Power distribution network fault reconstruction method and device and storage medium

Abstract

This application relates to a distribution network fault reconfiguration method, device, and storage medium, belonging to the field of power grid fault reconfiguration technology. The method includes: generating initial particles from the state of power grid line switches; dividing the initial particles using a hierarchical clustering method to obtain particles that meet preset conditions; recording the individual optimal and global optimal values ​​of the particles and updating their positions; iterating the particle positions based on the updated positions; obtaining the fitness values ​​of particles that have met the required number of iterations according to the fitness value calculation formula; determining whether the constraints are met during the above calculation process; if met, outputting the particle's fitness value and current particle position, and executing a power grid switch operation; if not met, iterating again until the constraints are met, outputting the particle's fitness value and current particle position, and executing a power grid switch operation. This addresses the problems of existing fault reconfiguration methods lacking dynamism, having low automation, and failing to meet the requirements of dynamic reconfiguration.

Description

Technical Field

[0001] This application belongs to the field of power grid fault reconfiguration technology, specifically involving distribution network fault reconfiguration methods, devices, and storage media. Background Technology

[0002] Most current literature focuses on static distribution network fault reconfiguration schemes at a specific time point. Static reconfiguration lacks dynamism, has low automation, and cannot meet the requirements of dynamic reconfiguration. In actual distribution systems, the load at each load node changes continuously over time. Furthermore, the output of distributed wind power turbines is affected by environmental factors and also varies over time. This means that the optimal network structure obtained by static reconfiguration algorithms is only optimal at the time the algorithm takes data. In addition, distributed wind power output is stochastic, while most current literature treats wind turbines as power sources with constant output. Therefore, it is essential to study an overall optimization scheme for distribution network fault reconfiguration that incorporates stochastic output characteristics. Summary of the Invention

[0003] To address this, this application provides a distribution network fault reconfiguration method, apparatus, and storage medium. In this application, considering the influence of various random factors, a stochastic power flow calculation method based on semi-invariant methods is adopted, the concept of hierarchical clustering is introduced, and an improved particle swarm optimization algorithm is used to restore and reconfigure the distribution network. This method considers particles that are currently second only to the optimal ones but may represent information about other exploration directions. This is intended to solve the problems that existing fault reconfiguration methods lack dynamism, have low automation, and cannot meet the requirements of dynamic reconfiguration.

[0004] To achieve the above objectives, this application adopts the following technical solution:

[0005] Distribution network fault reconstruction method, the reconstruction method includes:

[0006] Generate initial particles from the state of the power grid line switches;

[0007] The initial particles are divided using a hierarchical clustering method to obtain particles that meet preset conditions;

[0008] Record the individual optimal and global optimal values ​​of the particles and update the positions of the particles. Based on the updated positions of the particles, the particles with updated positions are divided again using the hierarchical clustering method. Repeat the above iterative process until the preset number of iterations is met.

[0009] The fitness value of a particle that meets the preset number of iterations is obtained according to the preset fitness value calculation formula. It is then determined whether the above calculation process meets the constraints. If it does, the fitness value of the particle and the particle position at this time are output, and the power grid switch operation is executed. If it does not meet the constraints, the iteration is performed again until the constraints are met, the fitness value of the particle and the particle position at this time are output, and the power grid switch operation is executed.

[0010] Furthermore, the step of using hierarchical clustering to divide the initial particles to obtain particles that meet the preset conditions includes: taking all the initial particles as a dataset of samples, taking each sample in the dataset as a separate cluster, calculating the distance between two clusters, finding the two closest clusters, merging these two clusters into one, and repeating this process until the preset number of clusters or threshold is reached.

[0011] Furthermore, the fitness value calculation formula is determined by the power outage load index, the network loss index, and the number of switching operations index, wherein the network loss index is determined by the expected total network loss value.

[0012] Furthermore, obtaining the expected total network loss includes:

[0013] Acquire power sample data, which includes the active load of the power grid, reactive load, wind turbine output, and feasible distribution network topology that meets the radial pattern.

