Power distribution network minute-level collaborative fault recovery method for network-oriented inverter power supply

Abstract

The present application relates to the technical field of power grid dispatching, and particularly relates to a power distribution network minute-level collaborative fault recovery method for network-constructing inverter-type power sources. When a power grid fault is triggered, a load transfer area and an island key area are extracted, wherein the island key area is a power grid outage area that is not covered by the load transfer area; a distributed collaborative power distribution solution is configured for the load transfer area, and a network-constructing collaborative power supply solution is configured for the island key area; through a graph neural network cost analyzer, a recovery duration and a loss of electric energy are generated; when the recovery duration is less than or equal to a duration threshold, and when the loss of electric energy is less than or equal to a loss threshold, the distributed collaborative power distribution solution is used to control the distributed power sources to supply power to the load transfer area, and the network-constructing collaborative power supply solution is used to control the network-constructing inverter-type power sources and the network-following inverter-type power sources to supply power to the island key area. The power distribution network's resilience and power supply reliability in the event of a fault are enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power grid dispatching, and in particular to a power distribution network minute-level collaborative fault recovery method for network-constructed inverter-type power sources. BACKGROUND

[0002] At present, in the field of power distribution network fault recovery, most of the traditional network-constructed control strategies rely on high-reliability communication networks and intelligent terminals. Once the communication is interrupted or the equipment fails, the recovery strategy will fail. At the same time, the real-time performance of the calculation model used is insufficient, and it is difficult to meet the minute-level recovery requirements. In addition, the traditional scheme only configures different power supply types for different regions, does not consider the fine parameters of power supply control, and has the problem of insufficient control precision, and has poor adaptability in dealing with high penetration rate distributed power sources and dynamic reconstruction scenarios. Some schemes can only handle specific types of faults and cannot fully cover complex fault conditions. There are technical problems of low speed, precision and economy of power grid fault recovery. SUMMARY

[0003] The present application provides a power distribution network minute-level collaborative fault recovery method for network-constructed inverter-type power sources to solve the technical problems of low speed, precision and economy of power grid fault recovery in the prior art.

[0004] The technical solution of the present application to solve the above technical problems is as follows:

[0005] In a first aspect, the present application provides a power distribution network minute-level collaborative fault recovery method for network-constructed inverter-type power sources, comprising: when a power grid fault is triggered, extracting a load transfer area and an island key area, wherein the island key area is a power grid outage area not covered by the load transfer area; configuring a distributed collaborative power distribution solution for the load transfer area, and configuring a network-constructed collaborative power supply solution for the island key area, analyzing the recovery time and the loss of electrical energy through a graph neural network cost analyzer, and generating the recovery time and the loss of electrical energy; when the recovery time is less than or equal to a time threshold, and when the loss of electrical energy is less than or equal to a loss threshold, controlling the distributed power sources to supply power to the load transfer area through the distributed collaborative power distribution solution, and controlling the network-constructed inverter-type power sources and the network-following inverter-type power sources to supply power to the island key area through the network-constructed collaborative power supply solution.

[0006] Optionally, when the grid failure trigger, extraction load transfer area and island key area, including: when the grid-connected point voltage is less than the voltage threshold, regarded as grid failure trigger, receive the power grid outage area; based on the power grid outage area line topology map, extraction access to the power grid outage area of distributed power supply; traversal of the distributed power supply to perform power flow calculation, determine the load transfer area of the power grid outage area; extraction of the load transfer area of the power grid outage area not covered area, set as the initial island key area; traversal of the initial island key area for source and load balance analysis, obtain the island key area.

[0007] Optionally, after the grid-connected point voltage is less than the voltage threshold, before receiving the power grid outage area, including: adjust the current overload capacity to the preset multiple rated current within k seconds, wherein, 0≤k≤2.

[0008] Wherein, the load transfer area is configured with distributed collaborative power distribution solution, and the island key area is configured with grid-connected collaborative power supply solution, and the recovery time and the loss of electric energy are generated by analyzing the graph neural network cost analyzer, including: according to the power grid failure area, from the graph neural network cost analyzer set, match the graph neural network cost analyzer, wherein, the graph neural network cost analyzer has a label identifying all line topologies of the power grid failure area; based on the load transfer area line topology, compared with all line topologies, separate the graph neural network cost analyzer to obtain the load transfer area cost analyzer, process the distributed collaborative power distribution solution, generate the first recovery time and the first loss of electric energy; based on the island key area line topology, compared with all line topologies, separate the graph neural network cost analyzer to obtain the island key area cost analyzer, process the grid-connected collaborative power supply solution, generate the second recovery time and the second loss of electric energy; the greater value of the first recovery time and the second recovery time is set as the recovery time; the greater value of the first loss of electric energy and the second loss of electric energy is set as the loss of electric energy.

[0009] The load transfer area cost analyzer is obtained by performing separation on the graph neural network cost analyzer based on comparison between the first load transfer area line topology and the overall line topology.

[0010] The construction steps of the set of graph neural network cost analyzers include:

[0011] Step one: PQ unit line topology enumeration disassembly is performed on the target transformer area line topology to obtain a plurality of PQ unit line topologies, the initial value of P is equal to 1, the initial value of Q is equal to 1, the PQ unit line topology includes Q power nodes and P load nodes, P and Q are integers, P≤total number of load nodes, and Q≤total number of power nodes;

[0012] Step two: the plurality of PQ unit line topologies are traversed, the power distribution position is taken as an input node, and the load distribution position is taken as an output node to perform graph neural network simulation and construct a plurality of PQ unit line graph neural network architectures;

[0013] Step three: the power supply capacity is taken as input data, and the power supply duration and loss energy are taken as output data to train the plurality of PQ unit line graph neural network architectures to generate a plurality of PQ unit line cost analyzers;

[0014] Step four: when P<total number of load nodes, P is incremented by one, and the loop is returned to step one for execution:

[0015] When Q<total number of power nodes, Q is incremented by one, and the loop is returned to step one for execution;

[0016] When Q=total number of power nodes and P<total number of load nodes, P is incremented by one, and the loop is returned to step one for execution;

[0017] Step five: when P=total number of load nodes and Q<total number of power nodes, Q is incremented by one, and the loop is returned to step one for execution;

[0018] Step six: when Q=total number of power nodes and P=total number of load nodes, the overall PQ unit line cost analyzer is output, is stored in association with the target transformer area, and is added to the graph neural network cost analyzer.

