A power distribution network topology dynamic reconstruction method based on space-time feature fusion

Abstract

The application provides a power distribution network topology dynamic reconstruction method based on space-time feature fusion, comprising: acquiring first time series data, second time series data and line operation state; generating first space-time feature parameters based on the first time series data and the line operation state; generating second space-time feature parameters based on the second time series data and the line operation state; generating a local initial topology reconstruction strategy based on the first space-time feature parameters and the second space-time feature parameters; performing neighborhood correction on the initial topology reconstruction strategy to obtain a first topology reconstruction strategy; performing consistency judgment on the first topology reconstruction strategy to generate a global topology reconstruction strategy; and generating corresponding switch operation instructions based on the global topology reconstruction strategy and outputting the instructions to an instruction execution end of the power distribution network. The method provided by the application realizes effective fusion of space-time features of the power supply side and the load side in the urban-rural junction power distribution network, improves the operation flexibility and efficiency of the power distribution network, and reduces the waste of power resources.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of intelligent power grid operation control, and particularly relates to a power distribution network topology dynamic reconstruction method based on space-time feature fusion. BACKGROUND

[0002] The power distribution network topology dynamic reconstruction refers to an operation control technology of changing the on or off state of a switch of a power distribution network, optimizing the topology structure of the power grid, so that the network loss is reduced, the power flow is reasonably distributed, the voltage quality is improved, and the power supply reliability is improved. The method usually depends on load prediction, power flow calculation and optimization of switch operation strategy, and under the premise of meeting the power supply safety, the operation efficiency and economy of the power distribution network are improved, and has become an important means in the operation of the power distribution network.

[0003] The current power distribution network topology dynamic reconstruction technology is mostly based on power flow analysis and optimization algorithm. Its advantage is that in the case that the load characteristics of a single area are relatively consistent, the approximately optimal topology structure can be quickly found through mathematical optimization. However, in the case of urban-rural junction, the operation of the power distribution network presents obvious differentiated characteristics. The rural power grid and the urban power grid are usually physically interconnected at the transformer substation or the junction line. Distributed photovoltaic and small hydropower and other clean energy are widely used in rural areas. The power supply side is characterized by scattered output, strong volatility and redundancy. The load of the urban area is mainly composed of residents, businesses and public infrastructure. The demand side is characterized by concentrated load, stability and large peak-valley difference. This makes the urban-rural junction naturally have a complementary relationship of "rural areas as power supply side and urban areas as load side". If the space-time characteristics of the urban-rural differentiation are not considered, the distributed power supply in rural areas may be abandoned during the output peak, and the urban power grid may still face power supply pressure during the load peak, which ultimately leads to low utilization efficiency of power resources and weakens the economy and reliability of the overall operation of the power distribution network. SUMMARY

[0004] The application provides a power distribution network topology dynamic reconstruction method based on space-time feature fusion. In order to solve the above technical problems, the application adopts the following technical method:

[0005] The application provides a power distribution network topology dynamic reconstruction method based on space-time feature fusion, comprising:

[0006] obtaining first time series data, second time series data and line operation state; the first time series data represents the power generation data of each power supply node on the power supply side at the target junction; the second time series data represents the power demand data of each load node on the load side at the target junction, and the line operation state represents the line state between each power supply node and each load node at the target junction;

[0007] generating a first space-time feature parameter based on the first time series data and the line operation state;

[0008] generating a second spatio-temporal feature parameter based on the second time sequence data and the line operation state;

[0009] generating a local initial topology reconfiguration strategy based on the first spatio-temporal feature parameter and the second spatio-temporal feature parameter;

[0010] performing neighborhood correction on the initial topology reconfiguration strategy to obtain a first topology reconfiguration strategy;

[0011] performing consistency judgment on the first topology reconfiguration strategy to generate a global topology reconfiguration strategy;

[0012] generating corresponding switch operation instructions based on the global topology reconfiguration strategy and outputting the instructions to an instruction execution end of the power distribution network.