[0014] The power sample data is input into a deep neural network model, which calculates the node voltage and the expected total network loss based on the sample data using a semi-invariant method for stochastic power flow calculation.

[0015] Furthermore, obtaining the feasible radial distribution network topology includes:

[0016] Read the network structure of the faulty system that has been located and isolated;

[0017] In the network structure, disconnect one switch in each loop, that is, the number of disconnected switches is equal to the number of loops;

[0018] Close the branch in the network structure that is directly connected to the power supply;

[0019] Close all branches in the network structure that do not belong to any loop;

[0020] Based on the characteristics of graph theory and tree structure, the connectivity of the network structure is determined, resulting in a series of feasible distribution network topologies that satisfy the radial shape.

[0021] Furthermore, the deep neural network model obtains the node voltage and the expected value of total network loss based on the sample data using a semi-invariant method for stochastic power flow calculation, including:

[0022] Calculate the power flow distribution under normal operating conditions of the power grid to obtain the node voltage state vector, the power vector of the branch under the reference operating point, the Jacobian matrix, and the sensitivity matrix.

[0023] The semi-invariants of the node state vector disturbance part ΔX and the semi-invariants of the branch power flow vector disturbance part ΔZ are calculated based on the node voltage state vector at the reference operating point, the branch power vector at the reference operating point, the Jacobian matrix, and the sensitivity matrix.

[0024] The probability distribution functions and probability density functions of each order of semi-invariants of ΔX and ΔZ are obtained by Gram-Charlier series expansion.

[0025] By shifting ΔX and ΔZ by X0 and Z0 units, we obtain the probability distribution function and probability density function of the node voltage and branch power.

[0026] The expected value of each output state variable, i.e. the expected value of the total network loss, is obtained based on the probability distribution function and probability density function of the branch power.

[0027] Furthermore, the constraints include: distributed wind power output constraints, node voltage constraints, network topology constraints, power flow balance constraints, and switching operation count constraints.

[0028] Furthermore, the node voltage constraint is generated based on the node voltage.

[0029] Distribution network fault reconfiguration device, the device comprising:

[0030] Initial particle generation module: used to generate initial particles from the state of the power grid line switches;

[0031] Particle partitioning module: used to partition the initial particles using a hierarchical clustering method to obtain particles that meet preset conditions;

[0032] Particle Iteration Module: Records the individual optimal and global optimal values ​​of particles and updates the positions of the particles. Based on the updated positions of the particles, the hierarchical clustering method is used to further divide the updated particles. The above iterative process is repeated until the preset number of iterations is met.

[0033] Result output module: Obtain the fitness value of the particle that meets the preset number of iterations according to the preset fitness value calculation formula, determine whether the above calculation process meets the constraint conditions, if it meets the constraint conditions, output the fitness value of the particle and the particle position at this time, and execute the grid switch operation, if it does not meet the constraint conditions, it iterate again until the constraint conditions are met, output the fitness value of the particle and the particle position at this time, and execute the grid switch operation.

[0034] A storage medium storing a computer program, which, when executed by a processor, implements the various steps in the power distribution network fault reconfiguration method.

[0035] The application employs the above technical solution and has at least the following beneficial effects:

[0036] This application generates initial particles from the state of power grid line switches, divides these initial particles using a hierarchical clustering method, iterates through the particles, and calculates their fitness values. The calculation process satisfies preset constraints, which make the fitness values ​​more closely reflect the actual state of the power grid, resulting in more accurate power grid switching results. An improved particle swarm optimization algorithm is used to restore and reconstruct the distribution network, considering particles that are currently only slightly below the optimal level but may represent information for other exploration directions. This addresses the problems of existing fault reconstruction methods lacking dynamism, having low automation, and failing to meet the requirements of dynamic reconstruction.

[0037] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a flowchart illustrating a distribution network fault reconfiguration method according to an exemplary embodiment;

[0040] Figure 2 This is a flowchart illustrating, according to an exemplary embodiment, the acquisition of a feasible radial distribution network topology;

[0041] Figure 3 This is a flowchart illustrating, according to an exemplary embodiment, the calculation of node voltages and expected total network losses using a semi-invariant method for stochastic power flow.