[0019] Optionally, the power distribution network minute-level collaborative fault recovery method for the networked inverter-type power supply further comprises: when the recovery time length is greater than the time length threshold, or / and when the loss of electric energy is greater than the loss threshold, updating the distributed collaborative power distribution solution and the networked collaborative power supply solution, and performing a cycle.

[0020] The power distribution network minute-level collaborative fault recovery method for the networked inverter-type power supply further comprises:

[0021] When the cycle does not converge for a preset number of times, a recovery power supply scheme in which the loss of electric energy is less than or equal to the loss threshold is extracted; when the recovery power supply scheme is not empty, a scheme with the shortest recovery time length in the recovery power supply scheme is extracted for power supply; when the recovery power supply scheme is empty, a first weighted feature is obtained by weighting a normalized value of the recovery time length of each power supply scheme based on a predefined time length weight; a second weighted feature is obtained by weighting a normalized value of the loss of electric energy of each power supply scheme based on a predefined loss weight; a power supply scheme adaptability is obtained by weighting a one-to-one corresponding first weighted feature and second weighted feature; and a power supply scheme corresponding to a minimum value of the power supply scheme adaptability is extracted for power supply.

[0022] The time length weight is included in [0.8, 1], and the loss weight is included in [0, 0.2].

[0023] The PQ unit line topology includes Q power supply nodes and P load nodes, and the P load nodes can perform power supply through the Q power supply nodes.

[0024] By implementing the present application, when the power grid fault is triggered, a load transfer area and an island key area can be extracted, the island key area is a power grid outage area not covered by the load transfer area, accurate division of the outage area is realized, a foundation is laid for subsequent targeted development of a power supply scheme, waste of power supply resources and blindness of power supply strategies are avoided, and it is ensured that each area can obtain a suitable power supply mode.

[0025] By implementing the present application, a distributed collaborative power distribution solution can be configured for the load transfer area, a networked collaborative power supply solution can be configured for the island key area, a recovery time length and a loss of electric energy can be generated through a graph neural network cost analyzer, and the advantages of different power supplies (distributed power supplies are suitable for transfer areas, and networked and networked inverter-type power supplies are suitable for island key areas) can be fully utilized through differentiated power supply scheme configuration; the graph neural network cost analyzer can efficiently and accurately analyze the recovery time length and the loss of electric energy, provide a key basis for subsequent judgment of whether a scheme is feasible, and ensure the scientificity and timeliness of the analysis.

[0026] By implementing the present application, when the recovery time length is less than or equal to the time length threshold, and the power loss is less than or equal to the loss threshold, the distributed power supply is controlled to supply power to the load transfer area, the grid-connected network cooperative power supply is controlled to supply power to the island key area, the power supply scheme can be ensured to complete recovery within a specified time, and the power loss is within an acceptable range, the timeliness and economy of power supply are ensured, and the reliability and efficiency of power grid fault recovery are effectively improved.

[0027] In summary, by implementing the present application, the speed, accuracy and economy of power distribution network fault recovery can be improved, and the resilience and power supply reliability of the power distribution network under fault conditions can be enhanced. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 A flowchart of a power distribution network minute-level cooperative fault recovery method for grid-connected inverter-type power supply provided by the present application is shown.

[0029] Figure 2 A schematic diagram of a line topology graph of a power grid outage area is shown.

[0030] Figure 3 A schematic diagram of an island key area in a line topology graph of a power grid outage area is shown. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0032] In the description of the present application, the terms "first" and "second" are used only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with "first" and "second" can explicitly or implicitly include one or more features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise specifically limited.

[0033] In the description of the present application, the term "for example" is used to mean "serving as an example, instance, or illustration." Any embodiment described as "for example" in the present application is not necessarily to be construed as preferred or advantageous over other embodiments. The following description is presented to enable any person skilled in the art to make and use the present application. In the following description, for the purpose of explanation, details are set forth in order to provide a thorough understanding of the present application. It should be noted that persons of ordinary skill in the art can realize other embodiments that do not require the specific details that are set forth. In other instances, well-known structures and processes are not elaborated in order not to obscure the description of the present application with unnecessary details. Thus, the present application is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed.

[0034] Embodiment one, as shown, the power distribution network minute-level collaborative fault recovery method for network-oriented inverter power supply provided by the embodiment of the present application comprises: Figure 1

[0035] S100: When the power grid fault is triggered, extract the load transfer area and the island key area, wherein the island key area is the power grid outage area not covered by the load transfer area;

[0036] S200: Configure a distributed collaborative power distribution solution for the load transfer area, and configure a network-oriented collaborative power supply solution for the island key area, analyze the recovery time and the loss of electric energy through a graph neural network cost analyzer, and generate the recovery time and the loss of electric energy;

[0037] S300: When the recovery time is less than or equal to the time threshold, and the loss of electric energy is less than or equal to the loss threshold, control the distributed power supply to supply power to the load transfer area through the distributed collaborative power distribution solution, and control the network-oriented inverter power supply and the network-following inverter power supply to supply power to the island key area through the network-oriented collaborative power supply solution.

[0038] In step S100 of the embodiment of the present application, when the power grid fault is triggered, the load transfer area and the island key area are extracted, comprising:

[0039] When the grid-connected point voltage is less than the voltage threshold, it is considered that the power grid fault is triggered, and the power grid outage area is received;

[0040] Based on the power grid outage area line topology graph, the distributed power supply connected to the power grid outage area is extracted;

[0041] Iterate the distributed power supply to perform power flow calculation, and determine the load transfer area of the power grid outage area;

[0042] Extract the area in the power grid outage area that is not covered by the load transfer area, and set it as the initial island key area; ​

[0043] Performing source-load balance analysis on the initial island critical area to obtain the island critical area.

[0044] In the embodiments of the present application, the core purpose is to accurately define the load transfer area and island critical area after power grid failure, and to provide clear area division basis for subsequent differentiated power supply scheme configuration. Specifically, by identifying the fault triggering condition, locating the power failure area, combining the distributed distributed power distribution and power flow calculation to determine the range of transferable load, and then screening out the island critical area that needs special coordinated power supply through source-load balance analysis, it ensures that the resource scheduling is more targeted during fault recovery, and avoids the blindness of power supply strategy

[0045] Firstly, the fault triggering condition needs to be set, i.e. the voltage threshold. When the grid voltage is less than the voltage threshold, it is considered that the power grid fails to trigger, and the power failure area of the power grid is received. For example, the voltage threshold can be set in the form of "grid point phase-to-phase voltage U PCC-PP <0.85pu" or "grid point phase voltage U PC-PG <0.9pu", where pu is the unit value, which is used to unify the voltage measurement standard. This determination is directly related to the rapid response in the initial stage of failure, which ensures that the failure is captured in time.

