Power distribution network dynamic line loss adaptive partition compensation method and system

Abstract

The present application relates to power distribution network line loss governance technical field, specifically to a kind of power distribution network dynamic line loss adaptive partition compensation method and system.Through the construction line loss sensitivity matrix to quantify the loss coupling strength between nodes, and using double-channel competition fusion structure, the sensitivity structure view based on electrical coupling and the behavior pattern view based on injection power time series mutual information are evolved game fusion, to generate dynamic partition scheme considering physical structure and operating state;Within the partition, after locking candidate compensation node, start the loop bounding process to obtain the initial value of the adjustment amount of the key node in the partition;In the device-level control layer, introduce the marginal contribution mechanism of Sharply value to realize the collaborative distribution of adjustment amount among heterogeneous compensation devices, and establish device three-section delay spectrum to respond to the technical scheme of jointly determining the time staggered compensation sequence by action slope and adjustment stability margin, so as to achieve the beneficial effect of accurately and stably reducing dynamic line loss in the whole network.

Description

Technical Field

[0001] This invention relates to the field of power distribution network line loss management technology, specifically to a dynamic line loss adaptive zoning compensation method and system for power distribution networks. Background Technology

[0002] With the accelerated construction of new power systems, the penetration rate of flexible loads such as distributed power sources and electric vehicle charging piles in distribution networks continues to rise, and the power flow distribution of distribution networks exhibits significant time-varying and uncertain characteristics.

[0003] Traditional distribution network line loss compensation strategies are mostly based on offline calculations using static network topology and typical daily load curves, generating fixed zoning schemes and compensation capacities. This traditional approach cannot adapt to operating scenarios with real-time fluctuations in source and load power. Specifically, existing methods often treat the entire network as a homogeneous whole or perform coarse zoning based on fixed administrative regions when formulating compensation schemes, ignoring the differences in coupling strength between nodes due to differences in the cross-influence of electrical distance and power regulation on line losses. The static zoning mechanism of existing methods results in compensation equipment being unable to accurately deliver reactive or active power compensation to the critical nodes with the greatest impact on line losses during load shifts or drastic fluctuations in renewable energy output. Furthermore, uncontrolled power exchange at zoning boundary nodes can lead to exceeding limits, making it difficult to meet the real-time requirements of lean distribution network operation for loss reduction strategies.

[0004] At the level of coordinated control of compensation equipment, existing technologies generally employ simultaneous command issuance or simple priority polling for equipment scheduling, failing to fully consider the heterogeneous characteristics of various types of compensation equipment in terms of response latency, adjustment accuracy, and health status. In actual operating conditions controlling network communication traffic fluctuations, uniform command triggering can easily cause oscillations and overshoots in the system adjustment process, leading to excessive operation of some degraded equipment while high-precision equipment remains idle. Simultaneously, due to the lack of precise temporal modeling and staggered arrangement of equipment response transition processes, the transient adjustment behaviors of various compensation equipment may overlap, creating superimposed impacts. This not only fails to effectively reduce dynamic line losses but may also worsen the voltage quality of partition nodes, leaving dynamic line loss mitigation projects in a long-term state of passive remediation, lacking forward-looking and self-regulating adaptive adjustment capabilities. Summary of the Invention

[0005] To overcome the problems of existing technologies, such as the inability of static zoning to adapt to real-time fluctuations in source load and the lack of time-series fine management in the collaborative control of compensation equipment, this invention provides a dynamic line loss adaptive zoning compensation method and system for distribution networks. Under the constraints of adaptive dynamic zoning, the compensation equipment can smoothly adjust peak shifts, avoiding dynamic adjustment oscillations and boundary overruns, thereby achieving the beneficial effect of accurately and stably reducing dynamic line losses across the entire network.

[0006] The technical solution of this application specifically includes:

[0007] According to one aspect of this application, a dynamic line loss adaptive zoning compensation method for distribution networks is provided, comprising:

[0008] The line loss sensitivity matrix is ​​formed by the change in injected power at the nodes and the line impedance, and the column vectors are used to represent the direction of line loss sensitivity. The loss coordination degree between nodes is calculated, and the loss coupling strength between nodes is obtained by combining the electrical distance.

[0009] A dual-channel competitive fusion structure is initiated for dynamic partitioning. The first channel obtains a sensitivity partitioning view based on the spectral clustering loss coupling strength, while the second channel obtains a behavioral partitioning view based on the clustering power fluctuation mutual information. A game payoff matrix is ​​constructed using the cross-partition line loss interaction entropy change and the intra-partition behavioral homogeneity change, and a dynamic partitioning scheme is derived. Partition boundary constraints are constructed based on the boundary power exchange sensitivity and voltage coupling coefficient.

[0010] Within each partition, a line loss potential energy field is constructed using the increase in active power loss, the node voltage deviation, and the change in homogeneity of behavior within the partition. A candidate compensation set is obtained by tracing back from the peak value of the potential energy field along the direction of line loss sensitivity. The line loss potential energy field is recalculated by iteratively adjusting the candidate compensation set, and the boundary constraints are verified. The valid ones are retained and the invalid ones are discarded until the target is met or the key nodes and the initial value of the adjustment are output.

[0011] A dynamic performance profile including response delay, adjustment accuracy, and health score is established for the adjustable equipment at key nodes. The adjustment amount is allocated to each zone according to the Shapley marginal contribution of each equipment to the potential energy field decline to obtain the initial allocation scheme. The boundary constraints are verified, and if the limit is exceeded, the adjustment is transferred to the second healthiest equipment in the same zone or the equipment in the neighboring zone. The compensation amount is obtained by iteratively satisfying the given value.

[0012] The communication traffic conditions are identified to construct a three-segment delay spectrum; the device with the largest upper bound of the delay spectrum is used as the timing anchor point, the health score is mapped to the response action slope, the adjustment accuracy is mapped to the adjustment stability margin, the command start time is determined, a timing interleaved compensation sequence is formed and issued for execution.

[0013] As a further option of the method of the present invention, the formation of the line loss sensitivity matrix includes:

[0014] For those with Each node For a radial distribution network with branches, forward-backward power flow calculations are performed to obtain the current amplitude of each branch. and its resistance With current Define nodes Changes in injected active power affect the branch The sensitivity factor of active power loss is ,in branch road The set of downstream nodes, For indicator functions, when node belong If the value is 1, then 0 is used;

[0015] Traverse all branches and nodes to construct a line loss sensitivity matrix And extract the first element of the matrix. column vector As a node The line loss sensitivity direction is obtained by normalizing the column vector to obtain the unit direction vector. It is used to characterize the spatial distribution of the impact of node power regulation on the branch loss of the entire network.

[0016] As a further option of the method of the present invention, the step of obtaining the inter-node loss coupling strength includes:

[0017] For nodes With nodes Utilizing its line loss sensitivity column vector and Define loss coupling strength ,in The electrical distance is represented by the sum of the impedance moduli of all branches on the shortest path between two nodes. The preset electrical distance attenuation coefficient, yes The transpose of the column vector is converted to 1× 3D row vectors For nodes and Degree of synergy in losses between them;

[0018] The loss coupling strength reflects the combined effect of the cosine similarity of the sensitivity directions of the two nodes and the electrical distance on the regulation interaction, forming the edge weight matrix for spectral clustering.

[0019] As a further option of the method of the present invention, the first channel of the dual-channel competitive fusion structure includes:

[0020] Construct a weighted undirected graph using loss coupling strength as edge weights, solve the eigenvalue problem of the graph Laplacian matrix, and take the first edge weight. The node embedding features are formed by the eigenvectors corresponding to the smallest non-zero eigenvalues, and the number of partitions is automatically determined using the eigenvalue intervals. Perform on embedded features - Mean clustering yields a sensitivity partition view;

[0021] The second channel of the dual-channel competitive fusion structure includes:

[0022] The time-series fluctuation sequences of active and reactive power injected into each node are discretized with equal width. The mutual information value of power fluctuation between any two nodes is calculated, and the mutual information matrix is ​​formed as a similarity matrix. Clustering is then applied to generate a behavioral partition view.

[0023] As a further option of the method of the present invention, the steps of constructing the game payoff matrix and evolving the dynamic partitioning scheme include:

[0024] Identify the set of boundary nodes that are inconsistent between the two views, for boundary nodes. From the current partition Assigned to adjacent partitions The corresponding cross-regional line loss cross-entropy change Changes in homogeneity of behavior within the region Perform weighted summation to construct the game payoff function;

[0025] A finite-round synchronous update strategy is adopted, in which each boundary node changes its partitioning according to the payment increment with probability, and the evolution is repeated until the boundary assignment is stable, outputting the optimal dynamic partitioning scheme that integrates structural coupling and behavioral patterns.

[0026] As a further option of the method of the present invention, the partition boundary constraint construction step includes:

[0027] Extract the power exchange sensitivity of each boundary node to the associated tie branch, and determine the upper limit of the allowable power exchange of the boundary node by combining the thermal stability limit and operating margin of the tie branch.