[0042] Figure 4 This is a schematic diagram of the structure of a deep neural network model according to an exemplary embodiment. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be described in detail below. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0044] Please see Figure 1 , Figure 1 This is a flowchart illustrating a distribution network fault reconfiguration method according to an exemplary embodiment. The distribution network fault reconfiguration method is applied in the field of power grid fault reconfiguration technology and includes:

[0045] S1 generates initial particles based on the state of the power grid line switch;

[0046] S2, The initial particles are divided using a hierarchical clustering method to obtain particles that meet the preset conditions;

[0047] S3, record the individual optimal and global optimal of the particles and update the position of the particles. Based on the updated position of the particles, the particles after position update are divided again by the hierarchical clustering method. Repeat the above iterative process until the preset number of iterations is met.

[0048] S4. Obtain the fitness value of the particle that satisfies the preset number of iterations according to the preset fitness value calculation formula. Determine whether the above calculation process satisfies the constraint conditions. If it does, output the fitness value of the particle and the particle position at this time, and execute the grid switch operation. If it does not satisfy the constraint conditions, iterate again until the constraint conditions are satisfied, output the fitness value of the particle and the particle position at this time, and execute the grid switch operation.

[0049] Specifically, this application generates initial particles based on the state of the power grid line switches. It sets the total number of particles, the number of subgroups, and the number of iterations. The particles are divided using hierarchical clustering. The individual optimality and global optimality of each particle are recorded, and the particle positions are updated. This process is iterated repeatedly until the preset number of iterations is reached. The fitness value of each particle is calculated, and it is determined whether the above calculation process satisfies the constraints. If both are satisfied, the fitness value and position of the particle are output, and a switching operation is executed. In this process, the influence of various random factors in the power grid system is considered through semi-invariant stochastic power flow calculation. Based on hierarchical clustering iteration, not only are particles that are currently only slightly below the optimal but may represent information for other exploration directions considered, but the particle swarm is also divided into several subgroups using hierarchical clustering. In addition to learning the individual optimality and global optimality, each particle also learns the optimal particle in its clustering domain to obtain the optimal reconfiguration strategy when a distribution network containing distributed wind power experiences a fault in the current time period. Parallel switching operations are then performed to transfer load and restore power supply.

[0050] In this embodiment, obtaining the expected total network loss includes:

[0051] Acquire power sample data, which includes the active load of the power grid, reactive load, wind turbine output, and feasible distribution network topology that meets the radial pattern.

[0052] The power sample data is input into a deep neural network model, which calculates the node voltage and the expected value of total network loss based on the sample data using a semi-invariant method for stochastic power flow calculation.

[0053] Specifically, this application obtains sample data and inputs it into a pre-built deep neural network model. The deep neural network model outputs power flow information such as node voltage and expected total network loss through a semi-invariant stochastic power flow calculation method. Based on the power flow information, the constraints and fitness value calculation formula of the scheme in this application are obtained.

[0054] In this embodiment, as shown in the appendix Figure 2 As shown, obtaining the feasible radial distribution network topology includes:

[0055] S101, Read the network structure of the faulty system that has been located and isolated;

[0056] S102, disconnect one switch in each loop of the network structure, that is, the number of disconnected switches is equal to the number of loops;

[0057] S103, close the branch in the network structure that is directly connected to the power supply;

[0058] S104, close all branches in the network structure that do not belong to any loop;

[0059] S105, based on the characteristics of graph theory and tree structure, the connectivity of the network structure is determined, and a series of feasible distribution network topologies that satisfy the radial shape are generated;

[0060] Specifically, the network structure of the faulty system that has been located and isolated is read. To meet the radial constraints of the distribution network, each loop can only disconnect one switch, that is, the number of disconnected switches equals the number of loops. At the same time, branches directly connected to the power source must be closed, and branches that do not belong to any loop must also be closed. If such branches are disconnected, islanding will occur, so such branches are not considered. Based on the characteristics of graph theory and tree structure, the connectivity of the distribution network topology is judged, and a series of feasible distribution network topologies that satisfy the radial constraints are generated.

