Hybrid distribution transformer power resource allocation method and system based on potential energy gradient

Abstract

The invention discloses a potential energy gradient-based hybrid distribution transformer power resource allocation method and system, and the method comprises the steps: distinguishing a power excess station region from a power vacancy station region through determining the adjacent communication topology of the station regions, and carrying out the statistics of an initial power deviation; obtaining net power of the whole network by utilizing consistency iteration, setting node target power and calculating potential energy; inter-station bidirectional power transmission is calculated based on potential energy gradient, damping and cutting processing is carried out in combination with the capacity utilization rate, node power is updated in parallel, and a symbol preserving strategy is adopted to prevent power overturning; in the iteration process, the optimization stage is adaptively switched according to the potential energy deviation and the residual vacancy change so as to improve the convergence efficiency, meanwhile, the vacancy fair proportion is periodically verified, and the distribution fairness is maintained through fine adjustment; and finally, a steady-state power distribution scheme, transmission, residual vacancy and other information are output when convergence or capacity is limited, and rapid, autonomous and near-optimal power mutual aid and vacancy relief under the condition of multi-zone-area interconnection are realized.

Description

Technical Field

[0001] This invention relates to the fields of distribution automation and distributed optimization technology, specifically a power resource allocation method and system for hybrid distribution transformer groups based on potential energy gradients. It is used to achieve rapid, autonomous, and near-optimal power redistribution and deficit mitigation under conditions of multiple interconnected transformers. Background Technology

[0002] With the rapid increase in high-proportion renewable energy (such as distributed photovoltaic and wind power), new energy storage, and flexible and adjustable loads at the distribution level, single-frequency transformers and traditional centralized dispatch structures are finding it increasingly difficult to fulfill the functions of dynamic power coordination across distribution areas and multi-port energy integration. The concept of a "hybrid distribution transformer" proposed in international research integrates a highly reliable and efficient frequency core / winding structure with a power electronic conversion unit possessing fast control characteristics. This allows it to perform conventional power conversion while also enabling flexible voltage regulation, power quality improvement, harmonic and fluctuation suppression, and multi-energy port access. Therefore, it has become an important basic equipment for new distribution areas, with significant engineering application prospects and research value.

[0003] Furthermore, by expanding DC / AC grid-connected interfaces and cascaded or parallel power electronic ports, hybrid distribution transformers can form reconfigurable interconnected clusters across multiple distribution areas, constructing a composite distribution network structure including AC / DC coupled microgrids. Under this structure, distributed generation and energy storage across distribution areas can absorb local excess power and rapidly compensate for remote deficits, supporting coordinated regulation of source-load-storage and mitigating the impact of high-penetration intermittent resources (such as midday photovoltaic peak decline or rapid cloud cover) on local voltage and power flow. Compared to traditional strategies relying on centralized dispatch or unidirectional "excess → deficit" transfer, this multi-port, interconnected cluster mode is easier to implement: ① Enabling transfer paths to improve utilization; ② Achieving power redistribution closer to the global optimum when local capacity constraints exist; ③ Reducing centralized communication and maintenance costs; ④ Enhancing resilience and scalability through distributed autonomy. Therefore, it is urgent to design a distributed power resource allocation method capable of bidirectional transmission and capacity awareness, tailored to the characteristics of hybrid distribution transformer clusters, to provide fundamental support for future AC / DC integrated distribution networks. Summary of the Invention

[0004] Purpose of the Invention: The purpose of this invention is to address the needs of multi-source, multi-port, capacity heterogeneous, and cross-regional dynamic power redistribution in hybrid distribution transformer clusters within AC / DC integrated distribution networks. It provides a distributed power resource allocation method and system that is centralized, bidirectional, and has fair constraints.

[0005] Technical solution: The method described in this invention is applied to a power distribution network containing multiple hybrid distribution transformer nodes, including:

[0006] Determine the adjacency communication topology between each transformer area and its neighboring transformer areas; classify transformer areas with excess and insufficient power and calculate the initial excess and insufficient power.

[0007] Perform a consistent iteration to obtain the total net power and total network statistics. The total network statistics include the total net power, the total excess power, and the total power deficit. Set the target power for each node based on the excess or deficit status of the entire network and form the initial potential energy.

[0008] The gradient base transmission volume is calculated based on potential energy, and damping is applied according to capacity utilization. The expected transmission volume of the line is obtained based on the gradient base transmission volume and damping coefficient. The expected transmission volume of the line is then trimmed to obtain the accumulated active power transmission volume on the tie line. The node net power is updated in parallel, and a symbol-preserving strategy is applied to prevent power flipping and record the indicators.

[0009] Monitor changes in potential energy and remaining deficit, and adaptively switch optimization phases to focus on different convergence objectives; periodically check the deficit fairness ratio, and if the fairness deviation exceeds the threshold, fine-tune the power allocation to maintain a fair structure.

[0010] After determining convergence or capacity limitation, the system outputs final power, transmission, remaining deficit, and audit logs for operation and maintenance and decision-making.

[0011] Furthermore, gradient base transfer volume The calculation is as follows: ,in, , , For nodes The degree, , For nodes The set of adjacent nodes; For nodes No. Potential energy after the next iteration For nodes No. The potential energy after the next iteration is calculated using the following formula: , for The node's current net power, The target power is set.