[0046] When the fault is triggered, the system receives the overall range information of the power failure area of the power grid, which is determined based on the node topology, load distribution and other real-time monitoring data uploaded by the distribution network before the failure, providing a basic boundary for subsequent area division.

[0047] Further, the distributed power connected to the power failure area of the power grid needs to be extracted based on the line topology graph of the power failure area of the power grid. This step needs to rely on the line topology graph of the power failure area of the power grid, which is presented in the form of "graph theory model", as shown in the accompanying drawings. Figure 2 Each load branch, external system and line branch in the graph is a node, the line with tie switches, sectional switches and protection switches is a disconnectable edge represented by a dashed line, and the remaining lines are ordinary edges represented by solid lines. By traversing all nodes in the power failure area of the line topology graph of the power failure area, the nodes connected to other distributed power are filtered out, and the position, capacity and operating state of the distributed power connected to the power failure area before the failure are determined according to the data before the failure. These distributed power will be used for subsequent load transfer.

[0048] Further, a simplified power flow calculation model of the distribution network is constructed based on the power flow calculation to determine the load transfer area. The input parameters of the model include: the line length and impedance parameters between nodes; the load power connected to each node, i.e. active power and reactive power; and the real-time output of distributed power.

[0049] Take the non-fault node adjacent to the power failure area as the starting point, such as the nodes in the load transfer node set, input its voltage measurement value, such as the normal range of 0.85-1.15pu, and calculate the voltage backward node by node.

[0050] If the node voltage is always in the normal range during the calculation process, continue to expand the calculation range across the tie switch, sectionalizing switch and the like; if the voltage of a certain area node exceeds the threshold value, such as lower than 0.85pu or higher than 1.15pu, stop expanding.

[0051] When there is a distributed power supply, the voltage support capability is stronger, and the transfer supply range is larger, so it is necessary to combine the support capability of the distributed power supply to finally determine the maximum range of load transfer, i.e. the load transfer area, which is set as the initial island key area.

[0052] Among them, the initial island key area is the part of the power failure area that is not covered by the load transfer area, i.e. the area that cannot be restored by external system or distributed power supply. When extracting the initial island key area, it is necessary to compare through the line topology map of the power failure area, and remove the coverage range of the load transfer area from the total range of the power failure area, and the remaining area is the initial island key area. For example, if the power failure area contains nodes 1-30, and the load transfer area covers nodes 1-10, then the initial island key area is nodes 11-30, which needs to be adjusted in combination with the actual topology map.

[0053] Finally, it is necessary to determine the island key area based on the source-load balance analysis. The source-load balance analysis is a matching calculation of the power supply output and load demand in the initial island key area, to judge whether the power supply in the area can meet the power demand of the load, i.e. active and reactive power balance.

[0054] The analysis logic is that if the source-load balance in the initial island key area, i.e. the power supply output ≥ the load demand, then the area is the final island key area, which needs to rely on the internal network type and network type inverter type power supply for collaborative support.

[0055] If the source-load is unbalanced, i.e. the power supply output < the load demand, then the area boundary needs to be further expanded to include the surrounding power supply, and the above source-load balance analysis is repeated until the island key area that meets the conditions is determined, such as the area composed of nodes 10, 12 and 29 in the attached figure. Figure 3

[0056] If the initial island key area cannot meet the source-load balance after analysis, there is no island key area in the area, which means that the power supply can be completely restored by load transfer.

[0057] ​In step S100 of the embodiment of the present application, after the grid-connected point voltage is less than the voltage threshold, before the power grid outage area is received, the current overload capability is adjusted to a preset multiple of the rated current within k seconds, where 0≤k≤2.

[0058] In the embodiment of the present application, the purpose of setting the mechanism is to provide immediate voltage support for the outage area within a very short time, such as 0-2 seconds, after the power grid failure is triggered, by enhancing the current output capability of the grid-forming inverter, to avoid the grid-following inverter from being disconnected due to voltage abnormalities, and to gain time for subsequent load transfer areas and island key areas to lay a stable initial power supply foundation.

[0059] Specifically, when it is detected that the grid-connected point voltage is less than the voltage threshold, i.e., the power grid failure is triggered, before the information of the power grid outage area is received, the grid-forming inverter will quickly adjust its current overload capability within k seconds, 0≤k≤2, usually within 2 seconds from the moment of failure, by adjusting the upper and lower limits of the d-axis and q-axis current reference values from the conventional 1.2 times of the rated current to the preset 2 times of the rated current, where the d-axis and q-axis current reference values are key control parameters obtained based on the "Park transformation" of three-phase alternating current, used to describe two components of the inverter output current in the rotating coordinate system. The d-axis current, i.e., the direct-axis current, is mainly related to active power, and determines the size of the active power output by the power supply to the grid. The q-axis current, i.e., the quadrature-axis current, is mainly related to reactive power, and affects the voltage support capability of the power supply to the grid, such as voltage amplitude adjustment. In the fault recovery scenario of the present application, the core role of the d-axis and q-axis current reference values is to accurately control the output characteristics of the grid-forming inverter.

[0060] Through this dynamic current limit adjustment, the grid-forming inverter can output larger current, thereby providing stronger voltage support for the power grid and maintaining the voltage in the outage area at a certain level to prevent the grid-following inverter from being disconnected due to voltage drop or fluctuation.

[0061] In step S200 of the embodiment of the present application, the load transfer area is configured with a distributed collaborative power distribution solution, and the island key area is configured with a grid-following collaborative power supply solution, and the recovery time and the loss of electric energy are generated by analyzing the graph neural network cost analyzer, including:

[0062] According to the power grid fault area, the graph neural network cost analyzer is matched from the set of graph neural network cost analyzers, where the graph neural network cost analyzer has a label identifying the topology of all lines of the power grid fault area;

[0063] Based on the line topology of the load transfer area, compared with the whole line topology, the graph neural network cost analyzer is separated to obtain a load transfer area cost analyzer, which processes the distributed collaborative power distribution solution to generate a first recovery time and a first loss of electric energy;

[0064] Based on the line topology of the island key area, compared with the whole line topology, the graph neural network cost analyzer is separated to obtain an island key area cost analyzer, which processes the grid collaborative power supply solution to generate a second recovery time and a second loss of electric energy;

[0065] The larger value of the first recovery time and the second recovery time is set as the recovery time;

[0066] The larger value of the first loss of electric energy and the second loss of electric energy is set as the loss of electric energy.