[0013] Optionally, the neighborhood correction on the initial topology reconfiguration strategy to obtain a first topology reconfiguration strategy comprises:

[0014] converting the initial topology reconfiguration strategy into an initial suggestion set containing a first number of suggestion items;

[0015] swapping the initial suggestion set of the adjacent power supply node and load node;

[0016] performing feasibility judgment on all initial suggestion sets received by each power supply node and each load node according to the corresponding spatio-temporal feature parameter and the corresponding line operation state to generate feasible suggestion items, infeasible suggestion items and out-of-limit suggestion items, and marking the infeasible suggestion items and the out-of-limit suggestion items as limited suggestion items;

[0017] correcting the limited suggestion items under the constraint of the spatio-temporal feature parameter and the corresponding line operation state of the current power supply node / load node;

[0018] converting the feasible suggestion items and the corrected limited suggestion items to generate a first topology reconfiguration strategy, and associating the first topology reconfiguration strategy to the corresponding power supply node or load node.

[0019] Optionally, the consistency judgment on the first topology reconfiguration strategy to generate a global topology reconfiguration strategy comprises:

[0020] constructing a power supply-load distribution matrix of the target interface based on the first topology reconfiguration strategy;

[0021] calculating a local deviation value for each row and each column of the power supply-load distribution matrix, the local deviation value being used to quantify the difference between the power supply node available power and the corresponding load node absorbable demand under the constraint of the line operation state;

[0022] determining whether all local deviation values satisfy a preset threshold value;

[0023] If yes, merging the allocation result of the current power-load distribution matrix as a global topology reconfiguration strategy.

[0024] Optionally, when the determination result of the step of determining whether all local deviation values satisfy a preset threshold value is no, the following steps are performed.

[0025] Step S401: adjusting elements with local deviation values exceeding the limit by using a preset adjustment strategy to obtain an adjusted power-load distribution matrix;

[0026] Step S402: calculating all local deviation values of the adjusted power-load distribution matrix;

[0027] Step S403: determining whether all local deviation values satisfy a preset threshold value or the number of iterations reaches a predetermined upper limit;

[0028] If no, performing steps S401-S403;

[0029] If yes, performing step S404;

[0030] Step S404: if all local deviation values satisfy a preset threshold value, merging the allocation result of the current power-load distribution matrix as a global topology reconfiguration strategy.

[0031] Optionally, step S404 further includes:

[0032] If the number of iterations reaches a predetermined upper limit, obtaining the sum of all local deviation values of the power-load distribution matrix in each iteration, and merging the allocation result of the power-load distribution matrix with the smallest sum of all local deviation values to generate a global topology reconfiguration strategy.

[0033] Optionally, the step of calculating a local deviation value for each row and each column of the power-load distribution matrix includes:

[0034] According to the first time series data, calculating the available output of each power supply node in the current predetermined time period under the line operation state constraint;

[0035] According to the second time series data, calculating the absorbable demand of each load node in the current predetermined time period under the line operation state constraint;

[0036] For each row of the power-load distribution matrix, calculating the difference between the available output of the corresponding power supply node and each element in the row to obtain a power supply node deviation value;

[0037] For each column of the power-load distribution matrix, a difference between the absorbable demand of the corresponding load node and each element of the column is calculated to obtain a load node deviation value;

[0038] The power node deviation value and the load node deviation value are weighted and summed according to a preset weight to obtain the local deviation value.

[0039] Optionally, the adjusting, by the preset adjustment strategy, of the element whose local deviation value exceeds the limit includes:

[0040] obtaining a Lagrange multiplier matrix updated last time;

[0041] solving the current power-load distribution matrix and the Lagrange multiplier matrix in parallel to obtain local tentative allocated power of each power node and each load node updated;

[0042] updating the current power-load distribution matrix based on the local tentative allocated power by using a consensus average method;

[0043] updating the Lagrange multiplier matrix based on the local tentative allocated power and the current power-load distribution matrix updated.

[0044] Optionally, the adjusting, by the preset adjustment strategy, of the element whose local deviation value exceeds the limit includes:

[0045] determining whether a target adjustment point is a power node with a maximum current local deviation value;

[0046] If yes, proportionally reducing tentative allocated power of each load node adjacent to the power node according to an unmet degree of the absorbable demand of each load node;

[0047] If no, the target adjustment point is a load node with a maximum current local deviation value;

[0048] proportionally increasing tentative allocated power of each power node adjacent to the load node according to a residual capacity of each power node adjacent to the load node.

[0049] Optionally, the proportionally reducing tentative allocated power of each load node adjacent to the power node according to an unmet degree of the absorbable demand of each load node includes:

[0050] calculating a difference between the absorbable demand of each load node adjacent to the power node and the current tentative allocated power, and dividing the difference by the absorbable demand to obtain a percentage of the unmet degree;

[0051] Then, the excess output of the power supply node is allocated to each load node according to the percentage of the unmet degree, and the allocated power of each load node is reduced accordingly.

