A kind of high toughness adaptive topology reconfiguration method of rural distribution network fusing real-time monitoring and dynamic correction

Abstract

The application discloses a kind of fusion real-time monitoring and dynamic correction's rural distribution network high toughness adaptive topology reconfiguration method, comprising: real-time monitoring obtains power grid load and distributed power output data;According to data characteristics, distribution network topology is dynamically divided into single radiation basic type, loop network enhancement type and microgrid autonomy type three kinds and model is established;Multi-objective fitness function containing network loss, voltage qualification rate and key load reliability is constructed, and initial optimal topology is obtained by using improved genetic algorithm solution;Super-short-term distributed power output is predicted using graph neural network, and the initial topology strategy is dynamically corrected accordingly;Finally, the optimal topology is quickly executed by automation terminal.The application solves the problems of insufficient dynamicity of rural distribution network topology reconfiguration, difficulty of multi-objective coordination and poor adaptability of new energy, significantly improves the operation toughness and power supply safety of rural distribution network under high proportion of new energy access.

Description

Technical Field

[0001] This invention relates to the field of power systems and their automation technology, specifically to a highly resilient adaptive topology reconfiguration method for rural distribution networks that integrates real-time monitoring and dynamic correction. Background Technology

[0002] The rural energy structure is rapidly transitioning towards clean and low-carbon energy, with the penetration rate of distributed photovoltaic and wind power and other new energy sources in rural power distribution networks continuing to rise. These new energy sources are not only a key pathway for rural areas to achieve "energy self-sufficiency + low-carbon emission reduction," but also directly support core rural production activities such as irrigation, livestock breeding, and agricultural product processing. According to statistics from the State Grid Hubei Electric Power Company, there are approximately 19.5 million households in rural areas of Hubei Province currently connected to the rural power grid, covering 83% of the province's total area. The rural power distribution network has become a core carrier for the local consumption of new energy and ensuring stable electricity supply for rural production and daily life; its operational resilience is directly related to farmers' livelihoods and rural economic development.

[0003] However, the operation and topology optimization of rural power distribution networks still face three core problems. These problems are deeply tied to the unique characteristics of rural power grids, directly restricting their resilience and economic efficiency, and significantly impacting rural production and daily life:

[0004] First, the rigid topology and low redundancy make it difficult to adapt to the strong fluctuations in rural loads. Rural loads have typical "seasonal and time-dependent" characteristics—irrigation and drainage equipment are activated intensively during the busy farming season, and the load can double compared to the off-season; peak electricity consumption by farmers during the day alternates with peak electricity consumption for processing at night. However, more than 60% of rural power distribution networks still use a single radial topology. This structure has no backup path and low redundancy. When irrigation loads surge during the busy farming season, it cannot quickly distribute power, often leading to line overload tripping, directly affecting the progress of farmland irrigation, and even causing crop yield reduction.

[0005] Secondly, the high proportion of distributed power sources has poor adaptability and is out of touch with rural renewable energy utilization scenarios. Rural areas often have abundant rooftop and wasteland resources, leading to concentrated grid connection of distributed photovoltaic and small-scale wind power. However, existing topology reconfiguration technologies have significant shortcomings in adaptability: on the one hand, there is a "spatiotemporal mismatch" between the peak photovoltaic output at midday and the peak electricity consumption for rural living / processing in the evening, which can easily lead to voltage exceeding limits or renewable energy curtailment; on the other hand, existing technologies mostly use static planning models, which do not consider the strong uncertainty of rural distributed power output affected by weather, and cannot cope with the output fluctuations of photovoltaic and wind power, which reduces the renewable energy absorption rate and threatens the stability of the power grid.