[0067] In the embodiments of the present application, the graph neural network cost analyzer matched for the fault area ensures the accuracy and pertinence of the subsequent power supply scheme analysis for the load transfer area and the island key area. By matching the graph neural network cost analyzer corresponding to the line topology of the whole grid fault area, a reliable model basis is provided for analyzing the recovery time and the loss of electric energy of the distributed collaborative power distribution solution and the grid collaborative power supply solution, which ensures that the analysis result can truly reflect the actual situation of the fault area, and supports the subsequent optimization decision of the power supply scheme.

[0068] In step S200 of the embodiments of the present application, the construction steps of the graph neural network cost analyzer set include:

[0069] Step one: PQ unit line topology enumeration and disassembly is performed on the line topology of the target area to obtain a plurality of PQ unit line topologies, the initial value of P is equal to 1, the initial value of Q is equal to 1, the PQ unit line topology includes Q power nodes and P load nodes, P and Q are integers, P≤total number of load nodes, Q≤total number of power nodes;

[0070] Step two: traverse the plurality of PQ unit line topologies, take the power distribution position as the input node, and take the load distribution position as the output node, perform graph neural network simulation, and construct a plurality of PQ unit line graph neural network architectures;

[0071] Step three: take the power supply capacity as the input data, take the power supply time and the loss of electric energy as the output data, train the plurality of PQ unit line graph neural network architectures, and generate a plurality of PQ unit line cost analyzers;

[0072] Step four: when P<total number of load nodes, P is incremented by one, and the loop is returned to step one:

[0073] When Q<total number of power nodes, Q is incremented by one, and the loop is returned to step one;

[0074] When Q = total number of power nodes, and P < total number of load nodes, P is incremented by one, and the loop is returned to step one for execution;

[0075] Step five: when P = total number of load nodes, and Q < total number of power nodes, Q is incremented by one, and the loop is returned to step one for execution;

[0076] Step six: when Q = total number of power nodes, and P = total number of load nodes, output all PQ unit line cost analyzers, store them in association with the target substation, and add them to the graph neural network cost analyzer.

[0077] In the embodiments of the present application, by enumerating all combinations of power nodes (Q) and load nodes (P) and training corresponding graph neural network cost analyzers, efficient analysis of power supply schemes under any power-load configuration is realized, providing computing power support for minute-level fault recovery.

[0078] First, in step one, the PQ unit line topology enumeration and disassembly of the target substation line topology is required. That is, based on the complete line topology of the target substation, it is disassembled into several "PQ unit line topologies". Among them, P represents the number of load nodes, the initial value = 1, Q represents the number of power nodes, the initial value = 1, and P ≤ total number of load nodes, Q ≤ total number of power nodes. Each PQ unit topology needs to meet the basic logic that P load nodes can be powered by Q power nodes, ensuring that the unit topology has actual power supply feasibility.

[0079] Then, in step two, the PQ unit line graph neural network architecture needs to be constructed. All disassembled PQ unit topologies need to be traversed,

[0080] The "power distribution location" in each unit is taken as the input node, and the "load distribution location" is taken as the output node.

[0081] Among them, the input node, i.e. the power distribution location, is the core input data of the graph neural network. For example, the power distribution location is reflected through the coordinate information in the node feature, and the graph neural network analyzes the power supply range based on this.

[0082] The output node, i.e. the load distribution location, is the analysis target of the graph neural network. The graph neural network calculates whether the power can cover the load, the required time and the loss through the input node feature and edge feature (line attribute), and finally outputs the result corresponding to the load distribution location.

[0083] Through the simulation training of the graph neural network (GNN), the neural network architecture corresponding to each PQ unit topology is constructed. The characteristics of the graph neural network enable it to accurately capture the topological correlation between nodes, laying a structural foundation for subsequent cost analysis.

[0084] In step three, the PQ unit line cost analyzer can be built using a gated graph neural network. This model can effectively handle the topology and timing characteristics in the power system, accurately capturing the complex interaction between power supply nodes and load nodes.

[0085] The features of the unit line cost analyzer include node features and edge features.

[0086] The node features of the unit line cost analyzer include power supply node features and load node features. The power supply node features include power supply types such as distributed power supply, network construction type, and network following type; capacity (kW), maximum overload capacity such as 2 times the rated current in the previous step; and real-time output (pu) for measuring power supply capacity. The load node features include load types such as residential, industrial, and commercial; power demand (kW); and priority such as primary load and secondary load. The unified feature dimension is 64 dimensions to ensure consistent node feature dimensions.

[0087] The edge features of the unit line cost analyzer include line impedance (R, X), length, rated current, and switch state such as closed / open. The edge feature dimension is 32 dimensions.

[0088] The structure of the PQ unit line cost analyzer is a 4-layer GGNN layer to capture local and global topology information of the power network; the GRU unit hidden dimension is 128 dimensions to enhance the timing modeling capability; and the output structure is a fully connected layer with output dimensions of 64, 32, and 2. The last two layers have dimension features corresponding to power supply duration prediction and loss energy prediction, respectively.

[0089] The input data is the core variable used by the unit line cost analyzer for calculation, specifically "power supply capacity". The actual output power of the power supply, such as the real-time output of the distributed power supply and the output power of the network construction type power supply, is a dynamic operating parameter that directly reflects the current power supply capacity of the power supply.

[0090] The output data is the result of the unit line cost analyzer calculation, specifically "power supply duration" and "loss energy". The time required to restore power supply to the load through the power supply scheme and the energy loss during the restoration process are key indicators for measuring the efficiency of the power supply scheme.

[0091] In the parameter setting of the PQ unit line cost analyzer, the learning rate is set to 0.001 with adaptive adjustment using the Adam optimizer. The batch size is 64. The regularization uses an L2 regularization coefficient of 0.0001 and a Dropout rate of 0.2. The hidden layer activation function is LeakyReLU with α=0.01. The output layer has no activation function. The number of training rounds is 200 rounds.

[0092] The following data of the PQ unit line cost analyzer is historical fault data, which can be extracted from the historical fault records of the power distribution network, including fault occurrence location, power distribution, load data, recovery time length and lost power. A digital twin model can also be built based on the actual power distribution network topology, and fault scenarios such as different location faults and different weather conditions can be generated through Monte Carlo simulation. Combining historical data and simulation data ensures that the samples cover various possible fault types and topologies.