[0012] Furthermore, the methods for calculating the expected transmission volume of the line include:

[0013] ;

[0014] ;

[0015] ;

[0016] And perform bidirectional capacity pruning on the expected transmission volume of the line: If Then take the upper limit of the difference; if it is less than Take the lower bound; otherwise, keep the original expected value; where, For capacity utilization, The damping factor, The expected transmission volume of the line, This represents the cumulative active power transmitted on the tie line at iteration number t, with the positive or negative sign indicating the direction. The bidirectional capacity limit of tie line (i,j) is the cumulative active power transmission amount on the tie line at iteration number t+1. Update using the following formula: ,make sure .

[0017] Furthermore, nodes Target power The settings follow these principles: when the absolute value of the total power deficit in the entire network is greater than the surplus, the target power of the surplus nodes is set to zero, and the target power of the deficit nodes is reduced according to the deficit compensation ratio; otherwise, the target power of the deficit nodes is set to zero, and the target power of the surplus nodes is scaled and retained according to the ratio of net power to total surplus.

[0018] Furthermore, by monitoring changes in potential energy and remaining deficit, the optimization phase is adaptively switched to focus on different convergence objectives; convergence is determined by the maximum change between iterations. With the maximum change in potential energy The joint criteria, in the initial stage, adopt a target deviation-driven stage, to... The main trend is downward, if The iteration will terminate if the number of iterations reaches a threshold. For nodes In the Net power at the next iteration For nodes In the Net power at the next iteration For nodes No. Potential energy after the next iteration This is the convergence tolerance threshold, used to determine whether convergence has occurred.

[0019] When the criterion fails to converge, continuous The next iteration satisfies Then switch to the deficit reduction-driven phase, based on the remaining deficit across the entire network. The trend is mainly downward; if it continues The improvement in the deficit during the second iteration was lower than Or meet Then, if the capacity is deemed limited, the process will terminate. ; The iteration count threshold for stage stagnation determination is used for stagnation identification in the potential energy deviation stage and the deficiency stage, respectively. For small change coefficients, the sensitivity amplification corresponds to the improvement of deviation or deficiency; For the first The maximum change in potential energy at the next iteration. For the first The remaining gap in the entire network at the next iteration.

[0020] Furthermore, in the deficit reduction driving phase, the transmission direction on the edge adopts the following approach: excess → deficit priority, dual deficits aim to alleviate the maximum deficit, and power balancing of multiple excess nodes avoids single-point overload.

[0021] Furthermore, it includes power sign preservation constraints: if a node that was initially positive is to be negative after the update, it is truncated to zero; if a node that was initially negative is to be positive after the update, it is truncated to zero, thus preventing the sign of the node's physical role from flipping.

[0022] Furthermore, under the condition of global deficit, the final state deficit allocation maintains the fairness of the initial deficit ratio: for any deficit node final power satisfy , for The initial net power of the node, for The initial net power of the node, and its fairness deviation. Not exceeding the set tolerance .

[0023] Furthermore, the consistency iteration adopts: ,in, For nodes In iteration Consistency estimates For nodes In iteration Consistency estimates For nodes In iteration Consistency estimates This indicates that for all adjacent nodes Sum.

[0024] The system described in this invention is applied to a power distribution network containing multiple hybrid distribution transformer nodes, comprising:

[0025] The communication topology determination module is used to determine the adjacency communication topology between each transformer area and its neighboring transformer areas;

[0026] The initial power statistics module is used to classify power-excess and power-deficient areas and to count the initial power excess and power deficiency.

[0027] The module for calculating net power and statistics of the entire network is used to perform consistent iteration to obtain net power and statistics of the entire network. The statistics of the entire network include the total net power of the entire network, the total power excess of the entire network, and the total power deficit of the entire network.

[0028] The target setting and initial potential energy module is used to set the target power of each node according to the over- or under-capacity status of the entire network, and to form the initial potential energy.

[0029] The active power transmission calculation module is used to calculate the gradient base transmission based on potential energy, apply damping according to capacity utilization, obtain the line expected transmission based on the gradient base transmission and damping coefficient, and trim the line expected transmission to obtain the accumulated active power transmission on the tie line.

[0030] The symbol-preserving module is used to update the node's net power in parallel, applying a symbol-preserving strategy to prevent power flipping and recording metrics;

[0031] An adaptive two-stage scheduling module is used to monitor changes in potential energy and remaining deficit, and adaptively switch optimization stages to focus on different convergence objectives;

[0032] The ratio monitoring module is used to periodically check the fairness ratio of the deficit. If the fairness deviation exceeds the threshold, the power allocation is finely adjusted to maintain the fairness structure.

[0033] The convergence judgment module is used to output the final power, transmission, remaining deficit, and audit log after determining convergence or capacity limitation for operation and maintenance and decision-making.

[0034] Beneficial effects: Compared with the prior art, the significant technical effects of the present invention are: (1) Achieving global optimization: By allowing bidirectional and transit flows, this method breaks the limitations of traditional unidirectional flows, can discover and utilize better allocation paths, and make the final resource allocation result closer to or even reach the theoretical global optimal solution; (2) Fully distributed and highly scalable: The algorithm does not require any central controller, each node only needs to communicate with its neighbors, the computational overhead is small, and it has strong adaptability to the expansion of network scale; (3) Robustness and adaptability: For changes in network topology or the addition / exit of nodes, the algorithm can adaptively adjust the resource flow and re-achieve balance; the capacity-aware damping mechanism also enhances the stability of the system when approaching the physical limit; (4) Wide application: The model is abstract and universal, and can be widely applied to various distributed resource management scenarios of power dispatch in smart grids. Attached Figure Description

[0035] Figure 1 This is a flowchart of the method of the present invention;

[0036] Figure 2 This is a power evolution diagram for area 1 in scenario 1;

[0037] Figure 3 This is a diagram showing the evolution of line traffic in scenario 1.