[0006] Third, the reconfiguration targets are singular and the response is slow, neglecting the needs of ensuring critical loads in rural areas. In rural power distribution networks, irrigation pumping stations, livestock and poultry breeding temperature control equipment, and emergency lighting are considered "critical loads"—interruptions to these loads directly lead to agricultural production losses. However, existing technologies mostly focus on the single objective of "minimizing network losses" and have not established a priority mechanism for critical load power supply; moreover, traditional distribution network reconfiguration strategies rely on manual experience or offline calculations, with reconfiguration delays generally exceeding 10 minutes, far from being able to cope with the second-level fluctuations in the output of rural distributed power sources, making it difficult to ensure continuous power supply to critical production loads.

[0007] To address the aforementioned issues, existing solutions have significant limitations and are difficult to adapt to the actual needs of rural areas: In terms of hardware upgrades, some solutions require the addition of a large number of power distribution equipment and lines, resulting in high costs. However, rural areas have a wide power supply coverage and a weak economic foundation, making it difficult to achieve large-scale promotion. In terms of software control, traditional intelligent algorithms such as particle swarm optimization (PSO) and single genetic algorithms are prone to getting trapped in local optima and do not take into account the characteristics of rural power distribution networks, such as "strong spatiotemporal heterogeneity of load and sparse measurement equipment." As a result, the optimization results are seriously out of touch with the actual working conditions in rural areas and cannot be applied in practice.

[0008] Therefore, the inventors of this application provide a highly resilient adaptive topology reconfiguration method for rural power distribution networks that integrates real-time monitoring and dynamic correction. Through closed-loop control of "real-time monitoring - dynamic modeling - intelligent optimization - rapid execution", it solves the problems of insufficient resilience and poor adaptability of existing technologies, and provides technical support for the construction of a new rural energy system. Summary of the Invention

[0009] This invention provides a highly resilient adaptive topology reconfiguration method for rural distribution networks that integrates real-time monitoring and dynamic correction. It aims to solve the technical problems of insufficient dynamics in existing rural distribution network topology reconfiguration, lack of multi-objective coordination, poor adaptability of distributed power sources, and slow response, so as to achieve improved resilience and safe and economical operation of rural distribution networks under high-proportion renewable energy access.

[0010] This invention provides a highly resilient adaptive topology reconfiguration method for rural power distribution networks that integrates real-time monitoring and dynamic correction, comprising the following steps:

[0011] Real-time monitoring technology is used to obtain load status and distributed power output data of rural power distribution networks;

[0012] Based on the load status and distributed generation output data, combined with the seasonal and temporal characteristics of the load and the penetration rate of distributed generation, the distribution network topology is divided into three types: single-radial basic type, ring network enhanced type, and microgrid autonomous type. A dynamic model of the distribution network including the dynamic load of nodes and the fluctuation range of distributed generation output is established.

[0013] Based on the dynamic model, a multi-objective fitness function including network loss, voltage qualification rate, and critical load reliability is constructed. An improved genetic algorithm is used to solve the multi-objective fitness function to determine the initial optimal topology switch state.

[0014] Based on the graph neural network model, the ultra-short-term output of distributed power sources is predicted, and the initial optimal topology switch state is dynamically corrected based on the prediction results to obtain the corrected optimal topology switch state.

[0015] The modified optimal topology switch state is executed through the distribution network automation terminal, so that the distribution network topology is adaptively adjusted to maintain the resilience index within the set range.

[0016] Furthermore, the acquisition of load status and distributed power output data of rural power distribution networks using real-time monitoring technology specifically includes:

[0017] Collect real-time load values, voltage values, and distributed generation output power of each node in the distribution network, and calculate the node load fluctuation coefficient and the distributed generation output fluctuation range.

[0018] Furthermore, the single-radial basic topology is suitable for remote rural transformer substations with low load density and distributed power penetration rate <15%; the ring network enhanced topology is suitable for township transformer substations with medium load density and distributed power penetration rate of 15%-30%; and the microgrid autonomous topology is suitable for new energy demonstration villages with distributed power penetration rate >30%.

[0019] Furthermore, based on the load status and distributed generation output data, combined with the seasonality and time-specific characteristics of the load and the penetration rate of distributed generation, the distribution network topology is divided into three categories: single-radial basic type, ring network enhanced type, and microgrid autonomous type, and a dynamic model is established, specifically including:

[0020] The dynamic load of nodes is determined based on seasonal coefficients, time-period coefficients, and baseline loads, providing dynamic load parameters for power flow calculations and voltage threshold settings in multi-objective optimization.