[0061] In this embodiment, as shown in the appendix Figure 3 As shown, the deep neural network model obtains the node voltage and the expected value of total network loss based on the sample data using a semi-invariant method for stochastic power flow calculation, including:

[0062] S201, calculate the power flow distribution under normal operating conditions of the power grid, and obtain the node voltage state vector of the reference operating point, the power vector of the branch under the reference operating point, the Jacobian matrix and the sensitivity matrix.

[0063] S202, calculate the semi-invariants of the node state vector disturbance part ΔX and the semi-invariants of the branch power flow vector disturbance part ΔZ based on the node voltage state vector of the reference operating point, the power vector of the branch under the reference operating point, the Jacobian matrix and the sensitivity matrix.

[0064] S203, the probability distribution functions and probability density functions of each order of semi-invariants of ΔX and ΔZ are obtained by Gram-Charlier series expansion;

[0065] S204, shift ΔX and ΔZ by X0 and Z0 units to obtain the probability distribution function and probability density function of node voltage and branch power;

[0066] S205, the expected value of each output state quantity, i.e. the expected value of total network loss, is obtained according to the probability distribution function and probability density function of the branch power;

[0067] Specifically, first, the power flow distribution under normal operating conditions is calculated to obtain the node voltage state vector X0, the power vector Z0 of the branch under the reference operating point, the Jacobian matrix J0, and the sensitivity matrix S0. Then, the AC power flow equations in polar coordinates are expanded using Taylor at the reference operating point to obtain:

[0068]

[0069] In the formula, W, X, and Z are the node injected power vector, node state vector, and branch power flow vector, respectively; the subscript 0 indicates the reference operating point; ΔW, ΔX, and ΔZ are the perturbation components of the variables; S0 and T0 are the sensitivity matrices. Where J0 is the Jacobian matrix.

[0070] Calculate the semi-invariants of the power at the wind turbine installation nodes and the corresponding moments of the load power, and calculate the semi-invariants Δr for each order. k Linearizing the nodal power flow equations and branch power flow equations at the reference operating point yields the following equations:

[0071]

[0072] In the formula, X is the state column vector composed of node voltage magnitude and phase angle, and Z is the column vector of active power flow and reactive power flow of the branch.

[0073] Based on the above formula, the semi-invariants of each order of ΔX and ΔZ can be calculated. At the same time, the probability distribution function and probability density function of ΔX and ΔZ can be obtained by using Gram-Charlier series expansion. Finally, by shifting ΔX and ΔZ by X0 and Z0 units respectively, the probability distribution function and probability density function of node voltage X and branch power Z can be obtained.

[0074] The above calculation process yields the node voltage and the probabilistic characteristics of the line power flow at a certain moment, which also allows us to obtain the expected values ​​of each output state variable, i.e., the expected value of the total network loss mentioned above is E(P). loss ).

[0075] In this embodiment, the constraints include: distributed wind power output constraints, node voltage constraints, network topology constraints, power flow balance constraints, and switching action count constraints.

[0076] Specifically, the expression for the distributed wind power output constraint is as follows:

[0077] 0≤P WTG (t)≤P WTG.max (t)

[0078] In the formula, P WTG (t) represents the output of the distributed wind turbine at time t; P WTG.max (t) represents the maximum output of the distributed wind turbine at time t. Since the wind turbine is affected by the natural environment and has a high degree of randomness, its maximum output also changes with time.

[0079] The expression for the node voltage constraint is:

[0080] U imin ≤U i (t)≤U imax

[0081] In the formula, U i (t) represents the voltage amplitude at the i-th node during time period t; U imin U is the lower bound of node i; imax Let i be the upper bound of node i;

[0082] The expression for the network topology constraint is:

[0083] g(t)∈G

[0084] In the formula, g(t) represents the distribution network topology during time period t; G is the set of all radial structures of the distribution network.