[0052] Optionally, the allocated power of each power supply node adjacent to the load node is proportionally increased according to the residual capacity of the power supply node, including:

[0053] The difference between the available output and the current allocated power of each power supply node adjacent to the load node is calculated to obtain the residual capacity;

[0054] The unmet demand of the load node is allocated to each power supply node according to the proportion of the residual capacity, and the allocated power of each power supply node is increased accordingly.

[0055] The present application has the following beneficial effects:

[0056] The method of the present application effectively integrates the time and space characteristics of the power supply side and the load side in the urban-rural junction distribution network, improves the operation flexibility and efficiency of the distribution network, reduces the waste of power resources, enhances the system's ability to cope with load fluctuations, and at the same time, fully utilizes the existing distribution network resources, avoids large-scale new transmission lines, and reduces the system operation cost. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 A flowchart of a distribution network topology dynamic reconstruction method based on time and space feature fusion provided by an embodiment of the present application;

[0058] Figure 2 A flowchart of generating a global topology reconstruction strategy provided by an embodiment of the present application. DETAILED DESCRIPTION

[0059] In order to facilitate the understanding of those skilled in the art, the present application will be further described below in conjunction with the embodiments and the drawings, and the content mentioned in the embodiments is not a limitation of the present application.

[0060] In the traditional existing distribution network topology dynamic reconstruction method, the power supply side and the load side at the urban-rural junction present significant differences in time and space characteristics, resulting in insufficient dynamic coordination ability. The distributed power supply connected to the rural power grid is scattered and has strong volatility, and the load demand of the urban power grid is concentrated and has large peak-valley difference. The traditional method uses a global unified optimization strategy, which cannot effectively integrate the time and space dynamic correlation of power supply output and load demand, causing conflicts between local operating states and global constraint conditions.

[0061] If the above problems are not solved, the distribution network will face the loss of operation efficiency caused by the mismatch of time and space for a long time. The distributed power supply in rural areas will produce abandoned electricity during the output peak, reducing energy utilization. The power supply capacity in urban areas will be limited during the load peak, increasing line loss and the risk of equipment overload. The superposition of the above two will increase the difficulty of power balance in the distribution network, forcing the system to rely on higher capacity standby power supply or frequently start-stop peak-shaving units, leading to the deterioration of operation economy.

[0062] To solve the above technical problems, as shown in Figure 1 The application provides a distribution network topology dynamic reconstruction method based on time-space feature fusion, which comprises the following steps:

[0063] Step S101: acquiring first time series data, second time series data and line operation state; the first time series data represents the power generation data of each power supply node on the power supply side of the target boundary; the second time series data represents the power consumption data of each load node on the load side of the target boundary, and the line operation state represents the line state between each power supply node and each load node at the target boundary;

[0064] The target boundary here is a certain urban-rural boundary distribution network. The rural side contains a certain number of distributed photovoltaic power supply nodes, and the urban side contains a certain number of load nodes. The power generation data of the power supply nodes and the power consumption data of the load nodes are collected every 15 minutes within 24 hours to form the first time series data and the second time series data respectively, and the operation state data of the lines between the nodes are acquired.

[0065] Step S102: generating first time-space feature parameters based on the first time series data and the line operation state;

[0066] The time-space feature parameters refer to parameters reflecting the operation state of the power supply nodes or the load nodes at different time and space dimensions. The first time series data and the line operation state are jointly analyzed, the power change trend of the time series and the correlation of the spatial topology are fused, the dynamic behavior characteristics of the nodes are captured, and the first time-space feature parameters are generated. For the photovoltaic light nodes, the first time-space feature parameters include sunrise and sunset time, maximum power generation, power fluctuation caused by cloud change and the like.

[0067] Step S103: generating second time-space feature parameters based on the second time series data and the line operation state;

[0068] The second time series data and the line operation state are jointly analyzed, the daily load curve of the time series and the correlation of the spatial topology are fused, the dynamic behavior characteristics of the nodes are captured, and the second time-space feature parameters are generated. For the load nodes, the second time-space feature parameters include daily load curve, peak-valley difference, power consumption type and the like.