[0038] Figure 4 The convergence curve is shown for scenario 1.

[0039] Figure 5 Scenario 1 - Network surplus, sufficient capacity - initial state topology diagram;

[0040] Figure 6 Scenario 1 - Network surplus, sufficient capacity - final state topology diagram;

[0041] Figure 7 The diagram shows the power evolution of the transformer area in scenario 2.

[0042] Figure 8 This is a diagram showing the evolution of line traffic in scenario 2.

[0043] Figure 9 The convergence curve for scenario 2;

[0044] Figure 10 Scenario 2 - Network-wide capacity shortage, sufficient capacity - initial state topology diagram;

[0045] Figure 11 Scenario 2 - Network-wide capacity shortage, sufficient capacity - final state topology diagram;

[0046] Figure 12 Power evolution diagram for Scenario 3;

[0047] Figure 13 This is a diagram showing the evolution of line traffic in scenario 3.

[0048] Figure 14 The convergence curve for scenario 3;

[0049] Figure 15 Scenario 3 - Abundant but Tight Lines - Initial State Topology Diagram;

[0050] Figure 16 Topology diagram of the final state of scenario 3 - surplus but tight lines;

[0051] Figure 17 Power evolution diagram for Scenario 4;

[0052] Figure 18 This is a diagram showing the evolution of line traffic in scenario 4.

[0053] Figure 19 The convergence curve for scenario 4;

[0054] Figure 20 The initial topology diagram for scenario 4, where there is a shortage of units and limited capacity.

[0055] Figure 21 Topology diagram of the final state of scenario 4 (shortage and limited capacity);

[0056] Figure 22 Power evolution diagram for Scenario 5 area;

[0057] Figure 23 This is a diagram showing the evolution of line traffic in scenario 5.

[0058] Figure 24 The convergence curve for scenario 5;

[0059] Figure 25 Scenario 5 - Abundant but Severely Limited Transmission - Initial State Topology Diagram;

[0060] Figure 26 The final state topology diagram for scenario 5 - with surplus but severely limited transmission. Detailed Implementation

[0061] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0062] This invention addresses the power redistribution needs of hybrid distribution transformer clusters under conditions of multi-area interconnection, bidirectional resource flow, and heterogeneous capacity. It proposes a distributed power resource allocation method and system based on a collaborative mechanism of "potential energy deviation—capacity damping—symbol holding—adaptive phase." , Construct nodal targets, and then use potential energy deviation. A unified characterization state is used to drive bidirectional gradient flow along arbitrary connection lines; capacity utilization damping and sign-holding constraints are combined to suppress overshoot and role reversal; and the optimization focus is switched based on iterative stagnation monitoring, thereby rapidly approaching theoretical equilibrium when capacity is sufficient. for The initial net power of the node, for The node's current net power, For the set target power, For nodes No. Potential energy after the next iteration.

[0063] This invention comprises an AC / DC integrated power distribution network formed by interconnecting and communicating multiple differentiated hybrid distribution transformer substations. Each "substation" is a local unit of the low-voltage power distribution network with a hybrid distribution transformer as the core of energy conversion, capable of connecting distributed power sources (such as photovoltaics and energy storage), AC loads, and DC loads. Substations establish a sparse adjacency structure through pre-configured or self-configured communication links, enabling the orderly flow of cross-regional power and status information without relying on centralized control at a single point.

[0064] In this interconnected system, each distribution area is equipped with at least one DC interconnection port connected to the high-voltage DC bus (or hub DC bus) to realize bidirectional power dispatching and cross-regional transfer of excess / deficient power between different distribution areas, and is regarded as a node in the distribution network. The so-called "adjacent node" refers to a set of distribution areas that have direct communication and potential power interaction needs in terms of spatial or engineering layout. One distribution area can correspond to multiple adjacent distribution areas, thus forming a scalable connection topology.

[0065] The port structure within a single hybrid distribution transformer substation includes medium-voltage AC ports (HVAC), low-voltage AC ports (LVAC), low-voltage DC ports (LVDC), and DC interconnection ports for cross-regional connections or access to DC buses. The medium-voltage and low-voltage AC ports are connected in parallel to the traditional distribution network AC bus, supporting conventional power supply and power quality regulation. The low-voltage AC ports can be connected to distributed power sources via AC-DC conversion. The low-voltage DC ports connect to the local DC bus and are connected to DC loads or DC distributed power sources via DC-DC conversion, enabling the coordination of multiple resources (source, load, and storage) within the same substation. The DC interconnection ports face external DC backbones or interconnection nodes in other substations, used for outputting or absorbing cross-regional power.

[0066] The power resource allocation method of the present invention is completely distributed—each transformer area handles measurement, communication and calculation locally, and achieves global coordination through neighboring information exchange.

[0067] Furthermore, to avoid repetition, this invention will not elaborate on the basic electrical structure of conventional AC distribution buses and power electronic interfaces; relevant conventional components can be integrated using existing mature products or standard modules. This invention emphasizes utilizing the multi-port characteristics and cluster interconnection capabilities of hybrid distribution transformers to achieve fair, rapid, and robust reallocation of power resources through a distributed iterative strategy, providing a system environment foundation for subsequent detailed module functions and algorithm processing.

