Power distribution network partitioning method, device and equipment and storage medium

Abstract

The invention provides a power distribution network partitioning method and device, equipment and a storage medium. The method comprises the steps that firstly, a node voltage-reactive power sensitivity matrix and historical operation data of a distributed power supply are obtained, the structural distance between nodes is calculated, the historical data are normalized, the attribute similarity is obtained in combination with an attribute weight matrix, and the two are added to generate an electrical distance. Resolution parameters are dynamically adjusted according to reverse power flow intensity, and the parameters are increased when the power flow is enhanced to refine clusters. According to the method, the limitation that traditional partition depends on geographical proximity or simple parameters is broken through, accurate partition is achieved by quantifying electrical coupling and resource characteristics, the problem of poor connection in the cluster is solved, the on-site consumption capacity of new energy is effectively improved, the transmission pressure of a main network is reduced, and the method is suitable for large-scale popularization and application. And meanwhile, the fault recovery capability and the calculation efficiency of the power grid are enhanced, and the method is suitable for district power balance optimization in a high-permeability distributed power supply access scene.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of power grid partitioning, and particularly relates to a power distribution network partitioning method, device, equipment and storage medium. BACKGROUND

[0002] With large-scale access of distributed new energy and energy storage devices in the power distribution network, the penetration rate in some areas breaks through 100%, leading to significant increase in source and load uncertainty on both sides of the area power system.

[0003] The prior art excessively relies on node geographical proximity or simple electrical parameters, and does not fully consider the time-space coupling effect of distributed power output fluctuation and energy storage charging and discharging characteristics, resulting in imbalance of area classification resources; the clustering algorithm with fixed resolution is difficult to adapt to the reverse power flow overload scene, when the power fluctuation of the power distribution feeder increases, the probability and amplitude of the reverse power flow increase, the demand for power exchange between clusters increases sharply, but the traditional method cannot dynamically adjust the cluster granularity, causing the internal connection density to decrease and the self-balancing ability to weaken; the partitioning scheme based on Louvain algorithm and the like has the problem of poor internal connection of clusters, and does not combine network topology and sensitivity index for modular evaluation, leading to deviation in node correlation judgment and affecting the local consumption efficiency of new energy. SUMMARY

[0004] The purpose of the present application is to overcome the defects in the prior art, and to provide a power distribution network partitioning method, device, equipment and storage medium.

[0005] The present application provides a power distribution network partitioning method, comprising:

[0006] obtaining a voltage-reactive sensitivity matrix of nodes in the power distribution network and historical operation data of distributed power resources;

[0007] calculating a structural distance between nodes according to the voltage-reactive sensitivity matrix, the structural distance being the total branch reactance through which the power flow flows between nodes;

[0008] normalizing the historical operation data, and calculating an attribute similarity between nodes according to the normalized historical operation data and a preset attribute weight matrix;

[0009] adding the structural distance and the attribute similarity to generate an electrical distance between nodes;

[0010] determining a resolution parameter according to the reverse power flow strength of the power distribution network, and increasing the resolution parameter when the reverse power flow strength increases;

[0011] based on the electrical distance and the resolution parameter, using a Leiden algorithm to divide the nodes of the power distribution network into clusters to obtain a partitioning result.

[0012] Optionally, according to the normalized historical operation data and the preset attribute weight matrix, an attribute similarity between nodes is calculated, including:

[0013] The attribute weight matrix in the form of a diagonal matrix is constructed;

[0014] Based on the diagonal matrix and the node attribute vector difference, a Minkowski distance is calculated;

[0015] The attribute similarity is generated according to the Minkowski distance.

[0016] Optionally, a structural distance between nodes is calculated, including:

[0017] The reactance component of the distribution network branch is extracted;

[0018] The structural distance is generated by summing the reactance components on the power flow path between nodes.

[0019] Optionally, the structural distance and the attribute similarity are added, including:

[0020] The structural distance and the attribute similarity are added without weighting;

[0021] The addition result is output as the electrical distance.