[0085] The expression for the power flow balance constraint is:

[0086] f(P,Q,U)=0

[0087] In the formula, P, Q, and U represent the active power, reactive power, and voltage of the system, respectively.

[0088] To extend switch lifespan and reduce unnecessary wear, a maximum number of switch operations is set. Once the maximum number of operations is reached, further operation is prohibited. The expression for this switch operation count constraint is:

[0089]

[0090] In the formula, K i.t and K i0.t Let W represent the open / closed state of the i-th switch before and after reconstruction during time period t, with 0 for open and 1 for closed; max W represents the maximum number of switching actions allowed by the system. i and W imax These are the number of actions for the i-th switch and the maximum number of actions, respectively.

[0091] In this embodiment, the fitness value calculation formula is determined by the power outage load index, the network loss index, and the number of switching operations index, wherein the network loss index is determined by the expected total network loss value.

[0092] Specifically, in this application, the objective functions include minimizing the power outage load, minimizing the expected total network loss, and minimizing the number of switching operations;

[0093] Minimum number of power outage loads:

[0094] The aforementioned fault reconstruction, under the premise that the power distribution system fault has been located and isolated, optimizes the restoration of loads in non-faulty power loss areas. Since loads in the power distribution system are typically classified into three categories—primary, secondary, and tertiary—the fault restoration process aims to maximize the total load restoration while prioritizing the restoration of power to higher-level loads. Therefore, the objective function is as follows:

[0095]

[0096] In the formula, n, m, and l represent the number of first-level loads, second-level loads, and third-level loads, respectively; P 1i (t), P 2j (t), P 3k (t) represents the primary load, secondary load, and tertiary load for time period t, respectively; x i (t) represents the power supply status of the i-th load in the primary load stage during time period t, when x i When (t) = 1, it indicates power loss, x i When (t) = 0, it indicates that no power loss has occurred. j (t) and x k (t) Similarly; C i C j and C k These are the outage weighting coefficients for the three types of loads, which can be set according to the importance of restoring power supply;

[0097] Minimize the expected total network loss:

[0098]

[0099] In the formula, b is the number of branches, and P i.t Q i.t U i.t and R i.t These represent the active power, reactive power, voltage, and resistance of the first segment of the i-th branch during time period t.

[0100] Minimum number of switch operations:

[0101] In actual power systems, since switching operations are costly and can affect the lifespan of switches, the number of switching operations is taken into consideration during system reconfiguration in order to extend the service life of switches and reduce unnecessary losses. Therefore, it is desirable to minimize the number of switch states that need to be changed during load transfer after a fault.

[0102]

[0103] In the formula, n represents the number of sectionalizing switches in the distribution network; m represents the number of tie switches in the distribution network; r i.t and h j.tThese represent the states of the sectionalizing switch and the tie switch during the reconstruction period t, respectively, with 1 indicating closed and 0 indicating open.

[0104] In this embodiment, the hierarchical clustering method is used to divide the initial particles to obtain particles that meet the preset conditions, which includes: taking all the initial particles as a dataset of samples, taking each sample in the dataset as a cluster, calculating the distance between two clusters, finding the two clusters with the closest distance, and then merging the two clusters into one. This process is repeated until the preset number of clusters or threshold is reached.

[0105] Specifically, traditional particle swarm optimization (PSO) algorithms only utilize the individual optimal value (pbest) and the global optimal value (gbest). This report goes beyond this, considering particles that are currently just below the optimal value but may represent information for other exploration directions. Hierarchical clustering is used to divide the particle swarm into several subgroups. Besides learning the individual and global optima, each particle also learns the optimal particle within its clustering neighborhood. Hierarchical clustering divides the original dataset at different levels until a certain condition is met, ultimately resulting in a tree-like clustering structure. Assume there is a dataset s = {s1, s2, ..., sn} containing n samples. n First, each sample is treated as a separate cluster. Then, the distance between any two clusters is calculated, and the two closest clusters are found. These two clusters are then merged into one. This process is repeated until a predetermined number of clusters or a threshold is reached. The calculation formula is shown below:

[0106]

[0107] In the formula, d(s) i ,s j ) represents cluster s i and s j The average distance between them; |xx′| is the distance between the two particles; m i s i Number of particles; m j s j Number of particles in the middle;

[0108] The quality of particles is evaluated using a fitness function. This report uses the minimum power outage load, minimum network loss, and minimum number of switching operations as objective functions, the specific objective functions of which have been introduced previously. Therefore, the distribution network fault reconfiguration at this point is a problem of finding a minimum value. In the particle swarm optimization algorithm, the fitness function used to evaluate the quality of particles can be chosen from the objective function in network reconfiguration; that is, the smaller the fitness function, the better the corresponding particle is, and the closer it is to the objective function. This report comprehensively considers multiple factors, assigns weights to each objective function, and combines the fitness functions into a multi-objective function. Therefore, the fitness function of a particle can be expressed as:

[0109] F = γ1f1 + γ2f2 + γ3f3

[0110] In the formula, γ is the weighting factor of the function; f1, f2 and f3 are the power outage load index, network loss index and switching operation number index, respectively.

[0111] The improved particle swarm optimization algorithm proposed in this application first divides the particle swarm into several subpopulations through hierarchical clustering in each iteration. Then, the best particle in each cluster is marked as lbest. In this way, each particle not only needs to learn from its own experience and the best particle in the population, but also learns from the best particle in the cluster's neighborhood. The improved formula is shown below:

[0112]

[0113] In the formula, ω is the inertia weight; Represents the particle's position; Represents particle velocity; q1, q2, q3 are random factors, i.e., random numbers in [0,1]; α1, α2, α3 are acceleration factors; This represents the individual optimal value of particle i at the k-th iteration; This represents the global optimal value of particle i at the k-th iteration; This represents the position of the best particle in the cluster to which particle i belongs at the kth iteration.

[0114] In particle swarm optimization (PSO), the value of the inertia weight ω directly affects the algorithm's ability to find the optimal value. A linear weight reduction strategy exhibits good optimization performance. Therefore, the formula for updating the ω value is as follows:

[0115]

[0116] In the formula, ω max ω represents the maximum value of the inertia weight. min The minimum value of the inertia weight; k is the current iteration number; k max This represents the maximum number of iterations set.

[0117]

[0118] In the formula, Let be the position of the i-th particle in the (k+1)-th iteration; Let be the velocity of the i-th particle in the (k+1)th iteration. Let be the position of the i-th particle in the k-th iteration.

[0119] As attached Figure 4 As shown, this application also provides a deep neural network model for distribution network fault reconstruction, the deep neural network model including an input layer, a hidden layer and an output layer;

[0120] The input layer is used to input the active and reactive loads of the power grid, the output of wind turbine generators, and a feasible radial distribution network topology.

[0121] The hidden layer is used to calculate the information input from the input layer according to the semi-invariant stochastic power flow calculation method;

[0122] The output layer is used to output the power flow information calculated by the hidden layer;

[0123] Specifically, Deep Neural Networks (DNNs) are a type of neural network that contains multiple hidden layers. DNNs can be divided into three categories according to the position of different layers: input layer, hidden layer, and output layer. They can extract the latent attributes of data by repeatedly abstracting and fitting training data.

[0124] Furthermore, the deep neural network model uses Sigmoid as the activation function;

[0125] Specifically, to improve the nonlinear fitting ability of deep neural networks, this report uses Sigmoid as the activation function, whose expression is:

[0126]

[0127] Furthermore, during the training process of the deep neural network model, a loss function is used to measure the loss between the output calculated from the training samples and the actual output of the training samples.

[0128] Specifically,

[0129] Forward propagation:

[0130] The so-called DNN forward propagation algorithm uses several weight coefficient matrices W and bias vector b to perform a series of linear operations and activation operations with the input value vector x. The sample data is input from the input layer and calculated layer by layer until the operation reaches the output layer K to obtain the output result of the Kth layer.