[0028] Extract the voltage sensitivity of the boundary nodes to the adjacent partition nodes, determine the voltage deviation correlation limit based on the allowable voltage deviation range of the adjacent partitions, and form a partition boundary constraint set that includes the upper limit of power exchange and the voltage deviation correlation limit.

[0029] As a further option of the method of the present invention, the step of constructing the line loss potential energy field includes:

[0030] For any node within the partition Construction line loss potential energy : ,in For the node The increase in total loss within the partition caused by increasing unit power injection. For partitioning The set of included branches; For nodes The L2 norm of voltage deviation; For nodes The estimated change in the homogeneity of behavior within the adjusted region; Preset weights;

[0031] Traversing all nodes in the partition forms a scalar field describing the severity and distribution of line loss, with the potential energy peak region corresponding to the priority compensation region.

[0032] As a further option of the method of the present invention, the step of obtaining the candidate compensation set includes:

[0033] Identify the set of local peak nodes in the potential energy field that are above the neighborhood mean;

[0034] For each peak node, backtrack along its normalized line loss sensitivity direction to the power source direction, selecting the upstream adjacent node with the smallest angle with the negative gradient direction hop by hop until the source node such as the substation bus or distributed power grid connection point is reached. During the tracing process, the power supply path with the best matching sensitivity direction is selected.

[0035] After deduplicating the source nodes traced back, a set of candidate compensation nodes is formed, ensuring that the candidate nodes are located at the beginning of the power supply path that contributes the most to the power of the high-loss area.

[0036] As a further option of the method of the present invention, the step of processing the candidate compensation set to output key nodes and initial values ​​of adjustment includes:

[0037] Initialize the selected compensation node set to empty, iteratively select candidate nodes that are most consistent with the negative potential gradient, apply a small amount of active power adjustment, and use the sensitivity matrix to estimate or recalculate the new potential field using fast power flow.

[0038] If the new potential energy peak decreases and the power exchange and voltage deviation of all boundary nodes do not exceed the partition boundary constraints, then the node is retained and the potential energy field is updated, and its adjustment is accumulated; otherwise, the adjustment is rolled back and the corresponding node is permanently removed from the candidate compensation set.

[0039] The loop continues until the potential energy peak drops to the preset threshold or the candidate compensation set is traversed. The retained nodes are then used as the key nodes that need to be compensated in the partition, and the accumulated adjustment amount is the initial value of the adjustment amount.

[0040] As a further option of the method of the present invention, the dynamic performance profile establishment step includes:

[0041] Collect the command response delay, minimum adjustment step size, or resolution of each adjustable compensation device as the adjustment accuracy, and count the historical cumulative number of actions;

[0042] Define health score ,in The initial health score is perfect. The attenuation coefficient is... For the first The cumulative number of actions performed by the device.

[0043] As a further option of the method of the present invention, the steps of forming the initial allocation scheme and obtaining the compensation amount given value include:

[0044] The decrease in the peak potential field after the equipment combination is put into use is used as the characteristic function of the alliance. The Shapley value of each equipment is calculated by Monte Carlo sampling. After normalization, the adjustment amount allocation weight is obtained. The total adjustment amount of the partition is allocated according to the weight to form the initial scheme, which is subject to the upper limit of equipment capacity.

[0045] The power exchange change of each boundary node is verified based on the power exchange sensitivity. If the limit is exceeded, the device that contributes the most is located and its allocation is reduced. The reduced amount is transferred to the device with the second highest health score in the same partition, or it is borne by the device in the adjacent partition through the boundary interaction node. The adjustment is iterated until all partition boundary constraints are satisfied, and the compensation amount of each device is output.

[0046] As a further option of the method of the present invention, the steps of constructing the three-segment delay spectrum and determining the time sequence anchor point include:

[0047] Monitor and control network communication traffic, and classify operating conditions into light load, normal, and heavy load based on bandwidth utilization and average latency;

[0048] For each adjustable compensation device, the historical response delay distribution under light load, normal, and heavy load conditions is extracted, and the 95th percentile is taken as the upper limit of the delay for each condition to form a three-segment delay spectrum.

[0049] Among all participating compensation devices, the device with the largest upper limit of delay under the current operating condition is selected as the timing anchor point, and its instruction start time is set as the reference time.

[0050] As a further option of the method of the present invention, the step of forming the time-interleaved compensation sequence includes:

[0051] Define the response action slope for each device. ,in , Preset upper and lower limits for slope. This represents the maximum reference value for health. Rate your health.

[0052] Adjust precision Mapping to adjust stability margin The higher the adjustment precision, the better. The smaller the value, the shorter the time required for the device to enter a stable confirmation state after reaching the target value;

[0053] The stable interval time is determined by adding the start time of the anchor point device to the quotient of the adjustment amount and the slope, and the stability margin. The stable interval time of the previous device is used as the instruction start time of the next device. This forms a sequentially connected time-interleaved compensation sequence, and the compensation amount is given to each device according to the time sequence.

[0054] Another aspect of this application provides a dynamic line loss adaptive zoning compensation system for a distribution network, the system comprising:

[0055] The sensitivity analysis and coupling quantization module is used to form a line loss sensitivity matrix from the changes in node injected power and line impedance, with the column vectors representing the line loss sensitivity direction, and to calculate the loss coordination degree between nodes, and to obtain the loss coupling strength between nodes by combining the electrical distance.

[0056] The dual-channel dynamic partitioning and constraint construction module is used to initiate a dual-channel competitive fusion structure for dynamic partitioning. The first channel obtains a sensitivity partitioning view based on the spectral clustering loss coupling strength, and the second channel obtains a behavioral partitioning view based on the clustering power fluctuation mutual information. The game payoff matrix is ​​constructed using the cross-partition line loss interaction entropy change and the intra-partition behavioral homogeneity change, and the dynamic partitioning scheme is evolved. The partitioning boundary constraints are constructed based on the boundary power exchange sensitivity and voltage coupling coefficient.

[0057] The partitioned potential energy field construction and node delimitation module is used to construct the line loss potential energy field within each partition based on the increase in active power loss, the amount of node voltage deviation, and the change in homogeneity of behavior within the partition. The candidate compensation set is obtained by tracing back from the peak value of the potential energy field along the direction of line loss sensitivity. The candidate set is cyclically and slightly adjusted to recalculate the line loss potential energy field, verify the boundary constraints, retain the valid ones and discard the invalid ones, until the target is met or the key nodes and the initial value of the adjustment are output.

[0058] The heterogeneous equipment collaborative allocation and cross-regional coordination module is used to establish a dynamic performance profile for key node adjustable equipment, including response delay, adjustment accuracy and health score. It allocates the regional adjustment amount according to the Shapley marginal contribution of each equipment to the potential energy field decline to obtain the initial allocation scheme; verifies the boundary constraints, and transfers the excess to the second healthiest equipment in the same region or the equipment in the neighboring region, and iteratively obtains the compensation amount given value.

[0059] The timing interleaving and execution module is used to determine the communication traffic conditions and construct a three-segment delay spectrum; taking the device with the largest upper bound of the delay spectrum as the timing anchor point, the health score is mapped to the response action slope, the adjustment accuracy is mapped to the adjustment stability margin, the instruction start time is determined, a timing interleaving compensation sequence is formed and issued for execution.

[0060] The beneficial effects of this application are as follows:

[0061] Compared with existing static partitioning methods, which are prone to uncontrolled boundary power exchange and difficult to adapt to real-time fluctuations in source load, this invention uses a line loss sensitivity matrix and a dual-channel evolutionary game fusion mechanism to dynamically combine physical coupling and behavioral patterns. The change rate of partition boundary assignment after the game is less than 4%, which realizes the adaptive adjustment of the structure following the operating state and strictly constrains the upper limit of boundary power exchange, effectively avoiding the risk of exceeding the limit across zones.

[0062] To address the problem that traditional extensive compensation methods cannot accurately locate the source of damage and have limited damage reduction effects, this invention constructs a line loss potential energy field and traces back along the sensitivity direction. Combined with cyclic boundary verification, it locks the key nodes that contribute the most to the high-loss area. In engineering applications, it reduces the average line loss rate by 2.3 percentage points during peak reverse flow periods, and stabilizes the voltage deviation of key nodes within ±2% of the rated value, thus achieving precise allocation of compensation resources to the source of the problem.