[0068] like Figure 1 As shown, the method of the present invention includes the following steps:

[0069] S1. Initialization: This includes determining the adjacency communication topology between each transformer station and its neighboring transformer stations, classifying transformer stations with excess and insufficient power, and calculating the initial excess and insufficient power amounts; specifically:

[0070] S11. Determine the adjacency communication topology between each transformer area and its neighboring transformer areas.

[0071] Among them, the adjacency communication topology can be used to characterize the communication connection between each transformer area and its neighboring transformer areas.

[0072] Specifically, each transformer substation can communicate with its neighboring substations through communication lines, thereby obtaining information about the adjacent communication topology.

[0073] S12. Classify power-excess and power-deficient areas and calculate the initial power excess and power deficiency.

[0074] In this context, a power excess area refers to a transformer station where the generated power exceeds the consumed power, while a power deficit area refers to a transformer station where the consumed power exceeds the generated power. Initial power excess and deficit refer to the absolute values ​​of the difference between the total power generated and the total power consumed in each transformer station.

[0075] Specifically, the initial net power of each node is collected. The excess and deficit sets are divided according to the initial net power (positive or negative). Line capacity is established. Current line traffic and set control parameters A preliminary statistical analysis of the total deficit and surplus was conducted to provide initial values ​​for the next step of global consistency estimation. Among these, , , For nodes The degree, , For nodes The set of adjacent nodes; This is the consistency step size coefficient; This is the convergence tolerance threshold, used to determine whether convergence has occurred. The iteration count threshold for stage stagnation determination is used for stagnation identification in the potential energy deviation stage and the deficiency stage, respectively. For small change coefficients, the sensitivity amplification (dimensionless) corresponds to the improvement of deviation or deficiency. To set tolerances.

[0076] The detailed steps for node classification are as follows: Each node Obtain its initial net power ,in accordance with When the generated power exceeds the absorbed power, it is denoted as a "power excess node". When the power absorbed is greater than the power generated, it is called a "power deficit node".

[0077] S2. Setting Target Power: This includes performing consensus iterations to obtain the total net power and total network statistics. The total network statistics include the total net power, the total excess power, and the total power deficit. Based on the network's excess or deficit status, the target power for each node is set, and initial potential energy is generated. Specifically:

[0078] S21. Perform a consensus iteration to obtain the total net power and total network statistics. The total network statistics include the total net power, the total excess power, and the total power deficit. Specifically, they include:

[0079] The consensus estimation phase uses an iterative method to obtain consensus estimates: initialization , The initial value for iteration, iteration ,when That is, convergence. The consistent estimate obtained after convergence is the output value of the iteration. The initial net power of each node. Using this as the initial value for iteration, an approximate value of the net power of the entire network can be obtained. ,in, For nodes The initial consensus estimate; For nodes In iteration Consistency estimates For nodes In iteration Consistency estimates For nodes In iteration Consistency estimates The set convergence threshold can be selected from 0.05 to 0.15.

[0080] Distributed computation of network statistics: obtaining the total net power of the entire network through iteration. Total excess power across the entire network and the total power deficit of the entire network The specific process is as follows: each node performs excess transformation on its local power. and deficit transformation Then, iterations are performed for each transformation (in the same form as the net power mentioned above). After convergence, the consensus estimate obtained at any node is multiplied by the total number of nodes. This corresponds to the total amount across the entire network:

[0081] ;

[0082] ;

[0083] ;

[0084] The total network volume is used to determine whether the system is in a "network-wide shortage" situation. ) or "over- or balanced across the entire network" The state directly determines the branch of the strategy for setting the target. Further explanation: This method is completely distributed, requiring no central node for aggregation or broadcasting, and converges to the required accuracy within a few rounds solely through adjacency swapping; the convergence speed is related to graph connectivity, maximum degree, and step size. Related.

[0085] S22. Set the target power of each node and form the initial potential energy according to the over- or under-capacity status of the entire network.

[0086] Specifically, when the total excess power exceeds the total power deficit, it is determined to be an over-limit state for the entire network; when the total power deficit exceeds the total excess power, it is determined to be a deficit state for the entire network.

[0087] Specifically, when the absolute value of the total power deficit in the entire network is greater than the surplus, the target of the surplus node is set to zero, and the deficit node is reduced according to the deficit compensation ratio; otherwise, the target of the deficit node is set to zero, and the surplus node is scaled and retained according to the ratio of net power to total surplus.

[0088] This refers to the target power that the method of this invention aims to achieve. The allocation strategy is detailed as follows:

[0089] (a) If (Network-wide shortage): Excess nodes ; Deficit node This means that each shortfall is partially made up proportionally, so that the remaining shortfall retains a relative proportion.

[0090] (b) If (Oversubscription across the entire network): Nodes with insufficient quotas Excess nodes That is, after the overall output meets the shortfall, the excess is retained according to the net and required proportions.

[0091] After the target power is set, the initial potential energy is obtained. , To initialize the potential energy. Potential energy symbol distinction: This indicates that the target value is relatively high (margin can be output). This indicates that the value is too low (and needs to be compensated).