[0021] The output range of distributed power sources is determined based on the ultra-short-term predicted output and fluctuation coefficient, providing boundary basis for the constraint conditions and reconfiguration correction triggering judgment of multi-objective optimization algorithms;

[0022] By combining topology types, a distribution network model incorporating the dynamic characteristics of loads and distributed generation is constructed, providing complete model support for the construction and reconstruction of the adjacency matrix predicted by graph neural networks and the verification of the execution effect.

[0023] Furthermore, the improved genetic algorithm uses integer encoding to represent the switch state and optimizes the population evolution process through elite retention selection, connectivity constraint crossover, and dynamic mutation probability.

[0024] 5. Furthermore, the step of using an improved genetic algorithm to solve the fitness function and determine the optimal topology switch state specifically includes:

[0025] Using the topology switch state as a genetic variable, a fitness function is constructed that includes network loss optimization terms, voltage qualification rate optimization terms, and critical load reliability optimization terms.

[0026] Initialize the population and set the iteration parameters, then calculate the fitness value of the initial population;

[0027] Select high-quality individuals using an elite retention strategy, perform crossover operations on the selected individuals, and then perform connectivity checks.

[0028] Dynamic mutation probability is used to perform mutation operations on individuals to avoid premature convergence of the algorithm.

[0029] Iteratively update the individual optimal values ​​and the global optimal value of the population until the convergence condition is met or the maximum number of iterations is reached;

[0030] The switch state corresponding to the global optimal solution is taken as the optimal topology switch state.

[0031] Furthermore, in the fitness function, the weights of the network loss optimization term, the voltage qualification rate optimization term, and the critical load reliability optimization term are 0.4, 0.3, and 0.3, respectively.

[0032] Furthermore, the dynamic mutation probability ranges from 0.01 to 0.05, and gradually decreases as the number of iterations increases.

[0033] Furthermore, the graph neural network model takes the distribution network topology coupling model, dynamic load sequence, historical meteorological data, and distributed generation output data as inputs. It extracts spatial correlation features through graph convolutional layers and outputs ultra-short-term predicted output of distributed generation within 15 minutes, with a prediction error of ≤8%. This provides a basis for the dynamic correction of distributed generation output fluctuations in subsequent topology reconfiguration strategies. The topology and load parameters provided by the dynamic model are used to enhance the graph neural network's ability to capture the coupling relationship between 'topology-load-distributed generation'.

[0034] Furthermore, the distribution network automation terminal includes a feeder terminal unit (FTU) and a distribution terminal unit (DTU), the reconfiguration strategy execution delay is ≤2 minutes, the critical loads include rural irrigation equipment and livestock and poultry breeding equipment; the resilience indicators include voltage qualification rate ≥98%, critical load reliability 100%, and network loss reduction rate ≥8%.

[0035] In summary, this invention employs real-time monitoring technology to accurately acquire load and distributed power generation data of rural power distribution networks; classifies topology types based on data characteristics and establishes dynamic models to adapt to the spatiotemporal heterogeneity of rural power grids; uses an improved genetic algorithm to solve multi-objective fitness functions to balance network losses, voltage quality, and critical load reliability; combines a graph neural network model to achieve ultra-short-term output prediction and dynamic strategy correction for distributed power sources, adapting to renewable energy fluctuations; and finally, rapidly executes the optimal topology through automated terminals, maintaining the resilience indicators of the power distribution network (voltage qualification rate ≥98%, critical load reliability 100%, network loss reduction rate ≥8%) within the set range. This allows for adaptive adjustment of the rural power distribution network topology, significantly improving the grid's anti-interference capability and power supply security under high-proportion renewable energy access. Furthermore, it relies on existing equipment, is low-cost, and possesses stronger adaptability to rural scenarios and greater engineering promotion value. Attached Figure Description

[0036] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a flowchart illustrating a highly resilient adaptive topology reconfiguration method for rural power distribution networks that integrates real-time monitoring and dynamic correction, according to an embodiment of the present invention.