[0022] Optionally, a resolution parameter is determined according to the reverse power flow strength of the distribution network, including:

[0023] The reverse power flow amplitude of the distribution network is detected;

[0024] When the reverse power flow amplitude increases, the resolution parameter is set to a value greater than 1;

[0025] When the reverse power flow amplitude decreases, the resolution parameter is set to a value less than or equal to 1.

[0026] Optionally, the Leiden algorithm is used to cluster and divide the nodes of the distribution network, including:

[0027] The node queue is initialized and the modularity gain is calculated;

[0028] The nodes are assigned to the cluster with the maximum modularity gain;

[0029] When the maximum modularity gain is less than 0, the nodes are kept in the original cluster;

[0030] The clusters with positive modularity gain are merged to generate the partition result in the form of a connected subgraph.

[0031] Optionally, the historical operation data includes: the output curve of the photovoltaic component, the charge and discharge power curve of the energy storage device, and the capacity data of the energy storage device.

[0032] The application also provides a power distribution network partition device, comprising:

[0033] An acquisition module acquires a voltage-reactive sensitivity matrix of nodes in a power distribution network and historical operation data of distributed power resources;

[0034] A structure module calculates a structural distance between nodes according to the voltage-reactive sensitivity matrix, wherein the structural distance is a total branch reactance through which a power flow between the nodes flows;

[0035] A normalization module normalizes the historical operation data, and calculates an attribute similarity between the nodes according to the normalized historical operation data and a preset attribute weight matrix;

[0036] An electrical module adds the structural distance and the attribute similarity to generate an electrical distance between the nodes;

[0037] A parameter module determines a resolution parameter according to a reverse power flow strength of the power distribution network, wherein the resolution parameter is increased when the reverse power flow strength is increased;

[0038] A partition module performs cluster division on the nodes of the power distribution network based on the electrical distance and the resolution parameter by using a Leiden algorithm to obtain a partition result.

[0039] The application also provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the above method when executing the computer program.

[0040] The application also provides a computer readable storage medium, which stores a computer program, and the computer program makes a computer execute the above method when the computer program is executed in the computer.

[0041] The application has the following beneficial effects:

[0042] The application provides a power distribution network partitioning method, comprising: obtaining a voltage-reactive power sensitivity matrix of nodes in a power distribution network and historical operation data of distributed power resources; calculating a structural distance between nodes according to the voltage-reactive power sensitivity matrix, wherein the structural distance is the total branch reactance through which a power flow between nodes flows; performing normalization processing on the historical operation data, and calculating an attribute similarity between nodes according to the normalized historical operation data and a preset attribute weight matrix; adding the structural distance and the attribute similarity to generate an electrical distance between nodes; determining a resolution parameter according to a reverse power flow strength of the power distribution network, wherein the resolution parameter is increased when the reverse power flow strength is increased; and performing cluster division on nodes of the power distribution network based on the electrical distance and the resolution parameter by using a Leiden algorithm to obtain a partitioning result. The application dynamically quantifies the electrical distance and the attribute similarity, adaptively adjusts the resolution parameter in combination with the reverse power flow strength, and realizes accurate cluster division by using the Leiden algorithm, thereby effectively improving a new energy local consumption rate, reducing a main network pressure, and enhancing power grid resilience and calculation efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 is a power distribution network partitioning flowchart in the application;

[0044] Figure 2 is a power aggregation schematic diagram in the application;

[0045] Figure 3 is a Leiden algorithm cluster division schematic diagram in the application;

[0046] Figure 4 is a schematic diagram of photovoltaic output of an aggregated subject in the application;

[0047] Figure 5 is a resulting pricing schematic diagram in the application. DETAILED DESCRIPTION

[0048] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it is understood that various forms implement the present disclosure and should not be limited by the embodiments set forth herein. On the contrary, the embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be accurately conveyed to those skilled in the art.

[0049] Referring to Figure 1 The application provides a power distribution network partitioning method, applied to the field of power systems, for solving the problems of slice area reverse power flow overload, new energy consumption difficulty and power balance caused by high penetration of distributed power sources. The method comprises:

[0050] S101, acquire voltage-reactive sensitivity matrix of nodes in power distribution network and historical operation data of distributed power resource.