[0131] a k =σ(z)k )=σ(W k a k-1 +b k )

[0132] In the formula, a k σ is the output of the Kth layer of the DNN; σ is the activation function; W k b represents the weight of the Kth layer; k The bias of the Kth layer; z k This is the output before the Kth layer is activated;

[0133] Backpropagation:

[0134] DNNs solve for the optimal neural network parameters through backpropagation. This requires selecting a loss function to measure the difference between the output calculated from the training samples and the actual output from the training samples. The error expression is as follows:

[0135]

[0136] In the formula, x is the sample input; y is the sample true value; and J is the error.

[0137] After the deep neural network is constructed, it is trained using sample data. The active load, reactive load, wind turbine output, and feasible radial distribution network topology in the training set are used as inputs to the deep neural network, and power flow information such as line loss and node voltage are used as outputs to train the deep neural network.

[0138] Distribution network fault reconfiguration device, the device comprising:

[0139] Initial particle generation module: used to generate initial particles from the state of the power grid line switches;

[0140] Particle partitioning module: used to partition the initial particles using a hierarchical clustering method to obtain particles that meet preset conditions;

[0141] Particle Iteration Module: Records the individual optimal and global optimal values ​​of particles and updates the positions of the particles. Based on the updated positions of the particles, the hierarchical clustering method is used to further divide the updated particles. The above iterative process is repeated until the preset number of iterations is met.

[0142] Result output module: Obtain the fitness value of the particle that meets the preset number of iterations according to the preset fitness value calculation formula, determine whether the above calculation process meets the constraint conditions, if it meets the constraint conditions, output the fitness value of the particle and the particle position at this time, and execute the grid switch operation, if it does not meet the constraint conditions, it iterate again until the constraint conditions are met, output the fitness value of the particle and the particle position at this time, and execute the grid switch operation.

[0143] A storage medium storing a computer program, which, when executed by a processor, implements the various steps in the power distribution network fault reconfiguration method.

[0144] Specifically, the storage medium may be a read-only memory, a disk or an optical disk, or any combination thereof.

[0145] It is understood that the same or similar parts in the above embodiments can be referred to each other, and the contents not described in detail in some embodiments can be referred to the same or similar contents in other embodiments.

[0146] It should be noted that in the description of this application, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, in the description of this application, unless otherwise stated, "multiple" or "more" means at least two.

[0147] It should be understood that when an element is referred to as "fixed to" or "set on" another element, it may be directly on the other element or may have an intervening element present at the same time; when an element is referred to as "connected to" another element, it may be directly connected to the other element or may have an intervening element present at the same time. In addition, the term "connected" as used herein may include wireless connections; the word "and / or" as used includes any unit and all combinations of one or more of the associated listed items.

[0148] Any process or method description in the flowchart or otherwise herein can be understood as: representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the function involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0149] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0150] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0151] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0152] The storage media mentioned above can be read-only memory, disk, or optical disk, etc.

[0153] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0154] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for reconfiguring distribution network faults, characterized in that, The reconstruction method includes: Generate initial particles from the state of the power grid line switches; The initial particles are divided using a hierarchical clustering method to obtain particles that meet preset conditions; The initial particles are divided using a hierarchical clustering method to obtain particles that meet preset conditions, including: The initial particles are treated as a dataset of samples. Each sample in the dataset is treated as a separate cluster. The distance between two clusters is calculated, and the two closest clusters are found. These two clusters are then merged into one. This process is repeated until a pre-set number of clusters or a threshold is reached. Record the individual optimal and global optimal values ​​of the particles and update the positions of the particles. Based on the updated positions of the particles, the particles with updated positions are divided again using the hierarchical clustering method. Repeat the above iterative process until the preset number of iterations is met. The fitness value of a particle that meets the preset number of iterations is obtained according to the preset fitness value calculation formula. It is determined whether the constraint conditions are met in the above calculation process. If they are met, the fitness value of the particle and the current position of the particle are output, and the power grid switch operation is executed. If they are not met, the iteration is performed again until the constraint conditions are met, the fitness value of the particle and the current position of the particle are output, and the power grid switch operation is executed. The fitness value calculation formula is determined by the power outage load index, the network loss index, and the number of switching operations index, wherein the network loss index is determined by the expected total network loss value.