[0038] Figure 2 This is the initial topology diagram of the power distribution network;

[0039] Figure 3 This is a diagram of the architecture of a real-time monitoring system for power distribution networks;

[0040] Figure 4 This is the convergence curve of the improved genetic algorithm;

[0041] Figure 5 This is a diagram of a graph neural network model architecture;

[0042] Figure 6 It is a graph comparing node voltages before and after the distribution network topology reconfiguration. Detailed Implementation

[0043] The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. The following detailed description of the embodiments and the accompanying drawings are used to illustrate the principles of the present invention by way of example, but should not be used to limit the scope of the present invention. That is, the present invention is not limited to the described embodiments, and covers any modifications, substitutions and improvements to parameter settings, model optimization and execution flow without departing from the spirit of the present invention.

[0044] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0045] Figure 1 This is a flowchart illustrating a highly resilient adaptive topology reconfiguration method for rural power distribution networks that integrates real-time monitoring and dynamic correction, as provided in an embodiment of the present invention. Figure 1 As shown, the method may include the following steps:

[0046] S100. Real-time monitoring technology is used to obtain load status and distributed power output data of rural power distribution networks;

[0047] S200. Based on the load status and distributed power output data, combined with the seasonal and temporal characteristics of the load and the penetration rate of distributed power, the distribution network topology is divided into three types: single-radial basic type, ring network enhanced type, and microgrid autonomous type, and a distribution network dynamic model including the dynamic load of nodes and the fluctuation range of distributed power output is established.

[0048] S300. Based on the dynamic model, construct a multi-objective fitness function that includes network loss, voltage qualification rate, and critical load reliability. Use an improved genetic algorithm to solve the multi-objective fitness function and determine the initial optimal topology switch state.

[0049] S400. Based on the graph neural network model, the distributed power source is predicted for ultra-short-term output. The initial optimal topology switch state is dynamically corrected based on the prediction results to obtain the corrected optimal topology switch state.

[0050] S500. The modified optimal topology switch state is executed through the distribution network automation terminal to enable the distribution network topology to adaptively adjust in order to maintain the resilience index within the set range.

[0051] In the above implementation, the dynamic characteristics of rural power grids are captured in real time, avoiding the lag of traditional static modeling; the combination of topology classification and dynamic model adaptation to the heterogeneity of rural scenarios solves the problem of adapting load and distributed power source fluctuations; the genetic algorithm is improved to balance multi-objective requirements and overcome the limitations of single-objective optimization; graph neural network prediction and dynamic correction ensure the timeliness of the strategy and adapt to the high proportion of new energy access; finally, through rapid execution by automated terminals, a closed loop of "monitoring-modeling-optimization-execution" for topology reconstruction is realized, significantly improving the resilience of rural distribution networks.

[0052] Based on the established dynamic model and multi-objective optimization framework, and by adopting a collaborative mechanism of improved genetic algorithm and graph neural network prediction, a topology reconfiguration scheme adapted to rural working conditions can be accurately output, so that the resilience indicators such as voltage qualification rate, critical load reliability, and network loss are maintained within the set range, thus meeting the safe and economical operation requirements of rural power grids under high-proportion renewable energy access.

[0053] As an optional implementation method, this embodiment is based on a typical single-radial base type distribution network topology, such as... Figure 2 As shown, it includes 33 nodes, 8 distribution transformers (total capacity 700kVA), and is connected to 2 distributed power sources: 12 nodes of 600kW wind power and 25 nodes of 400kW photovoltaic power. The critical loads are 16 nodes of irrigation pumping station (rated load 200kW) and 33 nodes of livestock and poultry farm (rated load 150kW).