[0092] S3. Iterative Calculation: Based on the potential energy, calculate the gradient-based basic transmission volume, and iteratively calculate the accumulated active power transmission volume on the tie line. If convergence to the target power is achieved, the process stops. If convergence fails, switch targets and continue iterating. Simultaneously, ensure that the active power on the tie line does not exceed the bidirectional capacity limit, and that the node power sign does not flip. Specifically, the following steps are included:

[0093] S31. Calculate the gradient base transmission volume based on potential energy, and apply damping according to capacity utilization. Obtain the line expected transmission volume based on the gradient base transmission volume and damping coefficient, and trim the line expected transmission volume to obtain the accumulated active power transmission volume on the tie line.

[0094] Potential energy refers to the difference between the node's current net power and the target power, and the line power is calculated based on this value. Capacity utilization rate refers to the value of the current calculated line power relative to the line power capacity. Damping refers to the iteration coefficient, and pruning refers to ensuring that the line power does not exceed the limit.

[0095] Gradient base transport for any tie line calculate ,in, For gradient base transmission volume, For nodes No. Potential energy after the next iteration For nodes No. The potential energy after the next iteration is calculated using the following formula: In this invention, it is permitted that... If positive or negative, it supports bidirectional power redistribution and the activation of relay paths, improving resource utilization in complex topologies. The power plan is determined by the node. Flow to Node If it is negative, then from the node Flow to Node This design allows for the use of relay lines to improve the overall deficit mitigation rate.

[0096] The formula for calculating capacity utilization is as follows:

[0097] ;

[0098] in, For capacity utilization, The cumulative active power transmission on the tie line at iteration number t.

[0099] Then we obtain the damping factor:

[0100] ;

[0101] in, It is the damping factor;

[0102] Desired transmission volume for the line:

[0103] ;

[0104] in, This represents the expected transmission volume of the line.

[0105] Then, bidirectional capacity pruning is performed on the expected transmission volume of the line, including: if Then take the upper limit of the difference; if it is less than Take the lower bound; otherwise, keep the original expected value; where, For capacity utilization, The damping factor, The expected transmission volume of the line, This represents the cumulative active power transmitted on the tie line at iteration number t, with the positive or negative sign indicating the direction. This represents the upper limit of the bidirectional capacity of the connection line (i,j). This represents the cumulative active power transmitted on the tie line at iteration number t+1. Update using the following formula: Ensure that the update still meets the requirements. .

[0106] S32. Update node net power in parallel, apply a symbol preservation strategy to prevent power flipping and record metrics.

[0107] Among them, the sign preservation strategy means that the sign of the node power remains unchanged, and the indicators refer to data such as the maximum deviation and the remaining deficit.

[0108] Each node simultaneously aggregates the increments of adjacent edges and updates them. Then recalculate .

[0109] The sign preservation strategy ensures that nodes initially positive are truncated to zero if they are updated to negative, and nodes initially negative are truncated to zero if they are updated to positive, thus preventing sign flips in the node's physical role. Specifically, if the initial net power... Then it will not appear during operation. If initially Then it will not appear After each parallel update, potential symbol changes are detected and truncated to zero, while the relevant line increments are called back to maintain consistency.

[0110] S4. Target Optimization: This includes the following steps:

[0111] S41. Monitor changes in potential energy and remaining deficit, and adaptively switch the optimization phase to focus on different convergence objectives.

[0112] Specifically, when the pre-set target, namely the node target power in S4, can be met, convergence can be achieved in the first stage; if the target cannot be achieved due to capacity limitations, the second stage is entered, with the target being the minimum net power.

[0113] Define the maximum change in potential energy for:

[0114] ;

[0115] Convergence is determined by the maximum change in net power between iterations. :

[0116] ;

[0117] in, This represents the net power at iteration number t-1.

[0118] The steps of adaptive two-stage scheduling are as follows: Convergence is determined by the maximum change between iterations. With the maximum change in potential energy The joint criteria, in the initial stage, adopt a target deviation-driven stage, to... The main trend is downward, if The iteration will terminate if the threshold number of iterations is reached. This is the convergence tolerance threshold used to determine whether convergence has occurred. When the criterion fails to converge, if continuous... The improvement in the maximum change of secondary potential energy is lower than ,Right now The system has entered the deficit reduction phase, based on the remaining deficit across the entire network. The main trend is downward; if it continues during this phase... Second iteration of deficit improvement Below Or meet Then, if the capacity is deemed limited, the process will terminate. Output the final allocation result; The iteration count threshold for stage stagnation determination is used for stagnation identification in the potential energy deviation stage and the deficiency stage, respectively. For small change coefficients, the sensitivity amplification corresponds to the improvement of deviation or deficiency; For the first The maximum change in potential energy at the next iteration. For the first The remaining deficit in the entire network at the next iteration. During the two-stage switching process, only the evaluation metrics are changed, without altering the underlying gradient and damping calculations, ensuring the simplicity of the control logic.

[0119] During the deficit reduction driving phase, the transmission direction of the opposite side adopts the following approach: excess → deficit priority, dual deficits aim to alleviate the maximum deficit, and power balancing of multiple excess nodes avoids single-point overload.

[0120] S42. Periodically check the fairness ratio of the deficit. If the fairness deviation exceeds the threshold, fine-tune the power allocation to maintain the fairness structure.

[0121] Fairness refers to the average power distribution across the distribution area.

[0122] Fairness is maintained for all nodes with shortages in the event of a global shortage. Maintain final power , Let J be the initial net power of node j, and the fairness deviation be... No more than , To set tolerances. This mechanism ensures even power distribution.

