also provides a distribution network reconstruction system based on a large-scale parallel computing architecture, including:

[0307] The power system reconstruction module is used to obtain the optimal solution for power system network reconstruction based on power system network parameters and a pre-built power system network reconstruction model using a binary particle swarm optimization algorithm;

[0308] The network system reconstruction module is used to solve the communication system network reconstruction optimal solution based on the communication system parameters and the pre-built communication system fault reconstruction model by combining the single-source shortest path with the dynamic routing method;

[0309] The reconstruction scheme module is used to form a distribution network reconstruction scheme based on the optimal solution for the reconstruction of the power system network and the optimal solution for the reconstruction of the communication system network.

[0310] Optionally, a power system reconstruction module is further included, for:

[0311] The first reconstruction optimization objective is established with minimum line loss;

[0312] The second reconstruction optimization objective is to minimize the number of switch actions and transformer voltage regulation times;

[0313] Setting a first constraint condition for the first reconstruction optimization target and the second reconstruction optimization target;

[0314] Constructing a power system network reconstruction model based on the first reconstruction optimization objective and the second reconstruction optimization objective and constraint conditions;

[0315] The first constraint condition includes: power flow constraint, capacity constraint, connectivity constraint and ring network identification constraint.

[0316] Optionally, the first reconstruction optimization objective is as follows:

[0317]

[0318] Where U i is the node voltage; P i and Q i is the active and reactive outflow power of node i; P L,i+1 and Q L,i+1 Inject active and reactive power into node i+1; P loss,i and Q loss,i is the active and reactive power loss before node i; P tloss represents the total network loss; N represents the number of nodes.

[0319] Optionally, the second reconstruction optimization objective is as follows:

[0320]

[0321] Where F is the second reconstruction optimization target, x1 is the influence weight of the number of switch actions, x2 is the influence weight of the number of transformer voltage regulation, and Y k and Z j are the states of the disconnecting switch and the tie switch after reconstruction, m and n are the numbers of the sectionalizing switch and the tie switch in the network, T a The number of transformer voltage regulators.

[0322] Optionally, the power system reconstruction module is specifically used to:

[0323] ① Describe the on / off state of the switch as an array A;

[0324] ②Start from the power supply point to search for the ring network formed by the closing of the tie switch, and record the switch numbers included in the array {b i};

[0325] ③ Disconnect {b i}, making the ring network an open loop; when all the ring networks are open loops, as a particle x i , the particle displacement vector represents the switch state, and the particle velocity vector is the probability of displacement change; and detect the particle x i feasibility;

[0326] ④ Repeat step ③ until the particle size meets the set size requirements;

[0327] ⑤ Start the assisted search between particles, and perform topological verification on the searched optimization results, calculate the fitness value of the optimization results after topological verification, and save the fitness value; if it is better than the historical optimal solution, the optimal solution is updated by the fitness value of the optimization result after topological verification;

[0328] ⑥ Repeat step ⑤ until the maximum number of iterations t is reached max Or achieve the required accuracy, and then output the optimal solution and the topological map corresponding to the optimal solution.

[0329] Optionally, a network model building module is also included, for:

[0330] The communication system fault reconstruction objective function is constructed with the goal of minimizing the sum of the on / off, delay, and error time of the communication between the physical station and the master station;

[0331] Setting a second constraint condition for the communication system fault reconstruction objective function;

[0332] Constructing a communication system fault reconstruction model based on the communication system fault reconstruction objective function and the second constraint condition;

[0333] The second constraint conditions include: communication current carrying capacity constraint and communication terminal power supply constraint.

[0334] Optionally, the communication system fault reconstruction objective function is as follows:

[0335]

[0336] In the formula, f d (c) is the sum of interruption time, delay time and error time, c is the communication link number, R is the number of routing nodes in the communication layer; B(c) is the on / off status of the c-th link; T(c) is the communication delay of the c-th link; P(c) is the data upload error rate of the c-th link; H(c) is the error rate of the control command issued by the c-th link; α, β, γ are weight coefficients.