[0063] Overcoming the shortcomings of existing compensation equipment collaborative control, such as lack of precise timing management and the tendency for simultaneous operation of multiple machines to cause shock oscillations, this invention establishes a dynamic performance profile and a three-segment delay spectrum of the equipment. By using a dynamic slope timing sequence with weighted health and adjustment accuracy, the adjustment process of multiple devices is seamlessly connected and peaks are smoothly staggered. The voltage fluctuation amplitude is reduced by 60% compared to the synchronous distribution method. Even under harsh conditions such as heavy communication load, it still maintains stable operation without overshoot, thus improving the stability and robustness of compensation execution. Attached Figure Description

[0064] Figure 1 This is a schematic diagram of the overall dynamic line loss adaptive zoning compensation method for distribution networks;

[0065] Figure 2 S100 flowchart of the dynamic line loss adaptive zoning compensation method for distribution networks;

[0066] Figure 3 S200 flowchart of the dynamic line loss adaptive zoning compensation method for distribution networks;

[0067] Figure 4 S300 flowchart of the dynamic line loss adaptive zoning compensation method for distribution networks;

[0068] Figure 5 Flowchart of the S400 method for dynamic line loss adaptive zoning compensation in distribution networks;

[0069] Figure 6 S500 flowchart of the dynamic line loss adaptive zoning compensation method for distribution networks;

[0070] Figure 7 , Figure 8 , Figure 9 , Figure 10 , Figure 11 and Figure 12The diagram shows the six interfaces of the dynamic line loss adaptive zoning compensation system for distribution networks. Detailed Implementation

[0071] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0072] In the context of new power systems, the random fluctuation characteristics of power distribution networks on both the source and load sides are becoming increasingly significant. This causes active power losses in the network to change in real time with the operating status, making it difficult for traditional strategies relying on fixed partitioning and static compensation to suppress dynamic line losses in a timely and accurate manner. The theoretical basis of this application is established on the theories of network loss sensitivity analysis, time-series mutual information feature mining, evolutionary game theory, potential field reverse tracing theory, and cooperative game Shapley value allocation theory. By constructing a line loss sensitivity matrix, the strength and direction of the impact of power regulation at any node on the losses of each branch in the entire network are quantitatively described. Then, the loss coupling strength between nodes is generated by combining electrical distance. Using a dual-channel competitive fusion structure, the spectral clustering view based on sensitivity coupling relationship and the clustering view based on time-series behavioral mutual information are fused through evolutionary game theory to generate a dynamic partitioning scheme that takes into account both physical structure and operating mode. Within the partition, a line loss potential energy field that integrates active power loss increment, voltage deviation, and behavioral homogeneity changes is constructed, and key compensation nodes are locked by tracing back along the sensitivity direction. The initial value of the regulation amount that satisfies the partition boundary constraints is obtained through a cyclic boundary determination process. At the equipment level, a Shapley value marginal contribution mechanism is introduced to realize the collaborative allocation among heterogeneous compensation equipment. Based on the mapping relationship between the equipment's three-segment delay spectrum and response slope and stability margin, dynamic slope time-series staggered arrangement with regulation accuracy weighting is executed, ultimately achieving adaptive partitioned smooth compensation of dynamic line loss for the entire network.

[0073] The core theoretical formula derivation includes the following five formulas, which respectively support sensitivity analysis, partitioned coupling quantization, potential field construction, equipment health assessment, and execution timing orchestration:

[0074] For a having Each node In a radial distribution network with several branches, when the per-unit voltage is approximately 1.0, the branch... Active power loss at nodes The sensitivity factor for injected active power can be simplified to: ,in branch road The resistance, branch road The current amplitude, branch road Downstream node set, This is an indicator function.

[0075] Using the sensitivity column vectors of each node and Combined with the electrical distance between nodes Define the loss coupling strength between nodes as ,in The distance attenuation factor is the electrical distance. Take the sum of the impedance magnitudes of the branches on the shortest path between the two nodes. For nodes and The degree of synergy between losses. This strength reflects the combined effect of sensitivity directional similarity and electrical distance on the interaction of power regulation.

[0076] To locate key compensation nodes within a zone, a scalar field describing the distribution of line loss is constructed, namely, the line loss potential energy field. ,in For the node Adjusting the increase in intra-zone losses caused by unit power Let L2 be the node voltage deviation. This represents the change in the homogeneity of behavior within the region. These are the weighting coefficients.

[0077] The health scores of multiple adjustable compensation devices are calculated by decaying based on the number of historical actions, and are expressed as follows: ,in The initial health score is perfect. The attenuation coefficient is... For the first The cumulative number of actions performed by the device.

[0078] During the time-series interleaving compensation phase, the health score is mapped to the slope of the device's response action. ,in and To preset the minimum and maximum slopes, This represents the maximum reference value for health. This mapping allows devices with higher health to adjust at a steeper rate, accelerating their transition to a stable state.

[0079] The above five formulas constitute the core mathematical foundation of this invention in terms of sensitivity quantification, partition coupling, potential field construction, equipment dynamic evaluation, and timing control, providing a clear basis for subsequent steps.

[0080] The specific embodiments of the present invention will be described in detail below.

[0081] Example 1

[0082] Please see Figure 1 This application illustrates a dynamic line loss adaptive zoning compensation method for distribution networks provided in an embodiment of this application. The method includes:

[0083] S100: Obtain distribution network topology and operation data, and calculate the line loss sensitivity matrix and the loss coupling strength between nodes.

[0084] S200: By fusing dual-view clustering and evolutionary game theory between the sensitivity structure channel and the behavior pattern channel, a dynamic partitioning scheme is generated and partition boundary constraints are constructed.

[0085] S300: Based on the line loss potential energy field, the key node is located and the cycle is bounded within the execution area to obtain the initial values ​​of the key nodes and compensation amounts.

[0086] S400: The compensation amount for multiple devices is allocated using the Shapley value marginal contribution mechanism, and boundary constraints are met through cross-regional coordination.

[0087] S500: Based on the three-segment delay spectrum and the dynamic performance of the equipment, it performs time-series interleaving and issues and executes compensation commands.

[0088] The specific plan includes:

[0089] In a dynamic adaptive zoning compensation method for line loss in a distribution network, S100 obtains the topology and parameters, calculates real-time changes, constructs a line loss sensitivity matrix, and generates the loss coupling strength between nodes, providing a basic quantitative relationship for subsequent zoning and compensation positioning.

[0090] Please refer to Figure 2 It illustrates a flowchart of an exemplary dynamic line loss adaptive zoning compensation method S100 for a distribution network, the contents of which include:

[0091] S110: Obtain the distribution network topology and line impedance parameters, synchronously collect the time-series data of the injected power at each node, and calculate the real-time changes.

[0092] In this step, a topology file describing the network node and branch connection relationships is obtained from the distribution network energy management system or distribution automation master station, and the resistance and reactance parameters of each branch are extracted. Simultaneously, intelligent terminal devices deployed at each node collect time-series data of injected active and reactive power at fixed time intervals, and calculate the real-time change in injected power between adjacent sampling times. This change reflects the microscopic details of source-load fluctuations over time and forms the basis for subsequent sensitivity calculations and dynamic zoning.

[0093] In one possible implementation of this step, the topological connections are stored in the form of a node-branch association matrix, representing the power supply hierarchy of the radial network. Branch impedance parameters include resistance per kilometer, reactance, and line length, which are converted to lumped parameter resistance and reactance. The sampling period for node injected power is set according to the capability of the distribution automation system to capture second-level or minute-level fluctuations and continuously store cross-sectional data within a certain time period. For the current moment, the real-time change in node injected active power is calculated using a first-order backward differential, i.e., subtracting the value from the previous sampling moment from the current value; the change in reactive power is calculated in the same way. Before entering the calculation module, the data needs to undergo basic quality screening to remove null values ​​and jump values ​​caused by communication interruptions or equipment malfunctions, ensuring that the sequence participating in the calculation is continuous and effective.

[0094] S120: Based on real-time changes and line impedance parameters, calculate the sensitivity factor of the injected power change at each node on the active power loss of each branch in the entire network, and form a line loss sensitivity matrix.

[0095] In this step, considering the current operating state of the distribution network, the sensitivity relationship between the branch losses and the injected power at each node is constructed using line impedance parameters and the injected power at each node. For a radial network, the branch current is determined by the net injected power at the downstream node; therefore, the sensitivity factor is closely related to the topological location of the node. The core of this step lies in using a simplified sensitivity expression to achieve fast and online-updable loss sensitivity calculation.

[0096] In one possible implementation of this step, a forward-backward power flow calculation is first performed to obtain the ground-state values ​​of the current amplitude in each branch. For each branch... Let its resistance be... The amplitude of the ground state current is Based on the definition of loss, neglecting voltage phase angle differences and voltage amplitude variations, the active power loss of the branch can be approximated as... When node When the injected active power changes slightly, if the node Located on a side road In the downstream region, the power change will flow entirely or partially through the branch. This causes a change in the branch current; conversely, if the node... If not in the downstream area, then for the branch road The current has no direct effect. Based on this physical law, the sensitivity factor is expressed as... With a constant voltage of 1.0 pu, the partial derivative... At the node Belongs to downstream set The value is approximately 1 when the sensitivity factor is applied, and 0 otherwise. Therefore, the practical formula for calculating the sensitivity factor is obtained: ;in This is an indicator function.

[0097] By applying the above calculation process to all branches and all nodes, the line loss sensitivity matrix can be formed. The first of the matrix Line number The column element is .

[0098] S130: Column vectors are extracted from the line loss sensitivity matrix to form the line loss sensitivity direction of each node.

[0099] In this step, each column of the line loss sensitivity matrix corresponds to a specific node, and each element in the column vector reflects the intensity of the independent impact of changes in node injected power on the loss of each branch.