[0093] For each PQ unit line topology, such as 1 power source-1 load, 1 power source-2 load, etc., at least 10,000 training samples are generated. When the mean absolute error (MAE) on the validation set decreases by less than 0.01% for 10 consecutive rounds, the training is stopped, and the PQ unit line cost analyzer is obtained. By inputting the power supply capacity of the corresponding PQ unit line cost analyzer, the power supply time length and the lost power can be predicted. Through the same method, a number of PQ unit line cost analyzers

[0094] For example, taking a PQ unit line topology with 1 power source node (Q=1) and 2 load nodes (P=2) as an example, assuming that the power source node features are: network type inverter power source, capacity 1000 kW, maximum overload capacity 2 times rated current, distribution location at the start of line A, and power supply capacity 800 kW. The load node features are: load 1, residential load, power demand 300 kW, distribution location 1 km from the power source node; load 2, industrial load, power demand 400 kW, distribution location 2 km from the power source node. The edge features are: line impedance 0.05 ohms per kilometer, and switch state closed.

[0095] From the graph neural network cost analyzer, match the PQ unit line cost analyzer with the label "1 power source node, 2 load nodes", input the power supply capacity 800 kW, and the node features such as power source capacity and load demand, and the edge features such as line impedance and length into the PQ unit line cost analyzer. The PQ unit line cost analyzer analyzes through the graph neural network and outputs a recovery time length of 15 minutes and a lost power of 12 kWh.

[0096] The core purpose of steps four to six in the embodiments of the present application is to enumerate all possible PQ unit line topology combinations in the target substation by incrementally increasing the number of power source nodes (Q) and the number of load nodes (P) in a loop, ensuring that the set of graph neural network cost analyzers constructed can cover all power source and load configuration scenarios in the substation. By exhaustive combination, any possible power source-load combination is avoided, thereby ensuring that in the event of a power grid failure, the corresponding analyzer for the actual fault scenario can always be matched from the set, laying the foundation for accurately analyzing the power supply scheme in step S200.

[0097] Assume the initial conditions: P initial value is 1, Q initial value is 1, and P ≤ total number of load nodes, Q ≤ total number of power supply nodes.

[0098] Then when P is less than the total number of load nodes, increase P by 1, such as P from 1 to 2, return to step one to disassemble the PQ unit line topology, that is, generate a topology of Q power supply nodes and new P load nodes, and repeat the subsequent training steps.

[0099] When P reaches the total number of load nodes, if Q is less than the total number of power supply nodes, increase Q by 1, such as Q from 1 to 2, return to step one to disassemble the topology, that is, generate a topology of "new Q power supply nodes, P = total number of load nodes", and repeat the training steps.

[0100] When Q reaches the total number of power supply nodes and P is less than the total number of load nodes, increase P by 1, return to step one to disassemble the topology, such as Q = total number of power supply nodes, P from 2 to 3.

[0101] When P reaches the total number of load nodes and Q reaches the total number of power supply nodes, stop the loop, store all trained PQ unit line cost analyzers associated with the target area, and include them in the graph neural network cost analyzer set.

[0102] For example, if the target area has 2 power supply nodes, that is, total Q = 2, and 3 load nodes, that is, total P = 3, the loop process is:

[0103] P = 1, Q = 1 → P = 2, Q = 1 → P = 3, Q = 1 → Q = 2, P = 1 → Q = 2, P = 2 → Q = 2, P = 3 → loop end, generating 6 PQ unit analyzers.

[0104] In step S200 of the embodiments of the present application, based on the load transfer area line topology, compared with the entire line topology, the graph neural network cost analyzer is executed separately to obtain the load transfer area cost analyzer, including:

[0105] From the load transfer area, extract a first load transfer area, wherein the first load transfer area has an accessible distributed power source position number identifier;

[0106] According to the accessible distributed power source position number identifier, the first load transfer area is configured for power supply with the pre-fault first load transfer area demand power as the target to obtain the first load transfer area cooperative power distribution solution.

[0107] According to the accessible distributed power source position number identifier and the first load transfer area, a first load transfer area line topology is extracted from a power grid outage area line topology graph.

[0108] Based on the first load transfer area line topology, compared with all line topologies, the graph neural network cost analyzer is separated to obtain a load transfer area cost analyzer.

[0109] In the embodiments of the present application, the core purpose of the above steps is to separate the cost analyzer specially applicable to the load transfer area from the matched graph neural network cost analyzer, to ensure that the distributed collaborative power distribution analysis of the load transfer area is targeted and accurate. By extracting the region containing accessible distributed power sources, configuring the power supply scheme, and separating the load transfer area cost analyzer corresponding to the topology, a dedicated tool is provided for calculating the recovery time and loss of electric energy in the region, ensuring that the analysis results are consistent with the actual power supply demand and topology characteristics of the load transfer area, and supporting the optimization decision of the subsequent power supply scheme.

[0110] Firstly, the first load transfer area needs to be extracted from the determined load transfer area, and the sub-area with the "accessible distributed power source position number identification" is selected, i.e. the first load transfer area. The "accessible distributed power source position number identification" refers to the existence of a clear distributed power source access point in the region, such as distributed power source interfaces numbered D1 and D2, which can clearly achieve power supply through distributed power sources.

[0111] Then, according to the accessible distributed power source position number identification, the first load transfer area collaborative power distribution solution is configured for the first load transfer area with the pre-fault demand power of the first load transfer area as the target.

[0112] Specifically, according to the position number identification of the accessible distributed power source, the demand power of the region before the fault is taken as the target, such as 1000kW of power demand before the fault in the region, and the output distribution and access mode of the distributed power source are configured, such as dispatching D1 power source to provide 600kW and D2 power source to provide 400kW, to form a power supply scheme for the sub-area, i.e. the first load transfer area collaborative power distribution solution, as a sub-scheme of the distributed collaborative power distribution solution.

[0113] Then, according to the accessible distributed power source position number identification and the first load transfer area, the first load transfer area line topology is extracted from the power grid outage area line topology graph. That is, the line topology of the sub-area is stripped from the overall line topology graph of the power grid outage area, including internal nodes, branch connection relationships, distributed power source access point positions, etc., in combination with the position number identification of the accessible distributed power source and the range of the first load transfer area.

[0114] Finally, the first load transfer area line topology needs to be compared with the overall line topology based on the first load transfer area line topology, and the separation of the graph neural network cost analyzer is performed to obtain the load transfer area cost analyzer. That is, the extracted first load transfer area line topology is compared with the overall line topology of the area, and the graph neural network cost analyzer containing only the topology of the sub-area is separated from the matched graph neural network cost analyzer, that is, the load transfer area cost analyzer. The graph neural network cost analyzer only analyzes the topology and power supply configuration of the first load transfer area.