[0054] The core parameters and applicable scenarios for the three topology types are as follows:

[0055] Single-radial basic type: suitable for remote rural areas with load density ≤0.3kW / mu and distributed power penetration rate <15%, including 1 main line (conductor type JKLYJ-240) + 4 branch lines (conductor type JKLYJ-70), and the total capacity of the distribution transformer ≤500kVA. The initial topology in this embodiment is of this type.

[0056] Ring network enhanced type: suitable for township transformer areas with load density of 0.3-0.8kW / mu and distributed power penetration rate of 15%-30%, including 2 backup main lines + 6-8 branch lines, with a total distribution transformer capacity of 500-1000kVA, and can achieve rapid fault switching through the interconnection of branch lines;

[0057] Microgrid autonomous type: suitable for new energy demonstration villages with load density > 0.8kW / mu and distributed power penetration rate > 30%, including 1 microgrid control unit (response latency ≤ 500ms) + 2-3 sets of distributed power-storage clusters (storage capacity ≥ 1MWh), supporting off-grid autonomous operation.

[0058] The selection of topology type should be determined based on the actual data of the distribution area: if the average proportion of distributed power output to load is less than 15% for 7 consecutive days and the load density is low, then a single-radial basic type should be adopted; if the proportion is 15%-30% and the load fluctuation coefficient (busy / slack season) is greater than 2.0, then a ring network enhanced type should be upgraded; if the proportion is greater than 30% and contains more than 2 critical loads, then a microgrid autonomous type should be adopted.

[0059] As an optional implementation, step S100 employs a hierarchical monitoring architecture of "FTU / DTU + edge-aware terminal", such as... Figure 3As shown. The data collection frequency is set to once per minute to ensure data real-time performance and integrity. The specific process is as follows:

[0060] Data collection scope: Current, voltage, and power factor data of the main line and branch lines are collected through 12 FTUs (feeder terminal units) deployed in the distribution area; load values ​​and distributed power output of each node are collected through 25 DTUs (distribution terminal units); real-time data (irradiance and wind speed, with sampling accuracy of 1W / m² and 0.1m / s, respectively) from Badong County Meteorological Station are accessed simultaneously.

[0061] Data preprocessing: Outliers were removed from the collected data using the 3σ criterion, and then missing values ​​were filled in using linear interpolation.

[0062] Data storage and transmission: Local data storage (30-day storage period) is achieved through edge computing units, and core data is uploaded to the regional new power system simulation platform through a 5G private network for subsequent modeling and verification.

[0063] The monitoring system architecture diagram is as follows: Figure 3 As shown, this architecture, through a three-level model of "terminal acquisition - edge processing - platform aggregation", can achieve accurate perception of the "last mile" of rural power distribution networks, with a data effectiveness rate of ≥95%, providing reliable input for subsequent dynamic modeling.

[0064] As an optional implementation, step S200 involves establishing a dynamic model by combining monitoring data with the characteristics of the rural power grid. Specifically, this includes the following steps:

[0065] S201. Introduce the seasonal coefficient ks and the time period coefficient kt, and correct the nodal baseline load P through their product relationship. L0 (i), the formula is as follows:

[0066]

[0067] The time period coefficient kt is divided into time periods: 1.1 for morning peak, 0.7 for midday peak, 1.3 for evening peak, and 0.5 for trough. In this example, the evening peak period in March (19:00-20:00) is selected. Therefore, the dynamic load of node 16 is 0.7×1.3×200=182kW, and that of node 33 is 0.7×1.3×150=136.5kW, with a deviation of ≤3% from the actual monitored value.

[0068] S202, Distributed power generation ultra-short-term output P based on graph neural network prediction 分布式电源 (j,t) (prediction duration 15 minutes), combined with volatility coefficient Constructing the output range:

[0069]

[0070] In this embodiment, the predicted output of the 25-node photovoltaic system is 320kW, with a fluctuation coefficient of ±20%, resulting in an output range of [256, 384]kW; the predicted output of the 12-node wind power system is 350kW, with a fluctuation coefficient of ±30%, resulting in an output range of [245, 455]kW. If the actual output exceeds this range (e.g., the actual photovoltaic output is 240kW < 256kW), a subsequent reconfiguration correction process is triggered.