[0100] In one possible implementation of this step, for the node Extract the first of the sensitivity matrix. List To eliminate dimensions while retaining directional information, column vectors can be normalized to obtain unit direction vectors. As a node The standard line loss sensitivity direction. In the reverse tracing process of S300, the power source will be traced back based on this direction vector to find the upstream node that contributes the most to the high-loss region.

[0101] S140: Using the line loss sensitivity matrix, calculate the loss coordination degree of any two nodes on the loss of the same branch, and generate the loss coupling strength between nodes by combining the electrical distance between nodes.

[0102] In this step, two factors are introduced: the similarity of sensitivity vectors and electrical distance, to comprehensively quantify the degree of coupling between the two nodes when adjusting power on the overall network loss. The more similar the sensitivity vectors, the more consistent the distribution pattern of the impact of the power adjustment of the two nodes on the losses of each branch; the closer the electrical distance, the stronger the mutual interference of the adjustment behavior at the physical level, and the tighter the coupling relationship.

[0103] In one possible implementation of this step, nodes are defined. With nodes loss coupling strength for: ;in and This is the unnormalized raw sensitivity column vector. For nodes and The degree of synergy between losses. For nodes and The electrical distance between them is taken as the sum of the impedance moduli of the branches along the shortest path, i.e. , Attenuation coefficient Used to control the rate at which electrical distance weakens coupling strength; the value is set according to network size and voltage level.

[0104] Calculate the relationship between all node pairs This matrix constitutes the inter-node loss coupling strength matrix. The inter-node loss coupling strength matrix is ​​a symmetric, non-negative matrix, and the size of its elements reflects the tightness of loss coupling between nodes. This matrix will serve as the edge weights for the first channel spectrum clustering in S200, enabling the partitioning process to group nodes with tight loss coupling into the same partition, thereby ensuring the synergy of adjustment behavior within a partition and reducing the cost of cross-partition interaction.

[0105] In a dynamic line loss adaptive zoning compensation method for distribution networks, S200 takes the line loss sensitivity matrix, sensitivity direction, inter-node loss coupling strength, and node injected power time series data output by S100 as input, starts a dual-channel competitive fusion structure to generate a dynamic zoning scheme, and establishes zoning boundary constraints.

[0106] Please refer to Figure 3 It illustrates a flowchart of an exemplary dynamic line loss adaptive zoning compensation method S200 for a distribution network, the contents of which include:

[0107] S210: First channel—sensitivity structure channel, using the loss coupling strength between nodes as the edge weight, generates a sensitivity partition view through spectral clustering.

[0108] In this step, the distribution network nodes are considered as vertices of the graph, and the inter-node loss coupling strength generated by S140 is used. As a connection vertex and We construct a weighted undirected graph by determining the edge weights. Then, we perform spectral clustering on the weighted undirected graph, dividing the nodes into clusters based on the eigenvectors of the graph's Laplacian matrix, thus forming a sensitivity partitioned view.

[0109] In one possible implementation of this step, the coupling strength matrix is ​​first used. Constructing an adjacency matrix ,in Set the diagonal elements to zero. Calculate the degree matrix. It is a diagonal matrix and The nonnormalized Laplace matrix of the graph is Solving the generalized eigenvalue problem Take the front The eigenvectors corresponding to the smallest non-zero eigenvalues ​​form a matrix, and each row of the matrix is ​​used as a low-dimensional embedding feature for the corresponding node. The dimension of the embedding feature... The number of partitions is automatically determined using the eigenvalue interval heuristic method, i.e., selecting the partitions that make the first partition the most suitable for the second partition. With the The largest difference among the eigenvalues The values ​​were then used to partition these embedded features using the k-means clustering algorithm, resulting in a sensitivity partitioning view. .

[0110] S220: Second channel—behavioral pattern channel, extracts the temporal fluctuation mutual information features of node injected power time series data, and generates a behavioral partition view based on feature similarity clustering.

[0111] In this step, the information interaction relationship between the power fluctuations of each node is mined from the time series data of the injected power of the nodes. The time series fluctuation mutual information feature reflecting the synergy of the behavioral patterns between the nodes is extracted, and then the feature is clustered for similarity to obtain a partitioned view based on the running behavior.

[0112] In one possible implementation of this step, for each node, the nearest [node name] is taken. For each sampling point, a sequence of active power changes is constructed, and a sequence of reactive power changes is similarly constructed. To calculate the mutual information of time-series fluctuations between two nodes, the power change sequence is first discretized into several intervals with equal width, and the joint probability distribution and marginal probability distribution are statistically analyzed. Mutual Information Based on the definition of information theory, a larger value indicates a stronger statistical interdependence between the power fluctuations of the two nodes, meaning their behavioral patterns are more similar. This yields a symmetric mutual information matrix. Treating the mutual information matrix as a similarity matrix, a spectral clustering algorithm similar to S210 is applied to aggregate the nodes into several clusters, resulting in a behavioral partitioning view. .

[0113] S230: The data is fed into the evolutionary game fusion layer in parallel, a game payoff matrix is ​​constructed, and the boundary node ownership is determined through finite-round evolutionary game, outputting a dynamic partitioning scheme.

[0114] In this step, the sensitivity partition view is... and behavior partition view This is considered as a proposal for a partitioning scheme by two participating parties. For all boundary nodes in the two views, the changes in cross-partition line loss cross-entropy and intra-partition behavior homogeneity caused by assigning them to different adjacent partitions are calculated. This is used to construct a game payoff matrix, and a finite-round evolutionary game is executed until the boundary assignment reaches a stable equilibrium, resulting in a final dynamic partition that takes into account the advantages of both views.

[0115] In one possible implementation of this step, the inconsistent set of boundary nodes in the two views is first identified. For any boundary node... and its adjacent partitions that may be assigned and Calculate the changes in global indicators caused by the two allocation schemes. Cross-partition line loss cross-entropy. Defined on the distribution of inter-regional exchange losses, it is used to quantify the interaction uncertainty caused by partitioning, and its variation. Reflects the nodes from Move to The degree of concentration or dispersion of post-inter-regional interaction losses. Homogeneity of behavior within the region. Defined as the average mutual information of all nodes within a partition regarding temporal fluctuations, it reflects the consistency of behavioral patterns within the partition, and its variation... This measures the extent to which the homogeneity of partition behavior is improved or reduced after the move.

[0116] The two changes mentioned above are weighted and summed according to preset preference weights to construct the payoff function for the boundary nodes. The game process employs a synchronous update strategy: each round, all boundary nodes calculate the payoff increment they could gain by changing their affiliation. When the payoff increment is positive and exceeds a preset threshold, the node moves its affiliation based on probability. This process is repeated for multiple rounds, with all boundary nodes continuously adjusting their positions through mutual influence in the evolutionary process, until no node has any further intention to move or the maximum number of evolutionary rounds is reached. At this point, the partition boundary reaches a stable equilibrium, and the final dynamic partitioning scheme is output. .

[0117] S240: At the boundary nodes of the final dynamic partition, extract the power exchange sensitivity and voltage coupling coefficient between adjacent partitions to construct partition boundary constraints.

[0118] In this step, to ensure that subsequent compensation control does not violate the safe operating limits between partitions, for each boundary node of the final partition scheme, the allowable upper limit of power exchange and the voltage deviation correlation limit are quantified using sensitivity information and voltage coupling relationship. These constraints will serve as hard boundaries for S300 and S400 compensation regulation.

[0119] In one possible implementation of this step, for boundary nodes Let it belong to a partition. Adjacent partitions are Extracting power exchange sensitivity For nodes The sensitivity factor of active power injection changes to power flow on the zonal tie line is obtained by summing the elements of the corresponding tie branch in the S120 line loss sensitivity matrix. Based on the thermal stability limit and current operating margin of the tie line, the maximum allowable power regulation at the boundary node is calculated, forming the upper limit of power exchange. The voltage coupling coefficient reflects the node... The impact of voltage changes on the voltages of adjacent partition boundary nodes is extracted using a voltage sensitivity matrix. Based on the allowable range of voltage deviation between adjacent partitions, voltage deviation correlation limits are derived.

[0120] In a dynamic adaptive zoning compensation method for line loss in a distribution network, the S300 receives the dynamic zoning scheme, zoning boundary constraints, and changes in the homogeneity of behavior within the zone. Combined with the direction of line loss sensitivity, a line loss potential energy field is constructed within each zone. Key compensation nodes and their initial adjustment values ​​are determined through reverse tracing and cyclic delimitation.

[0121] Please refer to Figure 4 It illustrates a flowchart of an exemplary dynamic line loss adaptive zoning compensation method S300 for a distribution network, the contents of which include:

[0122] S310: Within each zone, a line loss potential energy field is constructed using the active power loss increment, node voltage deviation, and intra-zone behavior homogeneity change as inputs.

[0123] In this step, the partition is treated as a whole, and a scalar potential energy value is defined at each node within the partition. This potential energy comprehensively characterizes the contribution of the node and its surrounding area to line loss, the degree of voltage quality degradation, and the degree of disruption to behavioral homogeneity, forming a scalar field describing the severity and distribution of line loss. The high-value areas of the potential energy field are the diseased areas that need priority compensation.