[0115] The collaborative power distribution solution of each load transfer area is integrated, and the distributed collaborative power distribution solution is obtained. The distributed collaborative power distribution solution is a whole power supply scheme covering the entire load transfer area, including the first load transfer area collaborative power distribution solution, and also covering the collaborative power distribution solutions of other sub-areas in the load transfer area.

[0116] Further, the distributed collaborative power distribution solution needs to be processed by the load transfer area cost analyzer to generate the first recovery time and the first loss of electric energy. Specifically, the distributed collaborative power distribution solution configured for a certain load transfer area is input into the separated load transfer area cost analyzer. Based on the line topology and power supply scheme of the area, the first recovery time and the first loss of electric energy during the recovery of the area are calculated.

[0117] Then, the line topology of the island key area needs to be compared with the overall line topology by referring to the logic of separating the load transfer area cost analyzer, and the "island key area cost analyzer" suitable for the topology of the island key area is separated from the graph neural network cost analyzer. The grid-forming collaborative power supply solution configured for the island key area, that is, the collaborative output scheme of the grid-forming and grid-following inverter-type power supply, is input into the island key area cost, and the second recovery time and the second loss of electric energy of the area are calculated.

[0118] The first recovery time and the second recovery time are compared, and the larger value is taken as the recovery time of the entire fault recovery process to ensure that both areas are recovered.

[0119] The first loss of electric energy and the second loss of electric energy are compared, and the larger value is taken as the "loss of electric energy" of the whole to ensure that the total loss is within an acceptable range.

[0120] Suppose that after a power distribution network fails, a load transfer area has been delineated: a distributed power supply D1 (capacity 500 kW) and D2 (capacity 300 kW) can be accessed, and the demand for electric power before the failure is 600 kW. The island key area: contains important hospitals and government agencies, and needs to be collaboratively powered by a grid-forming power supply G1 (capacity 1000 kW) and a grid-following power supply G2 (capacity 800 kW), and the demand for electric power before the failure is 1200 kW.

[0121] The distributed cooperative power distribution provides 400 kW for dispatch D1 power supply and 200 kW for dispatch D2 power supply, and supplies power through lines L1-L2-L3. Through the load transfer area cost analyzer, it is calculated that the first recovery time is 8 minutes and the first loss of electric energy is 5.2 kWh.

[0122] The island key area cooperative power supply provides 700 kW for dispatch G1 power supply and 500 kW for dispatch G2 power supply, and supplies power through lines L4-L5-L6. Through the island key area cost analyzer, it is calculated that the second recovery time is 12 minutes and the second loss of electric energy is 8.7 kWh.

[0123] The final determined recovery time is max(8 minutes, 12 minutes)=12 minutes; and the loss of electric energy is max(5.2 kWh, 8.7 kWh)=8.7 kWh.

[0124] The above steps ensure that the fault recovery scheme can meet the recovery requirements of the two areas at the same time, avoid affecting the overall power supply reliability due to the failure of a certain area to meet the standard, and provide a unified quantitative basis for judging whether the scheme is feasible by analyzing the power supply schemes of the two areas and taking the larger value of the more stringent one as the overall evaluation standard.

[0125] In step S300 of the embodiment of the present application, when the recovery time is less than or equal to the time threshold, and when the loss of electric energy is less than or equal to the loss threshold, the distributed cooperative power distribution is used to control the distributed power supply to supply power to the load transfer area, and the cooperative power supply is used to control the grid-connected inverter-type power supply and the grid-following inverter-type power supply to supply power to the island key area.

[0126] This step is to execute corresponding power supply control on the load transfer area and the island key area under the premise that the power supply scheme meets the recovery time and the loss of electric energy threshold,

[0127] The time threshold and the loss threshold are set with “minute-level recovery” as the core target, and exemplarily, the time threshold can be set to ≤60 minutes in reference to the power distribution network fault recovery industry standard and the grid-connected power supply support capability; the loss threshold is combined with the line transmission efficiency, the power output limit and the economic cost to ensure that the loss does not affect the system stability and meets the economic requirements, and can be set to ≤50 kWh, and at the same time, it is compatible with the fluctuation range under extreme working conditions, and balances the recovery speed and energy efficiency.

[0128] In this step, following the example from the previous steps, if the recovery time is 12 minutes or less than 60 minutes, and the energy loss is 8.7 kWh or less than 50 kWh, then in the load transfer area, distributed coordinated power distribution is implemented, controlling the output of distributed power source D1 to 400 kW and D2 to 200 kW, and closing the tie switch L1-2 to supply power to the area; in the island critical area, grid-connected coordinated power supply is implemented, controlling the output of grid-connected power source G1 to 700 kW and the output of grid-connected power source G2 to 500 kW, and closing the sectionalizing switch L4-5 to supply power to the area.

[0129] Power was eventually restored to both areas, completing the fault recovery.

[0130] In step S300 of this application embodiment, when the recovery time is greater than the time threshold, or / and when the power loss is greater than the loss threshold, the distributed coordinated power distribution solution and the grid-connected coordinated power supply solution are updated, and the loop is executed.

[0131] If the preset number of cycles does not converge, extract a power restoration scheme where the energy loss is less than or equal to the energy loss threshold.

[0132] If the power restoration scheme is not empty, extract the scheme with the shortest restoration time from the power restoration schemes and provide power;

[0133] When the power restoration scheme is empty, the normalized value of the restoration time of each power supply scheme is weighted based on the predefined time weight to obtain multiple first weighted features;

[0134] Based on predefined loss weights, the normalized values ​​of loss energy for each power supply scheme are weighted to obtain multiple second weighted features;

[0135] By weighting the one-to-one corresponding plurality of first weighted features and plurality of second weighted features, the fitness of multiple power supply schemes is obtained;

[0136] Extract the power supply scheme corresponding to the minimum fitness value among the multiple power supply schemes and supply power accordingly.

[0137] Wherein, the duration weight ∈ [0.8, 1], and the loss weight ∈ [0, 0.2].

[0138] In this embodiment of the application, the purpose of the above steps is to gradually approach the optimal solution that satisfies the threshold by iteratively updating the power supply scheme when the recovery time or power loss exceeds the threshold.

[0139] For example, assuming that the initial power supply scheme for the load transfer area and the islanded critical area does not meet the threshold, namely the aforementioned duration threshold of 60 minutes and loss threshold of 50kWh, the particle swarm algorithm can be used to update the distributed cooperative distribution solution and the grid-connected cooperative power supply solution, setting the number of particles to 5 and the number of iterations to 3.