[0123] S5. Output results: After convergence or capacity limitation determination, output final power, transmission, remaining deficit and audit log for operation and maintenance and decision-making; realize fast, autonomous and near-optimal power mutual assistance and deficit mitigation under the condition of multi-area interconnection.

[0124] The recording module saves data such as maximum deviation, remaining deficit, transmission volume, capacity utilization distribution, symbol truncation events, and fairness deviation in each iteration.

[0125] The termination condition is: (a) (b) Capacity-limited termination (power cannot reach the target due to capacity limitations); (c) Number of iterations , This is the maximum number of iterations. When (c) is triggered, output the lower bound estimate of the remaining deficit. And record the set of lines with full capacity. , This represents the final active power transmission volume on the interconnection line, for reference in subsequent capacity expansion decisions.

[0126] Output metric expansion: except , , In addition, it also includes: the maximum change in potential energy. Maximum deviation in fairness Average utilization rate saturation edge ratio Stagnation determination category (convergence / iteration upper limit). Let be the number of edges.

[0127] The system described in this invention is applied to a power distribution network containing multiple hybrid distribution transformer nodes, comprising:

[0128] The communication topology determination module is used to determine the adjacency communication topology between each transformer area and its neighboring transformer areas;

[0129] The initial power statistics module is used to classify power-excess and power-deficient areas and to count the initial power excess and power deficiency.

[0130] The module for calculating net power and statistics of the entire network is used to perform consistent iteration to obtain net power and statistics of the entire network. The statistics of the entire network include the total net power of the entire network, the total power excess of the entire network, and the total power deficit of the entire network.

[0131] The target setting and initial potential energy module is used to set the target power of each node according to the over- or under-capacity status of the entire network, and to form the initial potential energy.

[0132] The active power transmission calculation module is used to calculate the gradient base transmission based on potential energy, apply damping according to capacity utilization, obtain the line expected transmission based on the gradient base transmission and damping coefficient, and trim the line expected transmission to obtain the accumulated active power transmission on the tie line.

[0133] The symbol-preserving module is used to update the node's net power in parallel, applying a symbol-preserving strategy to prevent power flipping and recording metrics;

[0134] An adaptive two-stage scheduling module is used to monitor changes in potential energy and remaining deficit, and adaptively switch optimization stages to focus on different convergence objectives;

[0135] The ratio monitoring module is used to periodically check the fairness ratio of the deficit. If the fairness deviation exceeds the threshold, the power allocation is finely adjusted to maintain the fairness structure.

[0136] The convergence judgment module is used to output the final power, transmission, remaining deficit and audit log after convergence or capacity limitation judgment for operation and maintenance and decision-making, so as to realize rapid, autonomous and near-optimal power mutual assistance and deficit mitigation under the condition of multi-area interconnection.

[0137] Each module primarily interacts with adjacent transformer areas to avoid centralization.

[0138] Example Verification 1: Global Over-Limit and Sufficient Capacity Scenario;

[0139] This embodiment is used to verify whether, under ideal conditions where both excess power and transmission capacity are sufficient, the algorithm can completely satisfy the deficient nodes and fairly allocate the remaining power. For example... Figures 2 to 6 As shown, three excess transformer areas A, B, and C are connected to one deficient transformer area D, with initial power outputs of 150kW, 200kW, 180kW, and -120kW respectively. The overall net power is +410kW, indicating overall excess power. The capacity of the three tie lines is C. AD =150kW, CBD =150kW, C CD =150kW, the total capacity of 450kW is much greater than the total deficit of 120kW. After 177 iterations (tol=1e-3), the algorithm converged: the deficit area D was completely compensated from -120kW to about 0kW; the surplus areas output according to the initial surplus ratio, A decreased from 150kW to 116.04kW (output 33.96kW), B decreased from 200kW to 154.72kW (output 45.28kW), and C decreased from 180kW to 139.25kW (output 40.75kW). The utilization rate of each line remained at a low to medium level (about 23%, 30%, and 27%), without saturation, and the maximum deviation of 0.000979kW met the engineering accuracy. The results show that when capacity is not a constraint, the algorithm can achieve the theoretical optimal allocation: the deficit is completely eliminated and the remaining surplus is fairly retained according to the original ratio, verifying the accuracy and stable convergence ability of the algorithm.

[0140] Example Verification 2: Global Shortage and Sufficient Capacity Scenario;

[0141] This embodiment verifies the algorithm's optimal and fair allocation capabilities when overall resources are insufficient but transmission capacity is sufficient. For example... Figures 7 to 11 As shown, two excess nodes, A and B, connect to two deficit nodes, C and D. Their initial powers are 50kW, 60kW, -120kW, and -160kW, respectively, with a net power of -170kW, indicating a global deficit. Each tie-line has a capacity of 150kW, posing no additional constraint. The algorithm converges after 237 iterations: all excess nodes output to 0, and deficit nodes receive resources according to their initial deficit ratio: C's power improved from -120kW to -72.86kW, and D's improved from -160kW to -97.14kW. The maximum error of 0.000981kW indicates good convergence. The results show that the algorithm prioritizes the full absorption of excess power under global deficit conditions and fairly allocates limited resources according to the deficit ratio.