[0069] Step S104: generating an initial local topology reconstruction strategy based on the first spatio-temporal characteristic parameter and the second spatio-temporal characteristic parameter;

[0070] According to the first spatio-temporal characteristic parameter and the second spatio-temporal characteristic parameter, an initial topology reconstruction strategy is generated for each node. The initial topology reconstruction strategy refers to a local strategy independently generated by each power supply node or load node based on its own spatio-temporal characteristic parameter, which is used to represent the preliminary operation scheme of the node in the current time period. The strategy only considers the available output or absorbable demand of the node itself and the line operation state, and does not involve the influence of adjacent nodes.

[0071] Step S105: modifying the initial topology reconstruction strategy to obtain a first topology reconstruction strategy;

[0072] The first topology reconstruction strategy refers to the interactive adjustment of local strategies by adjacent nodes based on each other's spatio-temporal characteristic parameters. Specifically, a distributed negotiation mechanism can be used to achieve this. By exchanging suggestion sets and evaluating feasibility, the local strategies are preliminarily coordinated to achieve supply-demand balance within the neighborhood. The above modification process is as follows:

[0073] First, the initial topology reconstruction strategy of each node is converted into an initial suggestion set containing a first number of suggestion items. The suggestion items include at least suggestion type, suggestion power amount, suggestion duration and suggestion confidence. The suggestion type includes supply type, demand type and relay type. The conversion of the initial suggestion set refers to expressing the strategy content through structured data. The supply type suggestion corresponds to power output distribution, the demand type suggestion corresponds to load demand adjustment, and the relay type suggestion corresponds to line transmission capacity optimization. The suggestion power amount is expressed in per-unit value, the suggestion duration is in minutes, and the suggestion confidence is calculated based on the matching degree of historical data, with a threshold value that can be set by the user.

[0074] Then the adjacent power nodes are exchanged with the initial proposed set of load nodes, after the exchange is completed, all initial proposed sets received by each power node and each load node are judged for feasibility according to the corresponding space-time characteristic parameters and the corresponding line operation state, and feasible proposed items, infeasible proposed items and over-limit proposed items are generated, and the infeasible proposed items and the over-limit proposed items are marked as limited proposed items, and the limited proposed items are modified under the constraints of the space-time characteristic parameters of the current power node / load node and the corresponding line operation state. The modification of the limited proposed items refers to adjusting the power quantity by linear programming, and re-distributing the time length based on the upper limit of the line transmission capacity, converting the above feasible proposed items and the modified limited proposed items, thereby generating a first topology reconstruction strategy, and associating the first topology reconstruction strategy to the corresponding power node or load node. For example, a certain photovoltaic node sends its power generation prediction curve to the adjacent load node, and the load node modifies the received strategy according to its own power demand curve, in this way, each node can consider the operation characteristics of the adjacent nodes, and form a first topology reconstruction strategy.

[0075] Here, through the above neighborhood modification, information exchange and strategy modification between adjacent nodes are realized, and the accuracy and feasibility of the topology reconstruction strategy are improved.

[0076] Step S106: consistency judgment is performed on the first topology reconstruction strategy to generate a global topology reconstruction strategy;

[0077] The consistency judgment refers to global coordination of the first topology reconstruction strategy to minimize the difference between power output and load demand, which can be realized by using a matrix distribution model combined with an iterative optimization algorithm, by constructing a power-load distribution matrix and calculating a local deviation value, and gradually adjusting to a global optimal solution. The generation of the global topology reconstruction strategy refers to the formation of a final operation scheme that meets the line constraints by synthesizing the modified strategies of each node, which can be realized by using a multi-agent collaborative optimization framework, by combining distributed calculation and centralized coordination, to ensure that the strategy realizes optimal resource allocation under the premise of meeting the operation safety. In combination with Figure 2 The specific process of generating the global topology reconstruction strategy is as follows:

[0078] Step S201: according to the first topology reconstruction strategy, a power-load distribution matrix of the target boundary is constructed;

[0079] The rows of the power-load distribution matrix M represent the power nodes on the power side, and the columns represent the load nodes on the load side, and the matrix elements represent the tentative distribution of the power node i to the load node j in the current predetermined time period;

[0080] Step S202: For each row and each column of the power-load distribution matrix, a local deviation value is calculated, which quantifies the difference between the power supply capacity of the power node and the demand absorption capacity of the corresponding load node under the line operation state constraint.

[0081] The process of calculating the local deviation value in step S202 is as follows:

[0082] According to the first time series data, the power supply capacity of each power node under the line operation state constraint in the current predetermined time period is calculated. Specifically, for each power node, its power generation curve in the predetermined time period is obtained, and combined with the capacity limit of the transmission line connected to the power node, the maximum power that can be actually output by the power node in the predetermined time period is calculated.