[0051] The historical operation data includes output curve of photovoltaic component, charge-discharge power curve of energy storage device and capacity data of energy storage device.

[0052] The data is derived from historical monitoring record of power distribution network, the photovoltaic output curve reflects the power generation power change of photovoltaic component at different time periods, the energy storage charge-discharge power curve records the power fluctuation of energy storage device charge-discharge behavior, and the energy storage capacity data represents the maximum storable energy of energy storage device.

[0053] By collecting the historical data, the operation characteristics of distributed power supply are comprehensively described, and the basis for subsequent clustering analysis is provided. For example, the photovoltaic output curve can be extracted from the monitoring system of photovoltaic inverter, and the energy storage data comes from the log record of battery management system (BMS).

[0054] S102, calculate the structural distance between nodes according to the voltage-reactive sensitivity matrix, the structural distance is the total branch reactance of the power flow between nodes.

[0055] Extract the reactance component of the branch of the power distribution network, sum the reactance component on the power flow path between nodes, and generate the structural distance. The structural distance calculation formula is:

[0056]

[0057] Wherein, the structural distance between nodes i and j is The branch reactance between nodes i and j is The power flow path set between nodes i and j is And h is the branch index in the path.

[0058] This calculation is based on the physical relationship of voltage-reactive sensitivity matrix, and reflects the degree of electrical coupling between nodes.

[0059] For example, for nodes i and j, if the power flow path includes three branches, the reactance components are , and , then the structural distance is the sum of the three. This ensures that the physical constraints of the power grid (such as line impedance) are included in the partition process, avoiding the problem of electrical isolation after partition.

[0060] In the structural distance calculation, the component of the Jacobian matrix J is:

[0061] , , ,

[0062] Characterizes the relationship between the change in node power and the node voltage vector.

[0063] The sensitivity relationship is given by the Jacobian matrix:

[0064] Establish a node Voltage amplitude / phase angle and node The relationship between active and reactive power:

[0065]

[0066] in , The change in active / reactive power injected into the node. , This represents the change in node voltage phase angle and amplitude. Electrical distance calculation comprehensively considers factors such as network topology, sensitivity, resource operation characteristics, and energy storage capacity.

[0067] The sensitivity matrix of node voltage to node active / reactive power can be expressed as:

[0068]

[0069] in, For nodes The voltage amplitude; , For nodes Active / reactive power; The rated voltage of the area;

[0070] For nodes and The sum of the resistances of the branches through which the current flows.

[0071] For nodes and The sum of the reactance of the branches through which the current flows.

[0072] S103. Normalize the historical operation data and calculate the attribute similarity between nodes based on the normalized historical operation data and the preset attribute weight matrix.

[0073] Construct the attribute weight matrix in diagonal form; calculate the Minkowski distance based on the difference between the diagonal matrix and the node attribute vectors; generate the attribute similarity based on the Minkowski distance. The attribute similarity calculation formula is as follows:

[0074]

[0075] in, The attribute similarity between node i and node j; and Let i and j be attribute vectors, respectively, in the form of: The Indicates the first The characteristics of distributed power sources (such as photovoltaic output and energy storage power); D is the attribute weight matrix, which is in diagonal matrix form:

[0076]

[0077] Its diagonal value Each characteristic reflects its relative importance; The difference is in the node attribute vector; The distance is the Minkowski distance.

[0078] Normalization scales historical operational data (such as photovoltaic output curves) to the range [0, 1], eliminating the influence of dimensions. For example, if the normalized value of the photovoltaic output characteristic of node i is 0.8, and that of node j is 0.6, and the photovoltaic characteristic weight α in the weight matrix D is 0.5, then the Minkowski distance is used to generate similarity. This ensures that nodes with similar resource characteristics are clustered preferentially, maintaining resource balance in the area classification.

[0079] S104. Add the structural distance to the attribute similarity to generate the electrical distance between nodes.

[0080] Perform an unweighted arithmetic summation on the structural distance and the attribute similarity; output the summation result as the electrical distance. The electrical distance calculation formula is:

[0081]

[0082] in, The electrical distance between node i and node j; For structural distance; This represents attribute similarity.