[0071] As an optional implementation, the improved genetic algorithm in step S300 specifically includes the following steps:

[0072] S301, the multi-objective fitness function is the weighted sum of the network loss optimization term f1, the voltage qualification rate optimization term f2, and the critical load reliability optimization term f3:

[0073]

[0074] in:

[0075] ( (Initial topology network loss 195.10kW).

[0076] (U) qual (X) represents the voltage qualification rate, with a minimum threshold of 80%).

[0077] (Critical load power supply guarantee rate, 1 indicates full guarantee).

[0078] The encoding uses integer encoding, the chromosome length is equal to the number of segment switches, the gene value 0 indicates that the switch is open and 1 indicates that it is closed, and the encoding must satisfy the "acyclic connectivity" constraint.

[0079] S302, Algorithm parameters are: population size 50, maximum number of iterations 50, dynamic mutation probability. (t is the current iteration number), elite retention rate is 10%;

[0080] S303, The algorithm iteration steps are as follows:

[0081] (1) Initialization: 50 initial populations are randomly generated, and the individual optimal value pbest and the global optimal value gbest are initialized to 0;

[0082] (2) Fitness calculation: Substitute each particle (on / off state combination) into the fitness function;

[0083] (3) Elite selection: The top 10% of individuals with high fitness (fitness value ≥ 0.72) are retained, and the remaining individuals are selected according to their fitness percentage (selection probability). );

[0084] (4) Crossover operation: The real number crossover method is used to crossover the selected individuals in position;

[0085] (5) Mutation operation: In the early stage of iteration (t<20) p m =0.05 (enhanced global search), later iterations (t≥30) p m =0.01 (Focusing on local optimization);

[0086] (6) Update and Termination: pbest and gbest are updated every iteration. Termination occurs when the fluctuation of gbest is ≤0.5% for 5 consecutive iterations or when 50 iterations have been reached. In this embodiment, convergence occurs in the 38th iteration, with gbest=0.92. The convergence curve is shown below. Figure 4 As shown.

[0087] As an optional implementation, step S400 involves the specific implementation of graph neural network prediction and dynamic correction.

[0088] S401, Graph Neural Network Model Construction and Training

[0089] Graph neural network models employ a "Graph Convolutional Layer (GCN) + Fully Connected Layer" architecture, such as... Figure 5 As shown, the details are as follows:

[0090] ① Input Layer: The adjacency matrix of the distribution network topology is 33×33, where 1 indicates node connectivity and 0 indicates disconnection. Figure 2 Topology generation; the node feature matrix is ​​33×5 in size, and the features include distributed power source type, historical output, irradiance, wind speed, and load type; the first 72 hours of time series data, sampled at a frequency of 15 minutes / time, for a total of 288 sets of data.

[0091] ② Network structure: 2 layers of graph convolutional layers, where the hidden layer has a dimension of 64 and uses the ReLU activation function to extract the spatial correlation of "node-edge-feature"; 2 layers of fully connected layers, where the output has a dimension of 1 and is mapped to the predicted output value of the distributed power source 15 minutes later.

[0092] ③ Training process: The Adam optimizer was used with a learning rate of 0.001, a batch size of 32, 200 iterations, and the loss function was mean squared error (MSE). The ratio of training set, validation set, and test set was 7:2:1. The prediction error of the test set was 7.6% for photovoltaic power and 7.9% for wind power power, which met the accuracy requirements.

[0093] S402. Real-time comparison of actual output of distributed power sources with the predicted range. Under normal scenarios, if the actual output of photovoltaic power is within the range of [256,384] kW and the wind power is within the range of [245,455] kW, the optimal topology of the output is maintained; under corrected scenarios, if the actual output of photovoltaic power is lower than the lower limit of the range of 256 kW, a reconstruction correction is triggered, and the improved genetic algorithm is re-executed.