[0083] According to the second time series data, the demand absorption capacity of each load node under the line operation state constraint in the current predetermined time period is calculated. Specifically, for each load node, its power demand curve in the predetermined time period is obtained, and combined with the capacity limit of the transmission line connected to the load node, the maximum power that can be actually absorbed by the load node in the predetermined time period is calculated.

[0084] For each row of the power-load distribution matrix, the difference between the power supply capacity of the corresponding power node and each element in the row is calculated to obtain the power node deviation value. For example, for the i-th row power node, the power supply capacity is the sum of each element in the row is , and the power node deviation value is .

[0085] For each column of the power-load distribution matrix, the difference between the demand absorption capacity of the corresponding load node and each element in the column is calculated to obtain the load node deviation value. For example, for the j-th column load node, the demand absorption capacity is the sum of each element in the column is , and the load node deviation value is .

[0086] The power node deviation value and the corresponding load node deviation value are weighted and summed according to the preset weight to obtain the local deviation value. Specifically, the weight of the power node deviation value is , the weight of the load node deviation value is , and the local deviation value is .

[0087] Specifically, in the power supply node deviation calculation, a sliding time window algorithm is adopted to perform trend prediction on the first timing data, and the maximum transmissible power value is determined in combination with the line current thermal stability limit. In the load node deviation calculation, a typical demand curve is extracted through load clustering analysis, and a demand acceptance model is established by superimposing the line voltage drop constraint. The weighted summation process introduces normalization processing, maps the power supply deviation and the load deviation to the same dimension space, and then generates a comprehensive deviation index by using linear combination. This index reflects the power supply side output redundancy and the load side demand gap, and provides a quantitative basis for subsequent matrix adjustment. For example, when the deviation value of a certain power supply node is +50kW and the associated load node deviation value is -30kW, after 0.6 and 0.4 weight calculation, the comprehensive local deviation value is 18kW, indicating that there is a power supply surplus in this area that needs to be redistributed.

[0088] Step S203: judging whether all the calculated local deviation values meet the preset threshold value;

[0089] If yes, the allocation result of the current power-load distribution matrix is merged as the global topology reconstruction strategy.

[0090] If no, steps S401-S404 are executed.

[0091] If the judgment result is yes, it means that the overall supply-demand difference of the current power-load distribution matrix elements is the smallest, and the linear capacity constraint condition is met at the same time. At this time, the allocation result of the power-load distribution matrix is merged as the global topology reconstruction strategy, so that step S107 is executed.

[0092] If the judgment result is no, the following steps are executed:

[0093] Step S401: using a preset adjustment strategy to adjust the elements whose local deviation values exceed the limit, to obtain an adjusted power-load distribution matrix.

[0094] The application provides two preset adjustment strategies. In one example, the preset adjustment strategy is implemented based on the alternating direction multiplier method, specifically including:

[0095] First, the Lagrange multiplier matrix updated last time is obtained. Then, the current power-load distribution matrix and the Lagrange multiplier matrix are solved in parallel to obtain the updated local allocation power of each power supply node and each load node.

[0096] Then, according to the local allocation power, the consensus average method is used to update the current power-load distribution matrix. For example, the local allocation power of all nodes can be arithmetically averaged to obtain the updated power-load distribution matrix.

[0097] Finally, the Lagrange multiplier matrix is updated according to the local power to be allocated and the updated current power-load allocation matrix. Specifically, the Lagrange multiplier matrix can be updated using gradient ascent method, i.e. , wherein, is a step size parameter, is the Lagrange multiplier vector of the power node i, is the Lagrange multiplier vector of the load node j, and are the i-th row and j-th column of the updated allocation matrix, and are the corresponding row and column of the power-load allocation matrix of the last iteration.

[0098] At the beginning, the Lagrange multiplier matrix needs to be initialized, and the process of initialization is as follows:

[0099] A global optimization problem is constructed with the objective of minimizing the global power imbalance and the constraints of the available power output of each power node, the absorbable demand of each load node, and the line operation state. For example, the global power imbalance can be defined as the sum of squares of the difference between the power output of all power nodes and the demand of all load nodes.

[0100] A Lagrange multiplier matrix associated with the allocation matrix M is introduced and initialized. The dimension of the Lagrange multiplier matrix is the same as that of the allocation matrix M, and the initial value can be set to a zero matrix.