2. The distribution network fault reconfiguration method according to claim 1, characterized in that, The acquisition of the expected total network loss includes: Acquire power sample data, which includes the active load of the power grid, reactive load, wind turbine output, and feasible distribution network topology that meets the radial pattern. The power sample data is input into a deep neural network model, which calculates the node voltage and the expected total network loss based on the sample data using a semi-invariant method for stochastic power flow calculation.

3. The distribution network fault reconfiguration method according to claim 2, characterized in that, The acquisition of a feasible radial distribution network topology includes: Read the network structure of the faulty system that has been located and isolated; In the network structure, disconnect one switch in each loop, that is, the number of disconnected switches is equal to the number of loops; Close the branch in the network structure that is directly connected to the power supply; Close all branches in the network structure that do not belong to any loop; Based on the characteristics of graph theory and tree structure, the connectivity of the network structure is determined, resulting in a series of feasible distribution network topologies that satisfy the radial shape.

4. The distribution network fault reconfiguration method according to claim 3, characterized in that, The deep neural network model obtains the node voltage and the expected value of total network loss based on the sample data using a semi-invariant method for stochastic power flow calculation, including: Calculate the power flow distribution under normal operating conditions of the power grid to obtain the node voltage state vector, the power vector of the branch under the reference operating point, the Jacobian matrix, and the sensitivity matrix. The node state vector perturbation component is calculated based on the node voltage state vector at the reference operating point, the power vector of the branch at that reference operating point, the Jacobian matrix, and the sensitivity matrix. The semi-invariants of each order and the branch power flow vector perturbation part The semi-invariants of each order; We obtain the following through Gram-Charlier series expansion: as well as The probability distribution function and probability density function of each order of semi-invariant; Will and Translation and By analyzing the units, we can obtain the probability distribution function and probability density function of the node voltage and branch power; The expected value of each output state quantity, i.e. the expected value of the total network loss, is obtained based on the probability distribution function and probability density function of the branch power.

5. The distribution network fault reconfiguration method according to claim 4, characterized in that, The constraints include: distributed wind power output constraints, node voltage constraints, network topology constraints, power flow balance constraints, and switching operation count constraints.

6. The distribution network fault reconfiguration method according to claim 5, characterized in that, The node voltage constraint is generated based on the node voltage.

7. A distribution network fault reconfiguration device, characterized in that, The device includes: Initial particle generation module: used to generate initial particles from the state of the power grid line switches; Particle partitioning module: used to partition the initial particles using a hierarchical clustering method to obtain particles that meet preset conditions; The initial particles are divided using a hierarchical clustering method to obtain particles that meet preset conditions, including: The initial particles are treated as a dataset of samples. Each sample in the dataset is treated as a separate cluster. The distance between two clusters is calculated, and the two closest clusters are found. These two clusters are then merged into one. This process is repeated until a pre-set number of clusters or a threshold is reached. Particle Iteration Module: Records the individual optimal and global optimal values ​​of particles and updates the positions of the particles. Based on the updated positions of the particles, the hierarchical clustering method is used to further divide the updated particles. The above iterative process is repeated until the preset number of iterations is met. Result output module: Obtain the fitness value of the particle that meets the preset number of iterations according to the preset fitness value calculation formula, determine whether the constraint conditions are met in the above calculation process, if they are met, output the fitness value of the particle and the particle position at this time, and execute the grid switch operation; if they are not met, iterate again until the constraint conditions are met, output the fitness value of the particle and the particle position at this time, and execute the grid switch operation. The fitness value calculation formula is determined by the power outage load index, the network loss index, and the number of switching operations index, wherein the network loss index is determined by the expected total network loss value.

8. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements each step of the distribution network fault reconfiguration method as described in any one of claims 1-6.

Images