[0083] This summation method directly superimposes structural distance and attribute similarity without introducing additional weighting factors, ensuring that the physical constraints of the power grid and the functional characteristics of resources are considered with equal weight. For example, if the structural distance is 5.0 and the attribute similarity is 0.8, then the electrical distance is 5.8. This integrates network topology, sensitivity, and resource operating characteristics, producing a synergistic effect and improving the local absorption capacity after partitioning.

[0084] S105. Determine the resolution parameter based on the reverse power flow intensity of the distribution network, and increase the resolution parameter when the reverse power flow intensity increases.

[0085] Detecting a reverse power flow amplitude of the power distribution network; setting the resolution parameter to a value greater than 1 when the reverse power flow amplitude increases; setting the resolution parameter to a value less than or equal to 1 when the reverse power flow amplitude decreases.

[0086] The resolution parameter γ is used for subsequent modularity calculation. When the reverse power flow intensity increases (manifested as an increase in power fluctuation and reverse power flow probability), the dependence of the power distribution network line increases, the internal connection density of the cluster decreases, and the number of clusters needs to be increased, so γ>1 is set; when the reverse power flow intensity decreases, γ≤1 is set.

[0087] The reverse power flow amplitude can be obtained through the power distribution network monitoring system. For example, if the reverse power flow amplitude increases by 20% compared with the reference value in a certain period, then γ=1.2 is set. This dynamically adjusts the partition granularity to cope with different balance requirements.

[0088] S106, based on the electrical distance and the resolution parameter, using Leiden algorithm to cluster the nodes of the power distribution network, and obtaining a partition result.

[0089] Initialize the node queue and traverse the modularity gain calculation module; assign the nodes to the cluster with the maximum modularity gain; when the maximum modularity gain is less than 0, keep the node in the original cluster; in the optimization stage, merge the clusters with positive modularity gain to generate the partition result in the form of a connected subgraph.

[0090] The modularity index calculation formula is:

[0091]

[0092] Wherein, ρ is the extended modularity index; m is the sum of all edge weights in the power distribution network:

[0093]

[0094] γ is the resolution parameter; the and are the sum of all edge weights connected to nodes i and j respectively:

[0095]

[0096] δ(i, j) is an indicator function, which is 1 when nodes i and j are in the same cluster, and 0 otherwise.

[0097] Please refer to Figure 3 The Leiden algorithm division process is as follows:

[0098] First, initialize the node queue, and add all nodes in random order;

[0099] Traverse the queue, calculate the modularity gain, and assign the node to the cluster with the largest gain;

[0100] If the maximum gain is less than zero, remain in the original cluster;

[0101] In the optimization phase, merge clusters with positive modularity gain randomly, and generate a connected subgraph as the partition result.

[0102] Compared with the Louvain algorithm, the Leiden algorithm solves the problem of poor internal connection and ensures the connectivity of the partition. For example, if the maximum modularity gain after traversing the node queue is 0.3, it is assigned to the corresponding cluster; in the optimization phase, only when the gain is positive, the cluster is merged. This is based on the electrical distance weight and dynamic resolution parameter, which realizes accurate cluster division and reduces the pressure on the main network.

[0103] In the modularity formula, the Only when nodes i and j are in the same cluster; the Requires the sum of the electrical distance of node i and all adjacent nodes.

[0104] Please refer to Figure 2 Further, after completing the partition of the power distribution network, based on the cluster topology formed by aggregation, a joint clearing model of the electricity market and the frequency regulation service market is constructed. The objective function of the model is:

[0105]

[0106] Where, is the unit bidding function, is the aggregator bidding function, the is the winning output of the unit, is the winning amount of the unit frequency regulation capacity, is the winning output of the aggregator, is the winning amount of the aggregator frequency regulation capacity, T is the total period, is the total number of units, is the total number of aggregators.