[0142] Example Verification 3: Global Over-Limit and Capacity Shortage Scenarios;

[0143] This embodiment demonstrates the algorithm's path selection and load balancing capabilities when overall capacity is excessive but line capacity is relatively strained. For example... Figures 12 to 16As shown, the network consists of two excess nodes (A=220kW, B=180kW) and two deficient nodes (C=-150kW, D=-100kW), with a net power of +150kW. Each tie line has a capacity of 100kW, resulting in a total capacity of 400kW. Although this is higher than the total deficient power of 250kW, line limitations may affect path allocation. The algorithm converges after 85 iterations: the power of deficient nodes C and D is fully satisfied and increased to 0; the remaining excess power of excess node A is 82.50kW, and that of node B is 67.50kW, with a maximum error of 0.000888kW. The results show that the algorithm can automatically balance transmission under multi-path conditions.

[0144] Example Verification 4: Global Shortage and Capacity Shortage Scenarios;

[0145] This embodiment verifies the robustness of the algorithm under the dual constraints of insufficient total resources and limited single-line capacity. For example... Figures 17 to 21 The topology shown remains a mesh structure with 2 nodes having excess power (60kW, 90kW) and 2 nodes having deficit power (-160kW, -150kW), with a net power of -160kW. The capacity of each edge is 80kW, totaling 320kW, and the total excess power of 150kW is lower than the total deficit of 310kW. The algorithm converges after 70 iterations: all excess nodes output to 0; deficit nodes improve proportionally, with C decreasing from -160kW to -82.58kW and D decreasing from -150kW to -77.42kW; line utilization remains within the 31%–46% range to avoid saturation; the remaining deficit of 160kW is an inevitable result of insufficient resources, with a maximum error of 0.000923kW. This result demonstrates that even under dual constraints, the algorithm can still quickly reach a balance plateau that cannot be further improved, achieving partial mitigation of fair proportions and maintaining transmission stability.

[0146] Example Verification 5: Severely limited capacity and asymmetric allocation scenario;

[0147] This embodiment focuses on intelligent path selection and capacity-aware allocation under extreme asymmetric capacity constraints, representing the most representative scenario. For example... Figures 22 to 26As shown, four transformer substations, A and B (excess capacity 280.0kW and 370.0kW respectively) and C and D (deficit capacity -240.0kW and -320.0kW respectively), constitute the network, with an overall net power of +90.0kW. The tie line capacity exhibits a strong asymmetry, with the total transmission capacity of 485kW still lower than the total deficit of 560kW, indicating that the deficit cannot be completely eliminated. In the initial iteration, the high-potential A and B output to the low-potential C and D, and the low-capacity AC and BC lines quickly saturate. The algorithm then increases the utilization of the high-capacity backbone AD and the medium-capacity BD lines: AD provides a large amount of transmission at a stable utilization rate of about 83%, BD reaches full load of 130kW, while BC and AC are locked at their capacity limits of 70kW and 65kW respectively. The saturation of the three low / medium capacity lines forms a constraint "framework," and the high-capacity AD retains a small adjustment margin to avoid oscillations. When tol=1e-3, the algorithm rapidly reduced the maximum change from the initial 370kW to 0.00098kW within 89 iterations, achieving a stable equilibrium under capacity constraints: the final power allocation is approximately [31.89, 170.00, -105.00, -6.89] kW, with an improvement of approximately 56% in the capacity deficit (C) and approximately 98% in the capacity deficit (D). The remaining total deficit of approximately 111.89kW represents a physical bottleneck. The entire process demonstrates that the algorithm can intelligently identify efficient paths, use capacity-damped self-regulation near saturation, and maintain its fully distributed characteristics, relying solely on neighbor interactions.

[0148] In summary, this invention can achieve: (1) Fast autonomous balancing: convergence in a distributed iterative manner under the condition of multi-area interconnection, reducing the risk of centralized scheduling waiting and single-point failure. (2) Utilization of transit paths: allow power to be transmitted in multiple hops along indirect links, breaking through the traditional "source to sink" single-hop limitation and improving the available excess absorption rate. (3) Capacity awareness and stability: suppress high saturation edge oscillation through utilization damping and bidirectional pruning, ensuring numerical and operational stability when approaching the capacity limit. (4) Symbol preservation: prohibit the initial excess node from becoming deficient and the initial deficient node from becoming excess. (5) Fair deficiency allocation: maintain the final state deficiency ratio approximately equal to the initial demand ratio when global resources are insufficient. (6) Adaptive stage switching: automatically adjust and optimize the focus target based on the potential energy deviation and the stagnation judgment of the remaining deficiency, shortening the optimization time. (7) Low communication and scalability: only adjacent nodes interact, maintaining near-linear communication overhead as the node scale grows. (8) Multi-port compatibility: Supports the parallel existence of AC / DC and energy storage ports of hybrid distribution transformers, decoupling the redistribution logic from hardware expansion. (9) Simple and adjustable parameters: The core relies on only a few parameters, which are easy to calibrate according to the scale of the transformer area and the noise level.

Claims

1. A method for power resource allocation in a hybrid distribution transformer based on potential energy gradient, characterized by, The method is applied to a power distribution network comprising a plurality of mixed distribution transformer nodes, and comprises: determining the adjacent communication topology between each substation and its adjacent substations; classifying power excess and deficiency substations and counting the initial power excess and deficiency amounts; performing consistent iteration to obtain the total network net power and the total network statistics, the total network statistics including the total network net power amount, the total network power excess amount and the total network power deficiency amount; setting the target power of each node according to the total network excess or deficiency state, and forming the initial potential energy; calculating the gradient-based transmission amount based on the potential energy, and simultaneously applying damping according to the capacity utilization rate, obtaining the line expected transmission amount based on the gradient-based transmission amount and the damping coefficient, and performing clipping on the line expected transmission amount to obtain the cumulative active power transmission amount on the tie line; updating the node net power in parallel, applying the sign preservation strategy to prevent power rollover and recording the indicators; monitoring the potential energy and the remaining deficiency, and adaptively switching the optimization stage to focus on different convergence targets; periodically checking the deficiency fairness ratio, and if the fairness deviation exceeds the threshold, fine-tuning the power distribution to maintain the fairness structure; outputting the final state power, transmission, remaining deficiency and audit log for operation and decision-making after convergence or capacity limitation determination.