[0101] The global optimization problem is decomposed into sub-problems distributed in each power node and load node, and the objective function of each sub-problem includes a global consistency constraint term based on the Lagrange multiplier matrix. Specifically, for the power node i, the objective function of its sub-problem can be expressed as minimizing wherein, is the available power output of the node i, is the element of the i-th row and j-th column in the allocation matrix M, is the vector of the i-th row in the Lagrange multiplier matrix, is the i-th row of the allocation matrix M in the last iteration, and similarly, for the load node j, the objective function of its sub-problem can be expressed as wherein, is the absorbable demand of the node j, is the vector of the j-th column in the Lagrange multiplier matrix, is the j-th column of the allocation matrix M in the last iteration.

[0102] The preset adjustment strategy using the alternating direction multiplier method can effectively coordinate the power distribution between multiple power supply nodes and load nodes, realize global power balance, and reduce the calculation complexity while ensuring global consistency and improving the scalability of the algorithm by introducing Lagrange multipliers and distributed optimization. In addition, the parallel computing characteristics in the iteration process enable the scheme to adapt to the dynamic reconstruction requirements of large-scale distribution networks and improve the reconstruction efficiency.

[0103] In another example, the preset adjustment strategy is as follows:

[0104] First, it is determined whether the target adjustment node is the power supply node with the largest current local deviation value. If yes, the allocated power of the power supply node to each load node is proportionally reduced according to the unmet degree of the absorbable demand of each load node adjacent to the power supply node. Specifically, the difference between the absorbable demand and the current allocated power of each load node adjacent to the power supply node is calculated, and the difference is divided by the absorbable demand to obtain the unmet degree percentage. Then, the excess output of the power supply node is distributed to each load node in proportion to the unmet degree percentage, and the allocated power of each load node is correspondingly reduced.

[0105] If the determination result is no, it means that the target adjustment node is the load node with the largest current local deviation value. The allocated power of each power supply node adjacent to the load node is proportionally increased according to the remaining capacity of the available output of each power supply node adjacent to the load node. Specifically, the difference between the available output and the current allocated power of each power supply node adjacent to the load node is calculated to obtain the remaining capacity. Then, the unmet demand of the load node is distributed to each power supply node in proportion to the remaining capacity, and the allocated power of each power supply node is correspondingly increased.

[0106] Through the above proportional adjustment strategy, the node with the largest deviation can be adjusted preferentially in the iteration process, effectively reducing the overall deviation. Different adjustment strategies are used for power supply nodes and load nodes to fully utilize the remaining capacity or unmet demand of each node, so that the power supply output and load demand are more matched. Through proportional distribution, the single node is avoided from bearing too much adjustment amount, and more balanced and reasonable adjustment is achieved. This dynamic adjustment strategy can quickly converge to a solution that satisfies the constraint conditions, improving the efficiency and accuracy of the distribution network topology reconstruction.

[0107] Step S402: Calculate all local deviation values of the adjusted power supply-load distribution matrix;

[0108] The process of calculating the local deviation value here is the same as the calculation process in step S202, and reference can be made. Here, no detailed description is given.

[0109] Step S403: judging whether all the local deviation values meet the preset threshold or the iteration number reaches the predetermined upper limit;

[0110] The above two judgment conditions only need to meet one of them, that is, when all the local deviation values meet the preset threshold, or when the iteration number reaches the predetermined upper limit, step S404 can be executed.

[0111] If neither of the above two conditions is met, that is, all the local deviation values do not meet the preset threshold, and the iteration number also does not reach the predetermined upper limit, steps S401-403 are executed.

[0112] Step S404: if all the deviation values meet the preset threshold, the allocation results of the current power-load distribution matrix are merged into the global topology reconstruction strategy.

[0113] The predetermined upper limit of the iteration number is equivalent to a bottom line mechanism, which is to avoid too many iterations. After each iteration and execution of the adjustment strategy, the weighted sum of all power node deviation values and load node deviation values in the current distribution matrix is calculated, and the sum is stored in association with the current iteration number. When the iteration counter reaches the predetermined upper limit, the backtracking mechanism is triggered to extract all historical deviation and data from the storage space for sorting, locate the iteration number corresponding to the minimum value, extract the power-load distribution matrix corresponding to the number, and merge the allocation results of the power-load distribution matrix to generate the global topology reconstruction strategy. This mechanism ensures that even if the iteration does not completely converge, a relatively optimal topology reconstruction strategy can be output under limited computing resources, avoiding power distribution imbalance caused by forced termination of iteration.