[0337] Optionally, the network system reconstruction module is specifically used to:

[0338] Abstract the communication network into a graph, and divide all the vertices in the graph into two groups, one group is the vertices of the known shortest path, and the other group is the vertices of the unknown shortest path. Initially, all vertices belong to the vertices of the unknown shortest path, and mark the path length of each vertex as infinite, and set the path length of the starting point to 0;

[0339] In each stage, the vertex with the shortest path length is selected from all the vertices of the unknown shortest path, and the vertex is marked as the vertex of the known shortest path;

[0340] Then, the path length from the vertex of the known shortest path to the vertex of its adjacent vertex belonging to the unknown shortest path is calculated. If the obtained path length is shorter than the original length, the path length of the adjacent vertex is updated, and the preceding node of the adjacent vertex is set to the selected vertex; the loop operation is continuously performed until the number of vertices of the unknown shortest path is 0; and the communication routing table is output as the optimal solution for the communication system network reconstruction of the communication layer model.

[0341] Optionally, it also includes: a verification module, which is used to verify the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction using a pre-built decoupling model.

[0342] Optionally, the verification module is specifically used to:

[0343] Perform preliminary partitioning of the physical network based on the topology of the information network;

[0344] Based on the preliminary zoning, physical network division is carried out;

[0345] After the physical network is divided, the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction are substituted into a pre-built optimization model to obtain the CPU calculation balance under consideration of electrical connection;

[0346] The CPU calculates the degree of balance to verify the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction.

[0347] Optionally, an optimization model building module is also included, for:

[0348] Construct the verification objective function with the goal of minimizing CPU computing balance;

[0349] Setting verification constraints for the verification objective function;

[0350] The verification constraints include: the first feeder number calculated by the next CPU minus the last feeder number calculated by the previous CPU, the actual load number calculated by the i-th CPU cannot exceed the upper limit of the ideal load number, and the first feeder calculated by the first CPU must be the feeder numbered 1 and belongs to the first substation and partition variable constraints.

[0351] Optionally, the verification objective function is as shown below:

[0352]

[0353] Where: It measures the normalized computing power of the ith CPU; (m i,end -m i+1,st ) measures the feeder cross-substation situation calculated by the i-th CPU after normalization; ω1 is the weight coefficient of the first-level optimization target; ω2 is the weight coefficient of the second-level optimization target, ω1>ω2>0; is the equivalent load number of the ideal calculation of the i-th CPU; n cal,i is the equivalent load number actually calculated by the i-th CPU; m i,st is the substation number of the first feeder calculated by the i-th CPU; m i+1,st The substation number of the first feeder calculated by the i+1th CPU; m i,end is the substation number of the last feeder calculated by the i-th CPU; M is the total number of substations, Fun is the CPU calculation balance, N CPU Indicates the CPU number.

[0354] Furthermore, the verification module performs preliminary partitioning of the physical network based on the topological structure of the information network, and the specific implementation steps include:

[0355] ① Obtain the equivalent electrical distance between nodes as the node feature;

[0356] ② Classify each node in the power system into a separate category and calculate the distance between categories;

[0357] ③ Initialize the clustering acceleration factor δ;

[0358] ④Find all class pairs whose inter-class distance is less than δ. If there are none, select the class pair with the smallest inter-class distance;

[0359] ⑤ Check whether each class pair is electrically connected in turn. If there is no connecting line, set the electrical distance between the class pairs to be infinite. Otherwise, merge the class pairs into a new class.

[0360] ⑥Update the inter-class distance and δ value;

[0361] ⑦ Repeat steps ④, ⑤, and ⑥ until all nodes are merged into one class;

[0362] ⑧Analyze the statistic F, determine the target number of partitions, and obtain the partition results.