[0107] The constraint conditions include:

[0108]

[0109]

[0110]

[0111]

[0112]

[0113]

[0114]

[0115]

[0116]

[0117]

[0118]

[0119]

[0120]

[0121]

[0122]

[0123]

[0124]

[0125]

[0126]

[0127] in, This is the upper limit of the unit's output. It is the lower limit of the unit's output. / These are the upper and lower limits of unit power variation. It is the output of photovoltaic power from the aggregated merchants. / It is the power of the energy storage charging / discharging within the aggregater. It refers to the capacity of photovoltaic power plants within the aggregation company to participate in frequency regulation. It is the capacity of energy storage within the aggregator to participate in frequency regulation. This is the upper limit of photovoltaic power output. This is the lower limit of photovoltaic power output. This is the upper limit of photovoltaic frequency regulation capacity. This is the upper limit of energy storage charging and discharging power. / These are the upper and lower limits of energy storage capacity. It refers to energy storage capacity. This is the upper limit of the energy storage capacity participating in frequency regulation. It is the total load within the system. It is a system frequency modulation requirement. It refers to the power on the line. / are upper and lower limits of power flow on the line.

[0128] Please refer to Figure 4 , by scheduling model within the line capacity constraints, can ensure that the line flow will not be over the limit, due to the low cost of renewable energy, renewable energy will be given priority to bid in clearing, so as to realize the power system in the premise of ensuring stable operation to promote new energy consumption. The joint clearing model by coordinating the output of unit and aggregator, solve the contradiction between new energy consumption and line overload, for example, when the line capacity is scarce in the cluster, the constraint condition limits the output to prevent reverse power flow overload.

[0129] Please refer to Figure 5 , further, according to the scheduling output of joint clearing, to establish a double-layer pricing model to obtain the electricity price and frequency modulation price.

[0130] The upper model is:

[0131]

[0132] Among them, the indicates the unit self-scheduling income, the indicates the aggregator self-scheduling income, and the calculation formula is:

[0133]

[0134]

[0135] Among them, the is the system marginal price, and the is the frequency modulation price.

[0136] The lower model is:

[0137]

[0138] Among them, the , is the joint clearing income, and the calculation formula is:

[0139]

[0140]

[0141] The price boundary constraint is and , wherein the , , , respectively indicate the upper and lower limits of electricity energy and frequency modulation price.

[0142] There are a large number of photovoltaic distributions in the large cluster formed after polymerization, and the line capacity is easy to be scarce, which may cause the price to exceed the range allowed by market operation. The price boundary constraint can make the price meet the needs of market operation and ensure the smooth operation of the transaction mode. The price boundary is implicitly related to the cluster modularity. The high modularity cluster (internal connection is close) is more likely to maintain price stability and avoid price out-of-limit due to line overload.

[0143] Further, by integrating the physical constraints of the power grid and the functional characteristics of the resources, a synergistic effect is generated during the partitioning process, solving the serious problems of reverse power flow overload, power supply and demand shortage, and new energy consumption difficulty caused by distributed new energy penetration rate exceeding 100%. The specific technical effects include: improving local new energy consumption capacity and reducing main grid power transmission pressure; reducing communication, calculation and control costs; enhancing the resilience and fault recovery capability of the power grid, and achieving economic and efficient regional power balance. This is achieved through weight graph construction, Leiden algorithm division and market mechanism synergy, such as internal resource coordination to reduce the probability of reverse power flow and local consumption to reduce the dependence on the main grid.

[0144] The application also provides a power distribution network partitioning device, comprising:

[0145] An acquisition module acquires the voltage-reactive sensitivity matrix of nodes in the power distribution network and historical operation data of distributed power resources;

[0146] A structure module calculates the structural distance between nodes according to the voltage-reactive sensitivity matrix, wherein the structural distance is the total branch reactance through which the power flow flows between nodes;

[0147] A normalization module normalizes the historical operation data, and calculates the attribute similarity between nodes according to the normalized historical operation data and a preset attribute weight matrix;

[0148] An electrical module adds the structural distance and the attribute similarity to generate an electrical distance between nodes;

[0149] A parameter module determines a resolution parameter according to the reverse power flow strength of the power distribution network, and increases the resolution parameter when the reverse power flow strength increases;

[0150] A partitioning module performs cluster division on the nodes of the power distribution network based on the electrical distance and the resolution parameter using the Leiden algorithm to obtain a partitioning result.