[0124] In one possible implementation of this step, for partitioning Any node within Calculate the potential energy of line loss: In the formula, For the node When a unit power injection is added, the increase in total loss within the partition is obtained by summing the components of the corresponding column of the node in the line loss sensitivity matrix on the branch within the partition, i.e. , For partitioning A set of branches within the network. For nodes The L2 norm of the vector deviating between the current voltage and the rated voltage reflects the local voltage quality. For preliminary assessment at the node After power adjustment, the expected change in the homogeneity of behavior within the region is obtained through perturbation simulation using the mutual information model of S220. Weighting coefficients This is a preset constant, set according to the relative importance of loss reduction requirements, voltage requirements, and behavioral stability requirements. By traversing all nodes within a partition, the line loss potential energy field distribution of the partition can be constructed. The fluctuations of this scalar field indicate areas of dense loss and weak voltage.

[0125] S320: Starting from the local peak point of the line loss potential energy field, perform reverse tracing along the line loss sensitivity direction provided by S100 to lock the power supply path that contributes the most to the local peak and the source node of the power supply path, forming a set of candidate compensation nodes.

[0126] In this step, local peak nodes in the line loss potential energy field that are higher than the neighborhood mean are identified. These peak nodes are regarded as the convergence points of the concentrated line loss area. By tracing the upstream path of power flow back in the direction of line loss sensitivity, the root node that contributes the most to the peak loss is found and added to the candidate compensation node set, so that the compensation resources can be accurately deployed to the source of the loss.

[0127] In one possible implementation of this step, a set of local peak points is first identified by comparing the potential energy values ​​of each node with its electrically adjacent nodes. For each peak point, its normalized line loss sensitivity direction vector is extracted. In a radial topology, power flow has a clear directionality; reverse tracing means moving along the power source direction, i.e., selecting the upstream adjacent node with the smallest angle to the sensitivity direction vector as the next node in the tracing path. Starting from the peak point, the process continuously moves towards its parent node, calculating the consistency between the sensitivity direction of the candidate parent node and the negative gradient direction or the upstream desired direction at each step, prioritizing the upstream branch with the best matching influence mode and lower path impedance. The tracing process continues until the power source node, such as the substation bus or distributed generation grid connection point, is reached. This source node is recorded as a candidate compensation node. A peak point may be traced to multiple source paths; all traced source nodes are merged and deduplicated to form a set of candidate compensation nodes. .

[0128] S330: Initiate the cyclic boundary determination process, sequentially select compensation nodes from the candidate compensation node set, apply a small power adjustment, recalculate the line loss potential energy field and verify the partition boundary constraints, and iteratively output the key nodes and the initial values ​​of the adjustment amount.

[0129] In this step, based on the candidate compensation node set, the effect of compensation is tested node by node, and boundary verification is performed using partition boundary constraints to ensure that compensation adjustment does not cause cross-regional exceedances. Finally, the effective compensation nodes and their optimal initial adjustment values ​​are determined. The cyclic boundary determination process considers both loss reduction effect and boundary safety in a unified manner.

[0130] In one possible implementation of this step, the cycle delimitation process specifically includes:

[0131] S331: Initialize the selected compensation node set to an empty set, record the current potential energy field peak value, and initialize the remaining candidate compensation set to all candidate compensation nodes.

[0132] S332: Select a node from the residual candidate compensation set. The selection strategy is based on the consistency between the sensitivity direction and the negative gradient of the potential energy at that node, prioritizing the node that can reduce the peak value to the greatest extent. The selection criterion can be quantified as the negative value of the inner product of the node sensitivity direction vector and the local gradient vector of the potential energy field. The larger this value, the more effectively the adjustment at that node is expected to suppress the peak value.

[0133] S333: Apply a small amount of active power regulation to the selected node. The magnitude of the regulation is determined proportionally to the current peak potential energy and sensitivity level to ensure that the test does not cause large disturbances. Based on this regulation, perform a fast power flow calculation again or estimate using the sensitivity matrix to update the node voltages and branch currents of the entire partition, and recalculate the line loss potential energy field to obtain a new peak potential energy.

[0134] S334: Check whether the power exchange amount of all boundary nodes exceeds the power exchange upper limit given in S240, and whether the voltage deviation limit of adjacent partitions is exceeded. The verification is based on rapid extrapolation of power exchange sensitivity. If any boundary node exceeds the limit, it is determined that this attempt violates the boundary constraints, the adjustment amount is immediately rolled back, and the selected node is permanently removed from the residual candidate compensation set, returning to S332 to select the next node.

[0135] S335: If all boundary constraints of the partition are satisfied, and the new potential energy peak value is less than the original peak value, it indicates that the compensation node is effective. Add it to the set of selected compensation nodes, update the current potential energy field and peak value to the new value, accumulate and record the adjustment amount of the node, and remove the node from the set of residual candidate compensation nodes. If the new peak value does not decrease, it indicates that the adjustment of this node is ineffective or has a negligible effect in the current state, and it is also removed from the set of residual candidate compensation nodes.

[0136] S336: Repeat S332 to S335 until the current potential energy peak value drops below the preset threshold, or the remaining candidate compensation set is empty. The preset threshold can be set according to the proportion of the initial peak value to retain a certain margin for the reduction target. After the cyclic boundary is completed, the nodes in the selected compensation node set are the key nodes that need to be compensated in the partition, and the accumulated adjustment amount is the initial value of the adjustment amount of each key node. This information, along with the current potential energy field state, will be transmitted to S400 for device-level allocation.

[0137] In a dynamic line loss adaptive zonal compensation method for distribution networks, the S400 receives key nodes and initial values ​​of adjustment quantities, establishes a dynamic performance profile of the equipment, utilizes the Shapley value marginal contribution mechanism to achieve multi-equipment collaborative allocation, embeds a cross-regional coordination link, and outputs the compensation quantity setpoint of each adjustable compensation equipment.

[0138] Please refer to Figure 5 It illustrates a flowchart of an exemplary dynamic line loss adaptive zoning compensation method S400 for a distribution network, the contents of which include:

[0139] S410: Collects real-time operating status data for adjustable compensation devices associated with each key node, and establishes a dynamic performance profile, including response latency, adjustment accuracy, and health score.

[0140] In this step, adjustable compensation devices include static var generators, capacitor banks, on-load tap-changing transformers, and distributed energy storage. For each device, its nameplate parameters, real-time operating status, and historical action data are collected to construct a dynamic performance profile, so that the device's real-time capabilities and health level are considered when allocating adjustment amounts.

[0141] In one possible implementation of this step, the response delay is the average delay time from when the device receives the instruction to when the output changes. This delay is obtained by statistically analyzing the difference between the instruction issuance time and the output feedback confirmation time in the historical action record. Adjustment accuracy The accuracy is characterized by the ratio of the equipment's minimum adjustment step size or control resolution to its rated capacity; the higher the accuracy, the better. The smaller the value, the lower the health score. The health score is calculated by decaying the historical cumulative number of actions, using the following formula: ;in The initial health score is perfect. The attenuation coefficient is... For the first The cumulative number of actions performed by each device. For each device, its performance profile is stored in the form of triples and continuously updated, providing personalized parameters for subsequent Shapley value allocation and timing orchestration.

[0142] S420: The Shapley value marginal contribution mechanism is adopted to allocate the total adjustment amount of the zone to each adjustable compensation device to form an initial allocation scheme.

[0143] In this step, the total adjustment required for key nodes within a partition is considered in relation to the associated equipment set. The alliance characteristic function is defined as the decrease in the peak value of the potential energy field of the partition line loss after equipment combination compensation. By calculating the marginal contribution of each device to various possible alliances, the Shapley value is obtained, which serves as the basis for the proportional allocation of adjustment, ensuring that the allocation scheme fairly reflects the independent contribution of each device.

[0144] In one possible implementation of this step, a set of devices is set. For any subset Define its value For only sets The reduction in peak potential energy when the equipment participates in compensation according to its capacity. Shapley value This is the average of the marginal contributions across all possible alliance sequences. Since the computational cost of directly calculating all permutations and combinations increases exponentially with the number of devices, a Monte Carlo random sampling method can be used to generate a sufficient number of random device permutations. The marginal value increment after adding the current device is calculated sequentially along each sequence, and the average value is taken as an estimate of the Shapley value. The Shapley values ​​of all devices are normalized to obtain the allocation weights. The initial allocation for each device is the total adjustment multiplied by its corresponding weight, but does not exceed the device's capacity limit. If a device's allocation exceeds the limit, the excess is redistributed to other devices according to their weights.

[0145] S430: Embedded partition boundary constraint verification and cross-region coordination, iterative adjustment until all boundary constraints are satisfied, and output the given value of compensation amount.

[0146] In this step, the initial allocation scheme obtained from S420 is checked to see if the power exchange of each boundary node exceeds the limit. If the power exchange of some boundary nodes exceeds the allowable limit due to equipment allocation, the cross-regional coordination mechanism is triggered to reduce or transfer the allocation amount of the relevant equipment until all boundary constraints are satisfied, and the final compensation amount is output.