[0363] Furthermore, after the physical network is divided, the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction are substituted into a pre-built optimization model to obtain the CPU calculation balance considering the electrical connection, including:

[0364] Initialize the MPI environment and treat the verification of each physical sub-network as a subtask, with one subtask corresponding to one instruction;

[0365] Map images distribute instructions to multiple workers. The task granularity of workers during execution generally requires that the small data block is less than or equal to the size of a block in HDFS, and the number of Maps and Reduces is greater than the number of workers.

[0366] The reduce specification combines the results calculated by the Map workers;

[0367] The Master periodically detects the Worker and re-executes the Map if a failure occurs. If the Reduce is not completed, it is re-executed.

[0368] Example 3

[0369] like Figure 8 As shown, the present invention also provides an electronic device, which may be a computer device, a single-chip device, an intelligent mobile device, etc. The electronic device in this embodiment may include a processor, a memory, a transceiver component, etc. The memory, the processor, and the transceiver component are connected via a bus; the memory may be used to store an execution program, and an exemplary execution program may include instructions; the processor is used to execute the instructions stored in the memory. The memory may also be used to store data, which may be called and / or modified when the instructions are executed.

[0370] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in a storage medium to implement the corresponding method flow or corresponding functions, so as to realize the steps of the distribution network reconstruction method based on a large-scale parallel computing architecture in the above-mentioned embodiment.

[0371] Example 4

[0372] Based on the same inventive concept, the present invention also provides a readable storage medium, specifically an electronic device readable storage medium (Memory), which is a memory device in an electronic device for storing programs and data. It can be understood that the storage medium here can include both built-in storage media in electronic devices and, of course, extended storage media supported by electronic devices. The storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions can be one or more execution programs (including program codes). It should be noted that the storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor loads and executes one or more instructions stored in the storage medium, which can implement the steps of the distribution network reconstruction method based on a large-scale parallel computing architecture in the above embodiment.

[0373] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0374] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0375] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0376] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0377] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are included in the scope of the claims of the present invention to be approved.

Claims

1. A distribution network reconstruction method based on a large-scale parallel computing architecture, characterized in that: include: Based on the power system network parameters and the pre-built power system network reconstruction model, the binary particle swarm optimization algorithm is used to solve the optimal solution for the power system network reconstruction; Based on the communication system parameters and the pre-built communication system fault reconstruction model, the single-source shortest path and dynamic routing methods are combined to obtain the optimal solution for the communication system network reconstruction. The distribution network reconstruction plan is composed of the power system network reconstruction optimal solution and the communication system network reconstruction optimal solution.

2. The method according to claim 1, characterized in that The construction of the power system network reconstruction model includes: The first reconstruction optimization objective is established with minimum line loss; The second reconstruction optimization objective is to minimize the number of switch actions and transformer voltage regulation times; Setting a first constraint condition for the first reconstruction optimization target and the second reconstruction optimization target; Constructing a power system network reconstruction model based on the first reconstruction optimization objective and the second reconstruction optimization objective and constraint conditions; The first constraint condition includes: power flow constraint, capacity constraint, connectivity constraint and ring network identification constraint.

3. The method according to claim 2, characterized in that The first reconstruction optimization objective is as follows: Where U i is the node voltage; P i and Q i is the active and reactive outflow power of node i; P i+1 and Q i+1 is the active and reactive outflow power of node i+1; P L,i+1 and Q L,i+1 Inject active and reactive power into node i+1; P loss,i and Q loss,i is the active and reactive power loss before node i; P tloss represents the total network loss; N represents the number of nodes; P loss,i～(i+1) is the line loss from node i to node i+1; R i is the resistance of node i.

4. The method according to claim 2, characterized in that The second reconstruction optimization objective is as follows: Where F is the second reconstruction optimization target, x1 is the influence weight of the number of switch actions, x2 is the influence weight of the number of transformer voltage regulation, and Y k and Z j are the states of the disconnecting switch and the tie switch after reconstruction, k and j are the numbers of the section switches and tie switches in the network, m and n are the numbers of the section switches and tie switches in the network, T a is the transformer voltage regulating volume number, a is the transformer serial number, and nb is the total number of transformers.