[0114] The predetermined upper limit set here selects the optimal intermediate result as the final scheme in the case where the iteration cannot converge to meet the threshold requirement, thereby avoiding algorithm failure caused by failure to converge and improving the robustness of the algorithm.

[0115] Step S107: based on the global topology reconstruction strategy, corresponding switch operation instructions are generated and output to the command execution end of the power distribution network.

[0116] According to the global topology reconstruction strategy, a specific switch operation instruction sequence can be generated. For example, during the peak period of photovoltaic power generation, the instruction increases the power supply path from the rural side to the city side, and during the peak period of urban load, more standby power sources are connected. These instructions are sent to the switch device execution end through the SCADA (Supervisory Control and Data Acquisition, data acquisition and monitoring control system) system to realize dynamic adjustment of the power distribution network topology.

[0117] In summary, the method provided in the application realizes effective fusion of the time and space characteristics of the power supply side and the load side in the urban-rural junction distribution network, improves the operation flexibility and efficiency of the distribution network, reduces the waste of power resources, enhances the ability of the system to cope with load fluctuations, and at the same time, fully utilizes the existing distribution network resources, avoids large-scale new transmission lines, and reduces the system operation cost.

[0118] The above embodiments are preferred implementation schemes of the application, in addition to this, the application can also be implemented in other ways, any obvious replacement without departing from the technical scheme concept of the application is within the protection scope of the application.

[0119] In order for those skilled in the art to more conveniently understand the improvements of the present application over the prior art, some drawings and descriptions of the present application have been simplified, and some other elements have also been omitted from the present application file for the sake of clarity, and those skilled in the art should realize that these omitted elements can also constitute the content of the present application.

Claims

1. A method for dynamic topology reconfiguration of distribution networks based on spatiotemporal feature fusion, characterized in that, include: Acquire the first time series data, the second time series data, and the line operating status; The first time-series data represents the power generation data of each power node on the power supply side at the target boundary; The second time-series data represents the power demand data of each load node on the load side at the target boundary, and the line operation status represents the line status between each power node and each load node at the target boundary. Based on the first time-series data and the line operating status, a first spatiotemporal feature parameter is generated; Based on the second time-series data and the line operating status, a second spatiotemporal feature parameter is generated; Based on the first spatiotemporal feature parameters and the second spatiotemporal feature parameters, a local initial topology reconstruction strategy is generated; The initial topology reconstruction strategy is modified by neighborhood adjustment to obtain the first topology reconstruction strategy; Perform a consistency check on the first topology reconstruction strategy to generate a global topology reconstruction strategy; Based on the global topology reconfiguration strategy, corresponding switch operation commands are generated and output to the command execution terminal of the distribution network.

2. The method for dynamic reconfiguration of distribution network topology based on spatiotemporal feature fusion according to claim 1, characterized in that, The initial topology reconstruction strategy is modified by neighborhood adjustment to obtain a first topology reconstruction strategy. include: The initial topology reconstruction strategy is converted into an initial suggestion set containing a first number of suggestion items; Exchange the initial proposal sets of adjacent power nodes and load nodes; The feasibility of all initial suggestion sets received by each power node and each load node is determined based on the corresponding spatiotemporal characteristic parameters and the corresponding line operating status. Feasible suggestions, infeasible suggestions, and over-limit suggestions are generated, and infeasible suggestions and over-limit suggestions are marked as restricted suggestions. The aforementioned restricted recommendations are modified under the constraints of the spatiotemporal characteristic parameters of the current power node or load node and the corresponding line operating status. The feasible suggestions and the modified restricted suggestions are transformed to generate a first topology reconfiguration strategy, and the first topology reconfiguration strategy is associated with the corresponding power node or load node.

3. The method for dynamic topology reconfiguration of distribution networks based on spatiotemporal feature fusion according to claim 1, characterized in that, The step of performing a consistency judgment on the first topology reconstruction strategy and generating a global topology reconstruction strategy includes: Based on the first topology reconfiguration strategy, a power-load allocation matrix is ​​constructed at the target boundary. For each row and each column of the power-load distribution matrix, a local deviation value is calculated. The local deviation value is used to quantify the difference between the power output available at the power node and the demand absorbed by the corresponding load node under the constraints of the line operating state. Determine whether all local deviation values ​​meet the preset threshold; If so, the allocation results of the current power-load allocation matrix will be merged into a global topology reconfiguration strategy.