[0147] In one possible implementation of this step, the verification process is as follows: Based on the allocation amount and power exchange sensitivity of each device, the power exchange change caused by compensation at each boundary node is deduced. If there is a boundary node whose change exceeds a preset upper limit, the device that causes the largest over-limit contribution is first located. The initial allocation amount of this device is reduced, with the reduction calculated to just eliminate the over-limit. The reduced portion is preferentially transferred to the device with the second highest health score in the same partition and still has capacity; if the device in the same partition has no capacity, support is requested from adjacent partitions through the boundary interaction node, and the device with a higher health score and sufficient capacity in the adjacent partition takes on part of the adjustment amount. After adjustment, a second verification is performed. If there is still an over-limit, the above process is repeated, iteratively adjusting until all boundary node constraints are satisfied. The final compensation amount given value output to each device is the safe and feasible instruction value, and it is transmitted to S500 for timing execution.

[0148] In a dynamic line loss adaptive partition compensation method for distribution networks, the S500 constructs a three-segment delay spectrum based on the dynamic performance profile of the equipment and the current control network communication traffic conditions, performs time-series interleaving, and sends out the compensation sequence for execution.

[0149] Please refer to Figure 6 It illustrates a flowchart of an exemplary dynamic line loss adaptive zoning compensation method S500 for a distribution network, the contents of which include:

[0150] S510: Real-time monitoring and control of network communication traffic, determining whether the current operating condition of the network communication traffic is light load, normal, or heavy load.

[0151] In this step, the power distribution communication network management system obtains traffic characteristics such as average bandwidth utilization, packet loss rate, and latency of the current backbone network. Based on preset thresholds, the communication conditions are divided into three categories: light load, normal, and heavy load. This classification directly determines the selection of subsequent device delay spectra, enabling timing orchestration to match the real-time state of the communication environment.

[0152] In one possible implementation of this step, a light-load condition is defined as bandwidth utilization below a certain low threshold and average round-trip latency below a certain minimum value; a normal condition is utilization at a moderate level and latency not exceeding a certain medium limit; and a heavy-load condition is utilization above a certain high threshold or latency exceeding a certain maximum value. Statistics provided by the network management interface are periodically read, filtered by a moving average, and compared with the thresholds to output a current condition label for use by the S520.

[0153] S520: Utilize response delay to construct a three-segment delay spectrum corresponding to the current operating condition for each adjustable compensation device, and describe the delay distribution with probability intervals.

[0154] In this step, for each device, three probability distribution models are established based on its historical response delay data under different communication conditions, corresponding to light load, normal, and heavy load conditions, respectively, forming a three-segment delay spectrum.

[0155] In one possible implementation of this step, for the device In working conditions Next, historical response delay samples under this operating condition are collected. These samples are then fitted using kernel density estimation or by assuming a normal distribution to obtain a probability density function. The 95th quantile is taken as the upper bound of the delay under this operating condition. Thus, the three-segment delay spectrum of each device can be characterized by the upper bound values ​​under the three operating conditions, forming a triplet. This delay spectrum will be used for anchor point selection and timing interleaving arrangement of S530 and S540.

[0156] S530: Using the adjustable compensation device with the largest upper bound in the three-segment delay spectrum as the timing anchor point, determine the command start time of the anchor point.

[0157] In this step, the upper limit of the delay of all devices participating in this compensation under the current operating conditions is compared, and the device corresponding to the maximum value is selected as the timing anchor point. The timing anchor point is the time base of the entire interleaved compensation sequence, and its instruction start time is set to the current time plus a preset fixed lead time.

[0158] In one possible implementation of this step, after the anchor device is determined, its command initiation time is recorded as the baseline zero time or an absolute timestamp. Since this device has the largest upper bound on communication latency, placing it at the beginning of the sequence can avoid subsequent devices waiting too long due to communication congestion, thus reducing the uncertainty of the overall completion time.

[0159] S540: Performs dynamic slope timing interleaving with adjustment precision weighting to generate the start time of each device instruction.

[0160] In this step, for devices other than the anchor point, the health score is used to map the slope of the response action, and the adjustment accuracy is used to map the adjustment stability margin. The moment when the adjustment transition process of the adjacent device enters the stable range is used as the instruction start time of the next device, so as to achieve the connection between the beginning and end of the timing and smooth peak shifting.

[0161] In one possible implementation of this step, the health score of the device is first determined. Mapped to response action slope A linear mapping relationship is used: ;in and The preset minimum and maximum allowable slopes, This represents the maximum reference value for health. A higher health level indicates a steeper response slope, meaning a more abrupt transition from receiving a command to reaching the target compensation level, and a shorter completion time. Simultaneously, it increases the device's adjustment accuracy. Mapping to adjust stability margin The higher the adjustment precision, the greater the stability margin. The smaller the value, the shorter the confirmation time required for the device to determine that it has entered a stable state after reaching the target value. This mapping is implemented using linear or piecewise functions.

[0162] For anchor point devices, the time when they enter the stable region is equal to the startup time plus the adjustment amount divided by the response slope, plus the stability margin. For subsequent device sequences arranged in descending order of upper bound delay, the previous device... Once the stable interval time is calculated, it is directly used as the next device. The command start time is determined. The start time sequence for all devices is generated sequentially. Because each device has a different slope and stability margin, the command start times are naturally staggered, preventing multiple devices from entering the transient process simultaneously and effectively reducing superimposed impacts.

[0163] S550: Directly distributes the time-series interleaved compensation sequence to each adjustable compensation device for execution, completing dynamic line loss adaptive partition compensation.

[0164] In this step, the timing-interleaved compensation sequence, including device identifier, compensation setpoint, and precise command start time, is encoded into a control command message. This message is distributed to the corresponding device via the communication network at each specified start time. The device adjusts its output according to its set slope and automatically maintains stability after reaching the setpoint. Once the entire sequence is completed, a complete dynamic partitioned adaptive compensation cycle ends. The operating status is continuously monitored. When changes in the line loss potential field or node injected power trigger a new round of compensation conditions, the above S100 to S500 processes are repeated to form adaptive control.

[0165] Example 2

[0166] This embodiment was verified in a practical engineering project on a 10kV line in a city's power distribution network. The line has a rated voltage of 10kV, includes 94 nodes and 15 segmented branches, and connects to 23 distributed photovoltaic power generation systems and 11 electric vehicle charging stations. It exhibits significant fluctuations in both the source and load sides. Intelligent distribution terminals were deployed at all nodes along the line to collect injected power time-series data at 1-second intervals. The data was processed by an edge computing platform equipped with a dual-core CPU and a GPU accelerator card. A total of 17 adjustable devices participated in the compensation, including 8 static var generators, 3 on-load tap-changing transformers, and 6 distributed energy storage systems. Relevant parameters were set as health degradation coefficients. Response action slope range pu / s, pu / s, adjusting the stability margin range ms The baseline scheme for comparison is a fixed-zone compensation method, which divides the entire network into three fixed zones according to administrative regions. The total compensation amount is statically allocated according to the equipment capacity ratio, and all equipment adjusts simultaneously upon receiving instructions. The experiment ran continuously for 30 days, and data was collected and analyzed from four dimensions: loss reduction effect, equipment action balance, timing control quality, and long-term stability.

[0167] Regarding the loss reduction effect, typical 24-hour operating data were selected, and the average line loss rate and voltage qualification rate were statistically analyzed hourly. Table 1 summarizes the daily average line loss rate and voltage qualification rate under three scenarios: no compensation, fixed zone compensation, and the method of this invention. The daily average line loss rate was 3.82% under no compensation, decreased to 2.91% under fixed zone compensation, and further decreased to 2.15% under this method. The loss reduction was 43.7% compared to the no-compensation scenario, and 26.1% higher than the fixed zone compensation scenario; the voltage qualification rate increased from 97.4% to 99.3%.

[0168] Table 1 Comparison of daily average line loss rate and voltage qualification rate for different compensation methods

[0169] No compensation 3.82 97.4 — Fixed partition compensation 2.91 98.2 23.80% Method of the present invention 2.15 99.3 43.70%

[0170] Regarding the balance of equipment operation, Table 2 summarizes the coefficient of variation of the total number of operations, the average decay value of the health score, and the number of boundary power exchange overruns for 17 devices over 30 days. The fixed zoning method, due to its static allocation based on capacity ratio and disregard for equipment health status, resulted in a coefficient of variation of 0.42 for the number of operations, a maximum of 256 operations per device, an average health score decay of 7.8 points, and 12 boundary overrun events. This method, through Shapley value allocation and staggered timing, reduced the coefficient of variation of the number of operations to 0.18, with a maximum of only 118 operations per device, an average health score decay of 2.3 points, and no boundary overruns throughout the entire period.