5. The method according to claim 1, characterized in that The method is based on the power system network parameters and the pre-built power system network reconstruction model, and adopts the binary particle swarm optimization algorithm to solve and obtain the optimal solution for the power system network reconstruction, including: ① Describe the on / off state of the switch as an array A; ②Start from the power supply point to search for the ring network formed by the closing of the tie switch, and record the switch numbers included in the array {b i }; ③ Disconnect {b i }, making the ring network an open loop; when all the ring networks are open loops, as a particle x i , the particle displacement vector represents the switch state, and the particle velocity vector is the probability of displacement change; and detect the particle x i feasibility; ④ Repeat step ③ until the particle size meets the set size requirements; ⑤ Start the assisted search between particles, and perform topological verification on the searched optimization results, calculate the fitness value of the optimization results after topological verification, and save the fitness value; if it is better than the historical optimal solution, the optimal solution is updated by the fitness value of the optimization result after topological verification; ⑥ Repeat step ⑤ until the maximum number of iterations t is reached max Or achieve the required accuracy, and then output the optimal solution and the topological map corresponding to the optimal solution.

6. The method according to claim 1, characterized in that The construction of the communication system fault reconstruction model includes: The communication system fault reconstruction objective function is constructed with the goal of minimizing the sum of the on / off, delay, and error time of the communication between the physical station and the master station; Setting a second constraint condition for the communication system fault reconstruction objective function; Constructing a communication system fault reconstruction model based on the communication system fault reconstruction objective function and the second constraint condition; The second constraint conditions include: communication current carrying capacity constraint and communication terminal power supply constraint.

7. The method according to claim 6, characterized in that The communication system fault reconstruction objective function is as follows: In the formula, f d (c) is the sum of interruption time, delay time and error time, c is the communication link number, R is the number of routing nodes in the communication layer; B(c) is the on / off status of the c-th link; T(c) is the communication delay of the c-th link; P(c) is the data upload error rate of the c-th link; H(c) is the error rate of the control command issued by the c-th link; α, β, γ are weight coefficients.

8. The method according to claim 1, characterized in that The communication system fault reconstruction model based on the communication system parameters and the pre-built communication system fault reconstruction model is solved by combining the single-source shortest path with the dynamic routing method to obtain the optimal solution for the communication system network reconstruction, including: Abstract the communication network into a graph, and divide all the vertices in the graph into two groups, one group is the vertices of the known shortest path, and the other group is the vertices of the unknown shortest path. Initially, all vertices belong to the vertices of the unknown shortest path, and mark the path length of each vertex as infinite, and set the path length of the starting point to 0; In each stage, the vertex with the shortest path length is selected from all the vertices of the unknown shortest path, and the vertex is marked as the vertex of the known shortest path; Then, the path length from the vertex of the known shortest path to the vertex of its adjacent vertex belonging to the unknown shortest path is calculated. If the path length is shorter than the original path length, the path length of the adjacent vertex is updated, and the preceding node of the adjacent vertex is set to the selected vertex; the loop operation is continuously performed until the number of vertices of the unknown shortest path is 0; and the communication routing table is output as the optimal solution for the communication system network reconstruction of the communication layer model.

9. The method according to claim 1, characterized in that Also includes: The optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction are verified using a pre-built optimization model.

10. The method according to claim 9, characterized in that The optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction are verified using a pre-built optimization model, including: Perform preliminary partitioning of the physical network based on the topology of the information network; Based on the preliminary zoning, physical network division is carried out; After the physical network is divided, the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction are substituted into a pre-built optimization model to obtain the CPU calculation balance considering the electrical connection; The CPU calculates the degree of balance to verify the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction.