4. The method for dynamic topology reconfiguration of distribution networks based on spatiotemporal feature fusion according to claim 3, characterized in that, If the result of judging whether all local deviation values ​​meet the preset threshold is negative, the following steps are executed; Step S401: Using a preset adjustment strategy, adjust the elements whose local deviation values ​​exceed the limit to obtain the adjusted power-load distribution matrix; Step S402: Calculate all local deviation values ​​of the adjusted power-load distribution matrix; Step S403: Determine whether all the local deviation values ​​meet the preset threshold or the number of iterations reaches the predetermined upper limit; If not, proceed to steps S401-S403; If so, proceed to step S404; Step S404: If all deviation values ​​meet the preset threshold, merge the current power-load allocation matrix into a global topology reconfiguration strategy.

5. The method for dynamic topology reconfiguration of distribution networks based on spatiotemporal feature fusion according to claim 4, characterized in that, Step S404 also includes: If the number of iterations reaches a predetermined upper limit, the sum of all local deviation values ​​of the power-load allocation matrix in each iteration is obtained, and the allocation results of the power-load allocation matrix with the smallest sum of all local deviation values ​​are merged to generate a global topology reconstruction strategy.

6. The method for dynamic reconfiguration of distribution network topology based on spatiotemporal feature fusion according to claim 4, characterized in that, The calculation of local deviation values ​​for each row and each column of the power-load distribution matrix includes: Based on the first time series data, calculate the available power output of each power node under the constraints of the line operating state within the current predetermined time period; Based on the second time series data, calculate the absorbable demand of each load node under the constraints of the line operation status within the current predetermined time period. For each row of the power-load allocation matrix, calculate the difference between the available output of the corresponding power node and the sum of the elements in that row to obtain the power node deviation value; For each column of the power-load distribution matrix, calculate the difference between the absorbable demand of the corresponding load node and the sum of the elements in that column to obtain the load node deviation value. The local deviation value is obtained by weighting and summing the power node deviation value and the load node deviation value according to a preset weight.

7. The method for dynamic topology reconfiguration of distribution networks based on spatiotemporal feature fusion according to claim 4, characterized in that, The step of using a preset adjustment strategy to adjust elements with local deviation values ​​exceeding limits includes: Get the Lagrange multiplier matrix after the last update; The current power-load allocation matrix and the Lagrange multiplier matrix are solved in parallel to obtain the updated local proposed power allocation for each power node and each load node. Based on the locally proposed power allocation, the current power-load allocation matrix is ​​updated using a consensus averaging method; The Lagrange multiplier matrix is ​​updated based on the locally proposed power allocation and the updated current power-load allocation matrix.

8. The method for dynamic topology reconfiguration of distribution networks based on spatiotemporal feature fusion according to claim 4, characterized in that, The method of using a preset adjustment strategy to adjust elements whose local deviation values ​​exceed the limit also includes: Determine whether the target adjustment point is the power node with the largest current local deviation value; If so, the power to be allocated by the power source to each load node shall be reduced proportionally based on the degree of unmet demand of each load node adjacent to the power source node. If not, the target adjustment point is the load node with the largest current local deviation value; Then, based on the remaining output capacity of each power node adjacent to the load node, the proposed power allocation to each power node adjacent to the load node is increased proportionally.

9. The method for dynamic topology reconfiguration of a distribution network based on spatiotemporal feature fusion according to claim 8, characterized in that, The step involves proportionally reducing the power allocation from the power source to each load node based on the degree of unmet demand of each load node adjacent to the power source, including: Calculate the difference between the absorbable demand of each load node adjacent to the power node and the current planned power allocation; divide the difference by the absorbable demand to obtain the percentage of unmet demand; Then, based on the percentage of unmet needs, the excess output of the power node is allocated to each load node, and the planned power allocation to each load node is reduced accordingly.

10. The method for dynamic reconfiguration of distribution network topology based on spatiotemporal feature fusion according to claim 8, characterized in that, The step involves proportionally increasing the planned power allocation to each power source adjacent to the load node based on the remaining output capacity of each power source adjacent to the load node, including: Calculate the difference between the available output of each power node adjacent to the load node and the current planned power allocation to obtain the remaining capacity; Based on the proportion of remaining capacity, the unmet demand of this load node is allocated to each power supply node, and the planned power allocation to each power supply node is increased accordingly.

Images