[0171] Table 2. Statistics on Equipment Motion Balance and Boundary Limit Exceedance (30 days)

[0172] Fixed partition compensation 0.42 7.8 12 256 Method of the present invention 0.18 2.3 0 118

[0173] Regarding the effectiveness of timing interleaving control, the performance under heavy-load communication conditions was analyzed in detail, with network bandwidth utilization at 82% and average latency at 63ms. Table 3 shows a comparison of voltage quality at key nodes in the simultaneous transmission mode and the timing interleaving mode of this method. In the simultaneous transmission mode, the transient processes of multiple devices are superimposed, resulting in a maximum voltage deviation of 0.038 pu, an overshoot of 12.5%, and a voltage recovery time of 4.2 seconds. This method, through delay spectrum matching and slope margin arrangement, controls the maximum voltage deviation to 0.015 pu, with an overshoot of only 2.1%, and shortens the voltage recovery time to 1.9 seconds. Although the total adjustment time increases by 1.7 seconds due to interleaving, it significantly reduces the impact amplitude.

[0174] Table 3 Comparison of Voltage Quality under Heavy-Load Communication Conditions

[0175] Maximum voltage deviation (pu) 0.038 0.015 Voltage recovery time (s) 4.2 1.9 Overshoot (%) 12.5 2.1 Total time taken to complete adjustment (s) 3.1 4.8

[0176] Regarding long-term operational stability, Table 4 records the daily average line loss rate data for seven consecutive days. The line loss rate of the uncompensated method fluctuated between 3.76% and 3.92%, while that of the fixed-zone compensation method fluctuated between 2.85% and 2.96%. In contrast, the line loss rate of this method remained within a narrow range of 2.08% to 2.20%, exhibiting the smallest fluctuation amplitude and the lowest absolute value.

[0177] Table 4. Statistics of daily average line loss rate for each method over 7 consecutive days (%)

[0178] Day 1 3.78 2.89 2.12 Day 2 3.85 2.93 2.17 Day 3 3.8 2.88 2.1 Day 4 3.92 2.96 2.2 Day 5 3.76 2.85 2.08 Day 6 3.83 2.91 2.15 Day 7 3.79 2.87 2.11

[0179] Based on the above experimental analysis, this method can effectively reduce dynamic line losses, balance the operating load of each compensation device, significantly suppress transient voltage surges in distribution networks with strong fluctuations, and has excellent boundary safety control capabilities and long-term operational stability, showing significant advantages over fixed-zone compensation schemes.

[0180] Example 3

[0181] A dynamic line loss adaptive zoning compensation system for distribution networks, the system has 6 core functional pages, such as... Figure 7 , Figure 8 , Figure 9 , Figure 10 , Figure 11 and Figure 12 As shown, the system includes an overview interface, a sensitivity matrix interface, a dynamic partitioning interface, a potential energy field tracing interface, a Shapley assignment interface, and a timing orchestration interface. The system includes:

[0182] The sensitivity analysis and coupling quantization module is used to form a line loss sensitivity matrix from the changes in node injected power and line impedance, with the column vectors representing the line loss sensitivity direction, and to calculate the loss coordination degree between nodes, and to obtain the loss coupling strength between nodes by combining the electrical distance.

[0183] The dual-channel dynamic partitioning and constraint construction module is used to initiate a dual-channel competitive fusion structure for dynamic partitioning. The first channel obtains a sensitivity partitioning view based on the spectral clustering loss coupling strength, and the second channel obtains a behavioral partitioning view based on the clustering power fluctuation mutual information. The game payoff matrix is ​​constructed using the cross-partition line loss interaction entropy change and the intra-partition behavioral homogeneity change, and the dynamic partitioning scheme is evolved. The partitioning boundary constraints are constructed based on the boundary power exchange sensitivity and voltage coupling coefficient.

[0184] The partitioned potential energy field construction and node delimitation module is used to construct the line loss potential energy field within each partition based on the increase in active power loss, the amount of node voltage deviation, and the change in homogeneity of behavior within the partition. The candidate compensation set is obtained by tracing back from the peak value of the potential energy field along the direction of line loss sensitivity. The candidate set is cyclically and slightly adjusted to recalculate the line loss potential energy field, verify the boundary constraints, retain the valid ones and discard the invalid ones, until the target is met or the key nodes and the initial value of the adjustment are output.

[0185] The heterogeneous equipment collaborative allocation and cross-regional coordination module is used to establish a dynamic performance profile for key node adjustable equipment, including response delay, adjustment accuracy and health score. It allocates the regional adjustment amount according to the Shapley marginal contribution of each equipment to the potential energy field decline to obtain the initial allocation scheme; verifies the boundary constraints, and transfers the excess to the second healthiest equipment in the same region or the equipment in the neighboring region, and iteratively obtains the compensation amount given value.

[0186] The timing interleaving and execution module is used to determine the communication traffic conditions and construct a three-segment delay spectrum; taking the device with the largest upper bound of the delay spectrum as the timing anchor point, the health score is mapped to the response action slope, the adjustment accuracy is mapped to the adjustment stability margin, the instruction start time is determined, a timing interleaving compensation sequence is formed and issued for execution.

[0187] Those skilled in the art will understand that the embodiments of this application are provided as methods, systems, or computer program products. Therefore, this application takes the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application takes the form of a computer program product implemented on one or more computer storage media containing computer program code. The solutions in the embodiments of this application can be implemented using various computer languages, exemplarily the object-oriented programming language Java and the interpreted scripting language JavaScript, etc.

[0188] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0189] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A dynamic line loss adaptive zoning compensation method for distribution networks, characterized in that, include: The line loss sensitivity matrix is ​​formed by the change in nodal injected power and the line impedance, and the column vectors represent the direction of the line loss sensitivity. Calculate the loss coordination degree between nodes, and combine it with the electrical distance to obtain the loss coupling strength between nodes; A dual-channel competitive fusion structure is initiated for dynamic partitioning. The first channel obtains a sensitivity partitioning view based on the spectral clustering loss coupling strength, while the second channel obtains a behavioral partitioning view based on the clustering power fluctuation mutual information. A game payoff matrix is ​​constructed using the cross-partition line loss interaction entropy change and the intra-partition behavioral homogeneity change, and a dynamic partitioning scheme is derived. Partition boundary constraints are constructed based on the boundary power exchange sensitivity and voltage coupling coefficient. Within each partition, a line loss potential energy field is constructed using the increase in active power loss, the node voltage deviation, and the change in homogeneity of behavior within the partition. A candidate compensation set is obtained by tracing back from the peak value of the potential energy field along the direction of line loss sensitivity. The line loss potential energy field is recalculated by iteratively adjusting the candidate compensation set, and the boundary constraints are verified. The valid ones are retained and the invalid ones are discarded until the target is met or the key nodes and the initial value of the adjustment are output. A dynamic performance profile including response delay, adjustment accuracy, and health score is established for the adjustable equipment at key nodes. The adjustment amount is allocated to each zone according to the Shapley marginal contribution of each equipment to the potential energy field decline to obtain the initial allocation scheme. The boundary constraints are verified, and if the limit is exceeded, the adjustment is transferred to the second healthiest equipment in the same zone or the equipment in the neighboring zone. The compensation amount is obtained by iteratively satisfying the given value. The communication traffic conditions are identified to construct a three-segment delay spectrum; the device with the largest upper bound of the delay spectrum is used as the timing anchor point, the health score is mapped to the response action slope, the adjustment accuracy is mapped to the adjustment stability margin, the command start time is determined, a timing interleaved compensation sequence is formed and issued for execution.

2. The adaptive zoning compensation method for dynamic line loss in distribution networks according to claim 1, characterized in that, The steps for forming the line loss sensitivity matrix include: For those with Each node For a radial distribution network with branches, forward-backward power flow calculations are performed to obtain the current amplitude of each branch. and its resistance With current Define nodes Changes in injected active power affect the branch The sensitivity factor of active power loss is ,in branch road The set of downstream nodes, For indicator functions, when node belong If the value is 1, then 0 is used; Traverse all branches and nodes to construct a line loss sensitivity matrix And extract the first element of the matrix. column vector As a node The line loss sensitivity direction is obtained by normalizing the column vector to obtain the unit direction vector. It is used to characterize the spatial distribution of the impact of node power regulation on the branch loss of the entire network.

3. The adaptive zoning compensation method for dynamic line loss in distribution networks according to claim 2, characterized in that, The steps for obtaining the inter-node loss coupling strength include: For nodes With nodes Utilizing its line loss sensitivity column vector and Define loss coupling strength ,in The electrical distance is represented by the sum of the impedance moduli of all branches on the shortest path between two nodes. The preset electrical distance attenuation coefficient, yes The transpose of the column vector is converted to 1× 3D row vectors For nodes and Degree of synergy in losses between them; The loss coupling strength reflects the combined effect of the cosine similarity of the sensitivity directions of the two nodes and the electrical distance on the regulation interaction, forming the edge weight matrix for spectral clustering.