11. The method according to claim 10, characterized in that The construction of the optimization model includes: Construct the verification objective function with the goal of minimizing CPU computing balance; Setting verification constraints for the verification objective function; The verification constraints include: the first feeder number calculated by the next CPU minus the last feeder number calculated by the previous CPU, the actual load number calculated by the i-th CPU cannot exceed the upper limit of the ideal load number, and the first feeder calculated by the first CPU must be the feeder numbered 1 and belongs to the first substation and partition variable constraints.

12. The method according to claim 11, characterized in that The verification objective function is shown in the following formula: Where: It measures the normalized computing power of the ith CPU; (m i,end -m i+1,st ) measures the feeder cross-substation situation calculated by the i-th CPU after normalization; ω1 is the weight coefficient of the first-level optimization target; ω2 is the weight coefficient of the second-level optimization target, ω1>ω2>0; is the equivalent load number of the ideal calculation of the i-th CPU; n cal,i is the equivalent load number actually calculated by the i-th CPU; m i,st is the substation number of the first feeder calculated by the i-th CPU; m i+1,st The substation number of the first feeder calculated by the i+1th CPU; m i,end is the substation number of the last feeder calculated by the i-th CPU; M is the total number of substations, Fun is the CPU calculation balance, N CPU Indicates the CPU number.

13. The method according to claim 10, characterized in that The preliminary partitioning of the physical network based on the topological structure of the information network includes: ① Obtain the equivalent electrical distance between nodes as the node feature; ② Classify each node in the power system into a separate category and calculate the distance between categories; ③ Initialize the clustering acceleration factor δ; ④Find all class pairs whose inter-class distance is less than δ. If there are none, select the class pair with the smallest inter-class distance; ⑤ Check whether each class pair is electrically connected in turn. If there is no connecting line, set the electrical distance between the class pairs to be infinite. Otherwise, merge the class pairs into a new class. ⑥Update the inter-class distance and δ value; ⑦ Repeat steps ④, ⑤, and ⑥ until all nodes are merged into one class; ⑧Analyze the statistic F, determine the target number of partitions, and obtain the partition results.

14. A distribution network reconstruction system based on a large-scale parallel computing architecture, characterized in that: include: The power system reconstruction module is used to obtain the optimal solution for power system network reconstruction based on power system network parameters and a pre-built power system network reconstruction model using a binary particle swarm optimization algorithm; The network system reconstruction module is used to solve the communication system network reconstruction optimal solution based on the communication system parameters and the pre-built communication system fault reconstruction model by combining the single-source shortest path with the dynamic routing method; The reconstruction scheme module is used to form a distribution network reconstruction scheme based on the optimal solution for the reconstruction of the power system network and the optimal solution for the reconstruction of the communication system network.

15. The system of claim 14, wherein: Also included is a power model reconstruction module for: The first reconstruction optimization objective is established with minimum line loss; The second reconstruction optimization objective is to minimize the number of switch actions and transformer voltage regulation times; Setting a first constraint condition for the first reconstruction optimization target and the second reconstruction optimization target; Constructing a power system network reconstruction model based on the first reconstruction optimization objective and the second reconstruction optimization objective and constraint conditions; The first constraint condition includes: power flow constraint, capacity constraint, connectivity constraint and ring network identification constraint.

16. The system of claim 15, wherein: The first reconstruction optimization objective is as follows: Where U i is the node voltage; P i and Q i is the active and reactive outflow power of node i; P i+1 and Q i+1 is the active and reactive outflow power of node i+1; P L,i+1 and Q L,i+1 Inject active and reactive power into node i+1; P loss,i and Q loss,i is the active and reactive power loss before node i; P tloss represents the total network loss; N represents the number of nodes; P loss,i～(i+1) is the line loss from node i to node i+1; R i is the resistance of node i.

17. The system of claim 15, wherein: The second reconstruction optimization objective is as follows: Where F is the second reconstruction optimization target, x1 is the influence weight of the number of switch actions, x2 is the influence weight of the number of transformer voltage regulation, and Y k and Z j are the states of the disconnecting switch and the tie switch after reconstruction, k and j are the numbers of the section switches and tie switches in the network, m and n are the numbers of the section switches and tie switches in the network, T a is the transformer voltage regulating volume number, a is the transformer serial number, and nb is the total number of transformers.