2. The method of claim 1, wherein, Gradient base transfer The calculation is as follows: ,in, , , For nodes The degree, , For nodes The set of adjacent nodes; For nodes No. Potential energy after the next iteration For nodes No. The potential energy after the next iteration is calculated using the following formula: , for The node's current net power, The target power is set.

3. The method of claim 1, wherein, The line expected transmission amount calculation method comprises: ； ； ； and perform bidirectional capacity clipping on the line expected transmission: if then take the upper difference value; if less than take the lower limit; otherwise keep the original expected value; where, is the capacity utilization rate, is the damping factor, is the line expected transmission, is the cumulative active power transmission on the tie line at iteration t, and the positive and negative signs represent the direction, is the upper limit of the bidirectional capacity of the tie line (i, j), and the cumulative active power transmission on the tie line at iteration t+1 is updated by the following formula: , ensuring that .

4. The method of claim 1, wherein, Node Target power The setting follows: when the absolute value of the total amount of power shortage in the whole network is greater than the excess, the target power of the excess node is zero, and the target power of the node with power shortage is reduced in proportion to the power shortage; otherwise, the target power of the node with power shortage is zero, and the target power of the excess node is reserved in proportion to the net power and the total excess.

5. The method of claim 1, wherein, Monitor changes in potential energy and remaining deficit, and adaptively switch optimization phases to focus on different convergence objectives; convergence is determined by the maximum change between iterations. With the maximum change in potential energy The joint criteria, in the initial stage, adopt a target deviation-driven stage, to... The main trend is downward, if The iteration will terminate if the number of iterations reaches a threshold. For nodes In the Net power at the next iteration For nodes In the Net power at the next iteration For nodes No. Potential energy after the next iteration This is the convergence tolerance threshold, used to determine whether convergence has occurred. When the criterion fails to converge, the iteration is continued for a certain number of times and then switched to the shortage reduction driving phase, in which the total network remaining shortage is reduced as the main task; if the iteration fails to converge for a certain number of times or meets a certain criterion, the iteration is terminated, i.e. ; is the iteration count threshold for phase stagnation criterion, respectively used for potential energy deviation phase and shortage phase stagnation identification; is the small change coefficient, corresponding to the sensitivity amplification of deviation or shortage improvement; is the maximum change amount of potential energy at the th iteration, is the total network remaining shortage at the th iteration.

6. The method of claim 5, wherein, In the deficiency reduction driven stage, the transmission direction of the edge is as follows: excess→deficiency priority, double deficiency aiming at the largest deficiency relief, and power balancing of multiple excess nodes to avoid single point overload.

7. The method of claim 1, wherein, Further comprising a power sign preservation constraint: if a node initially positive is updated to be negative, it is truncated to zero; if a node initially negative is updated to be positive, it is truncated to zero, so as to maintain the physical role of the node without sign rollover.

8. The method of claim 1, wherein, Under the condition of global deficit, the final state deficit allocation maintains the fairness of the initial deficit ratio: for any deficit node final power satisfy , for The initial net power of the node, for The initial net power of the node, and its fairness deviation. Not exceeding the set tolerance .

9. The method of claim 1, wherein, Consistency iteration employs: where, is the consistency estimate for node at iteration , is the consistency estimate for node at iteration , is the consistency estimate for node at iteration , denotes the summation over all neighboring nodes .

10. A potential gradient based hybrid distribution transformer power resource allocation system, characterized by, The method is applied to a power distribution network comprising a plurality of mixed distribution transformer nodes, and comprises: a communication topology determination module for determining the adjacent communication topology between each substation and its adjacent substations; an initial power statistics module for classifying power excess and deficiency substations and counting the initial power excess and deficiency amounts; a total network net power and statistics calculation module for performing consistent iteration to obtain the total network net power and the total network statistics, the total network statistics including the total network net power amount, the total network power excess amount and the total network power deficiency amount; a target setting and initial potential energy module for setting the target power of each node according to the total network excess or deficiency state, and forming the initial potential energy; an active power transmission amount calculation module for calculating the gradient-based transmission amount based on the potential energy, and simultaneously applying damping according to the capacity utilization rate, obtaining the line expected transmission amount based on the gradient-based transmission amount and the damping coefficient, and performing clipping on the line expected transmission amount to obtain the cumulative active power transmission amount on the tie line; a sign preservation module for updating the node net power in parallel, applying the sign preservation strategy to prevent power rollover and recording the indicators; an adaptive two-stage scheduling module for monitoring the potential energy and the remaining deficiency, and adaptively switching the optimization stage to focus on different convergence targets; a ratio monitoring module for periodically checking the deficiency fairness ratio, and if the fairness deviation exceeds the threshold, fine-tuning the power distribution to maintain the fairness structure; a convergence judgment module for outputting the final state power, transmission, remaining deficiency and audit log for operation and decision-making after convergence or capacity limitation determination.

Images