4. The adaptive zoning compensation method for dynamic line loss in distribution networks according to claim 1, characterized in that, The first channel of the dual-channel competitive fusion structure includes: Construct a weighted undirected graph using loss coupling strength as edge weights, solve the eigenvalue problem of the graph Laplacian matrix, and take the first edge weight. The node embedding features are formed by the eigenvectors corresponding to the smallest non-zero eigenvalues, and the number of partitions is automatically determined using the eigenvalue intervals. Perform on embedded features - Mean clustering yields a sensitivity partition view; The second channel of the dual-channel competitive fusion structure includes: The time-series fluctuation sequences of active and reactive power injected into each node are discretized with equal width. The mutual information value of power fluctuation between any two nodes is calculated, and the mutual information matrix is ​​formed as a similarity matrix. Clustering is then applied to generate a behavioral partition view.

5. The adaptive zoning compensation method for dynamic line loss in distribution networks according to claim 4, characterized in that, The steps for constructing the game payoff matrix and evolving the dynamic partitioning scheme include: Identify the set of boundary nodes that are inconsistent between the two views, for boundary nodes. From the current partition Assigned to adjacent partitions The corresponding cross-regional line loss cross-entropy change Changes in homogeneity of behavior within the region Perform weighted summation to construct the game payoff function; A finite-round synchronous update strategy is adopted, in which each boundary node changes its partitioning according to the payment increment with probability, and the evolution is repeated until the boundary assignment is stable, outputting the optimal dynamic partitioning scheme that integrates structural coupling and behavioral patterns.

6. The adaptive zoning compensation method for dynamic line loss in distribution networks according to claim 5, characterized in that, The partition boundary constraint construction steps include: Extract the power exchange sensitivity of each boundary node to the associated tie branch, and determine the upper limit of the allowable power exchange of the boundary node by combining the thermal stability limit and operating margin of the tie branch. Extract the voltage sensitivity of the boundary nodes to the adjacent partition nodes, determine the voltage deviation correlation limit based on the allowable voltage deviation range of the adjacent partitions, and form a partition boundary constraint set that includes the upper limit of power exchange and the voltage deviation correlation limit.

7. The adaptive zoning compensation method for dynamic line loss in distribution networks according to claim 1, characterized in that, The steps for constructing the line loss potential energy field include: For any node within the partition Construction line loss potential energy : ,in For the node The increase in total loss within the partition caused by increasing unit power injection. For partitioning The set of included branches; For nodes The L2 norm of voltage deviation; For nodes The estimated change in the homogeneity of behavior within the adjusted region; Preset weights; Traversing all nodes in the partition forms a scalar field describing the severity and distribution of line loss, with the potential energy peak region corresponding to the priority compensation region.

8. The adaptive zoning compensation method for dynamic line loss in distribution networks according to claim 7, characterized in that, The steps for obtaining the candidate compensation set include: Identify the set of local peak nodes in the potential energy field that are above the neighborhood mean; For each peak node, backtrack along its normalized line loss sensitivity direction to the power source direction, selecting the upstream adjacent node with the smallest angle with the negative gradient direction hop by hop until the source node such as the substation bus or distributed power grid connection point is reached. During the tracing process, the power supply path with the best matching sensitivity direction is selected. After deduplicating the source nodes traced back, a set of candidate compensation nodes is formed, ensuring that the candidate nodes are located at the beginning of the power supply path that contributes the most to the power of the high-loss area.

9. The adaptive zoning compensation method for dynamic line loss in distribution networks according to claim 8, characterized in that, The steps for processing the candidate compensation set and outputting key nodes and initial values ​​of adjustment include: Initialize the selected compensation node set to empty, iteratively select candidate nodes that are most consistent with the negative potential gradient, apply a small amount of active power adjustment, and use the sensitivity matrix to estimate or recalculate the new potential field using fast power flow. If the new potential energy peak decreases and the power exchange and voltage deviation of all boundary nodes do not exceed the partition boundary constraints, then the node is retained and the potential energy field is updated, and its adjustment is accumulated; otherwise, the adjustment is rolled back and the corresponding node is permanently removed from the candidate compensation set. The loop continues until the potential energy peak drops to the preset threshold or the candidate compensation set is traversed. The retained nodes are then used as the key nodes that need to be compensated in the partition, and the accumulated adjustment amount is the initial value of the adjustment amount.

10. The adaptive zoning compensation method for dynamic line loss in distribution networks according to claim 1, characterized in that, The steps for establishing the dynamic performance profile include: Collect the command response delay, minimum adjustment step size, or resolution of each adjustable compensation device as the adjustment accuracy, and count the historical cumulative number of actions; Define health score ,in The initial health score is perfect. The attenuation coefficient is... For the first The cumulative number of actions performed by the device.

11. The adaptive zoning compensation method for dynamic line loss in distribution networks according to claim 10, characterized in that, The steps for forming the initial allocation scheme and obtaining the given value of the compensation amount include: The decrease in the peak potential field after the equipment combination is put into use is used as the characteristic function of the alliance. The Shapley value of each equipment is calculated by Monte Carlo sampling. After normalization, the adjustment amount allocation weight is obtained. The total adjustment amount of the partition is allocated according to the weight to form the initial scheme, which is subject to the upper limit of equipment capacity. The power exchange change of each boundary node is verified based on the power exchange sensitivity. If the limit is exceeded, the device that contributes the most is located and its allocation is reduced. The reduced amount is transferred to the device with the second highest health score in the same partition, or it is borne by the device in the adjacent partition through the boundary interaction node. The adjustment is iterated until all partition boundary constraints are satisfied, and the compensation amount of each device is output.

12. The adaptive zoning compensation method for dynamic line loss in distribution networks according to claim 1, characterized in that, The steps for constructing the three-segment delay spectrum and determining the time-series anchor points include: Monitor and control network communication traffic, and classify operating conditions into light load, normal, and heavy load based on bandwidth utilization and average latency; For each adjustable compensation device, the historical response delay distribution under light load, normal, and heavy load conditions is extracted, and the 95th percentile is taken as the upper limit of the delay for each condition to form a three-segment delay spectrum. Among all participating compensation devices, the device with the largest upper limit of delay under the current operating condition is selected as the timing anchor point, and its instruction start time is set as the reference time.

13. The adaptive zoning compensation method for dynamic line loss in distribution networks according to claim 12, characterized in that, The step of forming the temporal interleaving compensation sequence includes: Define the response action slope for each device. ,in , Preset upper and lower limits for slope. This represents the maximum reference value for health. Rate your health. Adjust precision Mapping to adjust stability margin The higher the adjustment precision, the better. The smaller the value, the shorter the time required for the device to enter a stable confirmation state after reaching the target value; The stable interval time is determined by adding the start time of the anchor point device to the quotient of the adjustment amount and the slope, and the stability margin. The stable interval time of the previous device is used as the instruction start time of the next device. This forms a sequentially connected time-interleaved compensation sequence, and the compensation amount is given to each device according to the time sequence.

14. A dynamic line loss adaptive zoning compensation system for a distribution network for implementing the method according to any one of claims 1-13, characterized in that, The system includes: The sensitivity analysis and coupling quantization module is used to form a line loss sensitivity matrix from the changes in node injected power and line impedance, with the column vectors representing the line loss sensitivity direction, and to calculate the loss coordination degree between nodes, and to obtain the loss coupling strength between nodes by combining the electrical distance. The dual-channel dynamic partitioning and constraint construction module is used to initiate a dual-channel competitive fusion structure for dynamic partitioning. The first channel obtains a sensitivity partitioning view based on the spectral clustering loss coupling strength, and the second channel obtains a behavioral partitioning view based on the clustering power fluctuation mutual information. The game payoff matrix is ​​constructed using the cross-partition line loss interaction entropy change and the intra-partition behavioral homogeneity change, and the dynamic partitioning scheme is evolved. The partitioning boundary constraints are constructed based on the boundary power exchange sensitivity and voltage coupling coefficient. The partitioned potential energy field construction and node delimitation module is used to construct the line loss potential energy field within each partition based on the increase in active power loss, the amount of node voltage deviation, and the change in homogeneity of behavior within the partition. The candidate compensation set is obtained by tracing back from the peak value of the potential energy field along the direction of line loss sensitivity. The candidate set is cyclically and slightly adjusted to recalculate the line loss potential energy field, verify the boundary constraints, retain the valid ones and discard the invalid ones, until the target is met or the key nodes and the initial value of the adjustment are output. The heterogeneous equipment collaborative allocation and cross-regional coordination module is used to establish a dynamic performance profile for key node adjustable equipment, including response delay, adjustment accuracy and health score. It allocates the regional adjustment amount according to the Shapley marginal contribution of each equipment to the potential energy field decline to obtain the initial allocation scheme; verifies the boundary constraints, and transfers the excess to the second healthiest equipment in the same region or the equipment in the neighboring region, and iteratively obtains the compensation amount given value. The timing interleaving and execution module is used to determine the communication traffic conditions and construct a three-segment delay spectrum; taking the device with the largest upper bound of the delay spectrum as the timing anchor point, the health score is mapped to the response action slope, the adjustment accuracy is mapped to the adjustment stability margin, the instruction start time is determined, a timing interleaving compensation sequence is formed and issued for execution.

Images