[0151] The application also provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above method.

[0152] The application further provides a computer readable storage medium, which stores a computer program, and the computer program causes a computer to execute the method when the computer program is executed in the computer.

[0153] The above description of the embodiments is for facilitating those of ordinary skill in the art to understand and apply the present application. Those skilled in the art can easily make various modifications to the above embodiments and apply the general principles described herein to other embodiments without creative efforts. Therefore, the present application is not limited to the above embodiments, and the improvements and modifications made to the present application by those skilled in the art according to the disclosure of the present application should be within the protection scope of the present application.

Claims

1. A method of partitioning an electrical distribution network, characterized in that, The method comprises the following steps: obtaining the voltage-reactive sensitivity matrix of nodes in the power distribution network and the historical operation data of distributed power resources; calculating the structural distance between nodes according to the voltage-reactive sensitivity matrix, wherein the structural distance is the total branch reactance through which the power flow between nodes flows; normalizing the historical operation data, and calculating the attribute similarity between nodes according to the normalized historical operation data and a preset attribute weight matrix; adding the structural distance and the attribute similarity to generate the electrical distance between nodes; determining the resolution parameter according to the reverse power flow strength of the power distribution network, wherein the resolution parameter increases when the reverse power flow strength increases; performing cluster division on the nodes of the power distribution network based on the electrical distance and the resolution parameter by using the Leiden algorithm to obtain the partition result.

2. The method of claim 1, wherein, The attribute similarity between nodes is calculated according to the normalized historical operation data and a preset attribute weight matrix, which comprises the following steps: constructing the attribute weight matrix in the form of a diagonal matrix; calculating the Minkowski distance based on the difference between the diagonal matrix and the node attribute vector; generating the attribute similarity according to the Minkowski distance.

3. The method of claim 1, wherein, The structural distance between nodes is calculated, which comprises the following steps: extracting the reactance component of the branch of the power distribution network; summing the reactance components on the power flow path between nodes to generate the structural distance.

4. The method of claim 1, wherein, The structural distance and the attribute similarity are added, which comprises the following steps: performing unweighted arithmetic addition on the structural distance and the attribute similarity; outputting the addition result as the electrical distance.

5. The method of claim 1, wherein, The resolution parameter is determined according to the reverse power flow strength of the power distribution network, which comprises the following steps: detecting the reverse power flow amplitude of the power distribution network; when the reverse power flow amplitude increases, setting the resolution parameter to a value greater than 1; when the reverse power flow amplitude decreases, setting the resolution parameter to a value less than or equal to 1.

6. The method of claim 1, wherein, The Leiden algorithm is used to perform cluster division on the nodes of the power distribution network, which comprises the following steps: initializing the node queue and traversing the module degree gain calculation module; allocating nodes to the cluster with the maximum module degree gain; when the maximum module degree gain is less than 0, keeping the nodes in the original cluster; merging the clusters with positive module degree gain to generate the partition result in the form of a connected subgraph.

7. The method of claim 1, wherein, The historical operation data comprises the output curve of a photovoltaic module, the charge and discharge power curve of an energy storage device, and the capacity data of the energy storage device.

8. A power distribution network partitioning device, characterized by, The method comprises the following steps: an acquisition module for obtaining the voltage-reactive sensitivity matrix of nodes in the power distribution network and the historical operation data of distributed power resources; a structure module for calculating the structural distance between nodes according to the voltage-reactive sensitivity matrix, wherein the structural distance is the total branch reactance through which the power flow between nodes flows; a normalization module for normalizing the historical operation data, and calculating the attribute similarity between nodes according to the normalized historical operation data and a preset attribute weight matrix; an electrical module for adding the structural distance and the attribute similarity to generate the electrical distance between nodes; a parameter module for determining the resolution parameter according to the reverse power flow strength of the power distribution network, wherein the resolution parameter increases when the reverse power flow strength increases; A partition module adopts a Leiden algorithm to cluster and divide nodes of the power distribution network based on the electrical distance and the resolution parameter, to obtain a partition result.

9. An electronic device, comprising: A computer program product, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method of any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, A computer program product, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method of any one of claims 1-7.

Images