[0140] Assume that the initial scheme parameters are: the load transfer area) distributed power output is D1=300kW, D2=200kW. The island key area network type / follow network type power output is G1=500kW, G2=400kW. The recovery time is 70 minutes and the loss of electric energy is 55kWh obtained by the method in the above steps.

[0141] Then, the particle coding needs to be performed first, and each particle represents a set of power output scheme (D1, D2, G1, G2). The initial particle group is generated, such as: particle 1 (350, 250, 550, 450); particle 2 (400, 200, 600, 500); particle 3 (320, 280, 520, 480); particle 4 (380, 220, 580, 420); particle 5 (420, 180, 620, 520).

[0142] Then, the particle fitness corresponding to each particle needs to be evaluated. The specific calculation method can be the recovery time value + the loss of electric energy value. The recovery time value and the loss of electric energy value can be obtained by the particle represented parameters from the aforementioned graph neural network cost analyzer, and the values in units of minutes and kilowatt-hours are not described here. The particle fitness is calculated as follows: particle 1 recovery time value 65, loss of electric energy value 52, particle fitness 117; particle 2 recovery time value 58, loss of electric energy value 49, particle fitness 107, which is the optimal individual; particle 3 recovery time value 68, loss of electric energy value 53, particle fitness 121; particle 4 recovery time value 62 minutes, loss of electric energy value 50, particle fitness 112; particle 5 recovery time value 60, loss of electric energy value 51, particle fitness 111.

[0143] By comparing the particle fitness, the global optimal particle is particle 2, which guides other particles to approach it.

[0144] After the second iteration, particle 2 is updated to (410, 210, 610, 510). As a result, the recovery time is 55 minutes, the loss of electric energy is 48kWh, and the particle fitness is 103.

[0145] After the third iteration, the result of particle 2 is that the recovery time is 55 minutes (≤60), the loss of electric energy is 48kWh (≤50), and the threshold is met. Then the updated power supply scheme is: the load transfer area: D1=410kW, D2=210kW, distributed collaborative power supply and control; the island key area: G1=610kW, G2=510kW, network collaborative power supply and control.

[0146] Suppose that after 3 iterations, no power supply scheme meets the time length ≤ 60 minutes and the loss ≤ 50 kWh at the same time, and the loss of the extracted power energy is less than or equal to the loss threshold of the recovery power supply scheme, that is, the power supply scheme with smaller loss of power energy is preferentially selected.

[0147] When the recovery power supply scheme is not empty, the recovery time length of the shortest scheme in the recovery power supply scheme is extracted for power supply, that is, when there are multiple recovery power supply schemes, the recovery time length of the shortest scheme in the recovery power supply scheme is preferentially selected for power supply, so as to shorten the time of power failure as much as possible.

[0148] When the recovery power supply scheme is empty, the recovery time length of each power supply scheme is normalized and weighted based on the predefined time length weight to obtain a plurality of first weighted features; the loss of power energy of each power supply scheme is weighted based on the predefined loss weight to obtain a plurality of second weighted features; the plurality of first weighted features and the plurality of second weighted features are weighted one by one to obtain a plurality of power supply scheme fitness; the power supply scheme corresponding to the minimum value of the plurality of power supply scheme fitness is extracted for power supply.

[0149] Specifically, the recovery time length of the plurality of power supply schemes can be normalized first, and the recovery time length normalization value = (current recovery time length value - minimum recovery time length value) / (maximum recovery time length value - minimum recovery time length value), and then the loss of power energy of the plurality of power supply schemes is normalized, and the loss of power energy normalization value = (current loss of power energy value - minimum loss of power energy value) / (maximum loss of power energy value - minimum loss of power energy value).

[0150] Suppose that in the plurality of power supply schemes, the maximum recovery time length value is 70 minutes, the minimum recovery time length value is 55 minutes, the recovery time length threshold is 60 minutes; the maximum loss of power energy value is 55 kWh, the minimum loss of power energy value is 51 kWh, and the loss of power energy threshold is 50 kWh.

[0151] Suppose there are 3 power supply schemes, if the recovery time length of scheme A is 65 minutes, then the recovery time length normalization value = (65-55) / (70-55) = 10 / 15 ≈ 0.67, and by analogy, the recovery time length normalization value of scheme B is 0.2 when the recovery time length is 58 minutes; the recovery time length normalization value of scheme C is 1.0 when the recovery time length is 70 minutes.

[0152] If the loss of power energy of scheme A is 53 kWh, then the loss of power energy normalization value = (53-51) / (55-51) = 2 / 4 = 0.5. By analogy, the loss of power energy normalization value of scheme B is 1.0 when the loss of power energy is 55 kWh, and the loss of power energy normalization value of scheme C is 0.0 when the loss of power energy is 51 kWh.

[0153] According to the urgency of the power demand of the region, the time length weight can be set as 0.9 and the loss weight as 0.1. Then, the scheme A power supply scheme fitness = 0.67*0.9+0.5*0.1 = 0.603+0.05 = 0.653; the scheme B power supply scheme fitness = 0.2*0.9+1.0*0.1 = 0.18+0.1 = 0.28; and the scheme C power supply scheme fitness = 1.0*0.9+0.0*0.1 = 0.9.

[0154] Then, the scheme B with the minimum weighted feature is selected to supply power.

[0155] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0156] Those skilled in the art should understand that the embodiments of the present application can provide a method, a system or a computer program product. Therefore, the present application can adopt a completely hardware embodiment, a completely software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes.

[0157] The present application is described with reference to flowcharts and / or block diagrams of the method, device (system) and computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams can be realized by computer program instructions, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded computer or other programmable data processing devices to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices produce a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for performing the functions specified in one or more flows and / or blocks.

[0158] These computer program instructions can also be stored in a computer readable storage medium capable of guiding the computer or other programmable data processing devices to work in a specific way, so that the instructions stored in the computer readable storage medium produce a product including instruction means, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 means for performing the functions specified in one or more flows and / or blocks.

[0159] These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operational steps are performed on the computer or other programmable devices to generate computer-implemented processes, thus the instructions executed on the computer or other programmable devices provide the function for implementing the processes specified in the flowchart Figure 1 one or more flows and / or blocks Figure 1 one or more blocks or steps of the functions specified in the flowchart

[0160] Although the preferred embodiments of the application have been described, those skilled in the art will be able to make additional modifications and variations to these embodiments without departing from the spirit and scope of the application.