18. The system of claim 14, wherein: The power system reconstruction module is specifically used for: ① Describe the on / off state of the switch as an array A; ②Start from the power supply point to search for the ring network formed by the closing of the tie switch, and record the switch numbers included in the array {b i }; ③ Disconnect {b i }, making the ring network an open loop; when all the ring networks are open loops, as a particle x i , the particle displacement vector represents the switch state, and the particle velocity vector is the probability of displacement change; and detect the particle x i feasibility; ④ Repeat step ③ until the particle size meets the set size requirements; ⑤ Start the assisted search between particles, and perform topological verification on the searched optimization results, calculate the fitness value of the optimization results after topological verification, and save the fitness value; if it is better than the historical optimal solution, the optimal solution is updated by the fitness value of the optimization result after topological verification; ⑥ Repeat step ⑤ until the maximum number of iterations t is reached max Or achieve the required accuracy, and then output the optimal solution and the topological map corresponding to the optimal solution.

19. The system of claim 14, wherein: Also included are network model building modules for: The communication system fault reconstruction objective function is constructed with the goal of minimizing the sum of the on / off, delay, and error time of the communication between the physical station and the master station; Setting a second constraint condition for the communication system fault reconstruction objective function; Constructing a communication system fault reconstruction model based on the communication system fault reconstruction objective function and the second constraint condition; The second constraint conditions include: communication current carrying capacity constraint and communication terminal power supply constraint.

20. The system of claim 19, wherein: The communication system fault reconstruction objective function is as follows: In the formula, f d (c) is the sum of interruption time, delay time and error time, c is the communication link number, R is the number of routing nodes in the communication layer; B(c) is the on / off status of the c-th link; T(c) is the communication delay of the c-th link; P(c) is the data upload error rate of the c-th link; H(c) is the error rate of the control command issued by the c-th link; α, β, γ are weight coefficients.

21. The system of claim 14, wherein: The network system reconstruction module is specifically used for: Abstract the communication network into a graph, and divide all the vertices in the graph into two groups, one group is the vertices of the known shortest path, and the other group is the vertices of the unknown shortest path. Initially, all vertices belong to the vertices of the unknown shortest path, and mark the path length of each vertex as infinite, and set the path length of the starting point to 0; In each stage, the vertex with the shortest path length is selected from all the vertices of the unknown shortest path, and the vertex is marked as the vertex of the known shortest path; Then, the path length from the vertex of the known shortest path to the vertex of its adjacent vertex belonging to the unknown shortest path is calculated. If the obtained path length is shorter than the original length, the path length of the adjacent vertex is updated, and the preceding node of the adjacent vertex is set to the selected vertex; The loop operation is continuously performed until the number of vertices of the unknown shortest path is 0; the communication routing table is output as the optimal solution for the communication system network reconstruction of the communication layer model.

22. The system of claim 14, wherein: Also includes: The verification module is used to verify the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction using a pre-built decoupling model.

23. The system of claim 22, wherein: The verification module is specifically used for: Perform preliminary partitioning of the physical network based on the topology of the information network; Based on the preliminary zoning, physical network division is carried out; After the physical network is divided, the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction are substituted into a pre-built optimization model to obtain the CPU calculation balance considering the electrical connection; The CPU calculates the degree of balance to verify the optimal solution for the power system network reconstruction and the optimal solution for the communication system network reconstruction.

24. An electronic device, characterized in that: include: at least one processor and memory; The memory and the processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, the distribution network reconstruction method based on a large-scale parallel computing architecture as described in any one of claims 1 to 13 is implemented.

25. A readable storage medium, characterized in that: An execution program is stored thereon, and when the execution program is executed, the distribution network reconstruction method based on a large-scale parallel computing architecture as described in any one of claims 1 to 13 is implemented.

Images