[0161] Obviously, various modifications and changes can be made to the present application by those skilled in the art without departing from the spirit and scope of the present application. Accordingly, it is intended that the present application embrace all such modifications and changes as fall within the scope of the present application and its equivalents.

Claims

1. A power distribution network minute-level collaborative fault recovery method for network-oriented inverter power supply, characterized in that, Comprise: When the power grid failure trigger, extract load transfer area and island key area, wherein the island key area is the power grid outage area not covered by the load transfer area; The load transfer area is configured with a distributed collaborative power distribution solution, and the island key area is configured with a grid-connected collaborative power supply solution. Through the graph neural network cost analyzer, the recovery time and the loss of electric energy are generated; When the recovery time is less than or equal to the time threshold, and the loss of electric energy is less than or equal to the loss threshold, the distributed power supply is controlled by the distributed collaborative power distribution solution to supply power to the load transfer area, and the grid-connected inverter-type power supply and the grid-connected inverter-type power supply are controlled by the grid-connected collaborative power supply solution to supply power to the island key area; Wherein, the load transfer area is configured with a distributed collaborative power distribution solution, and the island key area is configured with a grid-connected collaborative power supply solution. Through the graph neural network cost analyzer, the recovery time and the loss of electric energy are generated, comprising: According to the power grid fault area, the graph neural network cost analyzer is matched from the graph neural network cost analyzer set, wherein the graph neural network cost analyzer has a label identifying the topology of all lines of the power grid fault area; Based on the load transfer area line topology, compared with all line topologies, the graph neural network cost analyzer is separated to obtain the load transfer area cost analyzer, and the distributed collaborative power distribution solution is processed to generate the first recovery time and the first loss of electric energy; Based on the island key area line topology, compared with all line topologies, the graph neural network cost analyzer is separated to obtain the island key area cost analyzer, and the grid-connected collaborative power supply solution is processed to generate the second recovery time and the second loss of electric energy; The greater value of the first recovery time and the second recovery time is set as the recovery time; The greater value of the first loss of electric energy and the second loss of electric energy is set as the loss of electric energy.

2. The method of claim 1, wherein, When the power grid failure trigger, extract load transfer area and island key area, comprising: When the grid-connected point voltage is less than the voltage threshold, it is considered that the power grid failure trigger, and the power grid outage area is received; Based on the power grid outage area line topology graph, the distributed power supply connected to the power grid outage area is extracted; The load transfer area of the power grid outage area is determined by traversing the distributed power supply and performing power flow calculation; Extract the area not covered by the load transfer area in the power grid outage area, and set it as the initial island key area; The island key area is obtained by traversing the initial island key area and performing source and load balance analysis.

3. The method of claim 2, wherein, After the grid-connected point voltage is less than the voltage threshold, before the power grid outage area is received, the current overload capacity is adjusted to a preset multiple of the rated current within k seconds, wherein 0≤k≤2.

4. The method of claim 1, wherein, Based on the load transfer area line topology, compared with all line topologies, the graph neural network cost analyzer is separated to obtain the load transfer area cost analyzer, comprising: From the load transfer area, the first load transfer area is extracted, wherein the first load transfer area has an accessible distributed power supply bit number identification; According to the accessible distributed power supply bit number identification, the first load transfer area demand power before failure is taken as the target, the first load transfer area is configured for power supply, and a first load transfer area collaborative power distribution solution is obtained; According to the accessible distributed power supply bit number identification and the first load transfer area, a first load transfer area line topology is extracted from a power grid outage area line topology graph; Based on the first load transfer area line topology, compared with all line topologies, a load transfer area cost analyzer is obtained by executing separation on the graph neural network cost analyzer.

5. The method of claim 1, wherein, The construction steps of the graph neural network cost analyzer set include: Step one: the PQ unit line topology enumeration disassembly is performed on the target area line topology, a plurality of PQ unit line topologies are obtained, the initial value of P is equal to 1, the initial value of Q is equal to 1, the PQ unit line topology includes Q power supply nodes and P load nodes, P and Q are integers, P≤total number of load nodes, Q≤total number of power supply nodes; Step two: the plurality of PQ unit line topologies are traversed, the power supply distribution position is taken as the input node, and the load distribution position is taken as the output node, the graph neural network simulation is performed, and a plurality of PQ unit line graph neural network architectures are constructed; Step three: the power supply capacity is taken as the input data, the power supply duration and the loss of electric energy are taken as the output data, the plurality of PQ unit line graph neural network architectures are trained, and a plurality of PQ unit line cost analyzers are generated; Step four: when P<total number of load nodes, P is increased by one, and the loop is returned to step one: When Q<total number of power supply nodes, Q is increased by one, and the loop is returned to step one; When Q=total number of power supply nodes and P<total number of load nodes, P is increased by one, and the loop is returned to step one; Step five: when P=total number of load nodes and Q<total number of power supply nodes, Q is increased by one, and the loop is returned to step one; Step six: when Q=total number of power supply nodes and P=total number of load nodes, all PQ unit line cost analyzers are output, are stored in association with the target area, and are added to the graph neural network cost analyzer.

6. The method of claim 1, wherein, Further comprising: When the recovery duration is greater than the duration threshold, or / and when the loss of electric energy is greater than the loss threshold, the distributed collaborative power distribution solution and the construction network collaborative power distribution solution are updated, and the loop is executed.

7. The method of claim 6, wherein, Further comprising: When the loop does not converge for a preset number of times, a recovery power supply scheme in which the loss of electric energy is less than or equal to the loss threshold is extracted; When the recovery power supply scheme is not empty, a recovery duration shortest scheme in the recovery power supply scheme is extracted for power supply; When the recovery power supply scheme is empty, a plurality of first weighted features are obtained by weighting the recovery duration normalized values of each power supply scheme based on a predefined duration weight; A plurality of second weighted features are obtained by weighting the loss of electric energy normalized values of each power supply scheme based on a predefined loss weight; The plurality of first weighted features and the plurality of second weighted features are weighted one by one to obtain a plurality of power supply scheme fitnesses; The power supply scheme corresponding to the minimum value of the plurality of power supply scheme fitnesses is extracted for power supply.

8. The method of claim 7, wherein, The duration weight ∈ [0.8, 1], and the loss weight ∈ [0, 0.2].

9. The method of claim 5, wherein, The PQ unit line topology includes Q power supply nodes and P load nodes, and further includes that the P load nodes can perform power supply through the Q power supply nodes.

Images