[0094] As an optional implementation, step S500 involves the execution and verification of the reconstruction strategy. Specifically, it includes the following steps:

[0095] S501, Execution process: The optimal switch state is converted into a control command and sent to the FTU / DTU terminal via the 5G private network. The terminal drives the segmented switch action. The entire execution process takes 1.8 minutes (including command transmission, switch action, and status feedback).

[0096] S502, Effect Verification:

[0097] (1) Real-time monitoring: After execution, the node voltage ranged from a maximum of 1.04 pu to a minimum of 0.935 pu. Figure 6 As shown by the dashed line, all values ​​are within the allowable range; the maximum line load rate is 78%, which is below the 120% threshold.

[0098] (2) Resilience indicators: voltage qualification rate 98.5%, critical load reliability 100%, network loss reduction rate 23.8%, exceeding the target of 8%;

[0099] (3) Extreme scenario test: Simulate a fault in the main line 5, and the reconfiguration strategy completes the switch to the backup path within 2.1 minutes, without interrupting the power supply to the critical load.

[0100] Figure 5 The voltage comparison curves before and after reconstruction show that the voltage fluctuation is significantly reduced after reconstruction, and all of them fall within the range of [0.93, 1.07] pu, which verifies the effectiveness of the method of the present invention.

[0101] It should be clarified that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. This invention is not limited to the specific parameters and processes described above and shown in the figures. For example, the topology type can be adjusted according to the regional load density, the algorithm parameters can be optimized according to the transformer area size, and the monitoring frequency can be adapted according to the distributed power source penetration rate. Furthermore, for the sake of brevity, detailed descriptions of known power system simulation technologies and automation terminal principles are omitted here.

[0102] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art without departing from the scope of the invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of this application should be included within the scope of the claims of this application.

Claims

1. A highly resilient adaptive topology reconfiguration method for rural power distribution networks that integrates real-time monitoring and dynamic correction, characterized in that, Includes the following steps: Real-time monitoring technology is used to obtain load status and distributed power output data of rural power distribution networks; Based on the load status and distributed generation output data, combined with the seasonal and temporal characteristics of the load and the penetration rate of distributed generation, the distribution network topology is divided into three types: single-radial basic type, ring network enhanced type, and microgrid autonomous type. A dynamic model of the distribution network including the dynamic load of nodes and the fluctuation range of distributed generation output is established. Based on the dynamic model, a multi-objective fitness function including network loss, voltage qualification rate, and critical load reliability is constructed. An improved genetic algorithm is used to solve the multi-objective fitness function to determine the initial optimal topology switch state. Based on the graph neural network model, the ultra-short-term output of distributed power sources is predicted, and the initial optimal topology switch state is dynamically corrected based on the prediction results to obtain the corrected optimal topology switch state. The modified optimal topology switch state is executed through the distribution network automation terminal, so that the distribution network topology is adaptively adjusted to maintain the resilience index within the set range.

2. The high-resilience adaptive topology reconfiguration method for rural distribution networks integrating real-time monitoring and dynamic correction as described in claim 1, characterized in that, The method of acquiring load status and distributed power output data of rural power distribution networks using real-time monitoring technology specifically includes: Collect real-time load values, voltage values, and distributed generation output power of each node in the distribution network, and calculate the node load fluctuation coefficient and the distributed generation output fluctuation range.

3. The high-resilience adaptive topology reconfiguration method for rural distribution networks integrating real-time monitoring and dynamic correction as described in claim 1, characterized in that, The single-radial basic topology is suitable for remote rural transformer substations with low load density and distributed power penetration rate <15%; the ring network enhanced topology is suitable for township transformer substations with medium load density and distributed power penetration rate of 15%-30%; and the microgrid autonomous topology is suitable for new energy demonstration villages with distributed power penetration rate >30%.

4. The high-resilience adaptive topology reconfiguration method for rural distribution networks integrating real-time monitoring and dynamic correction as described in claim 3, characterized in that, Based on the load status and distributed generation output data, combined with the seasonal and temporal characteristics of the load and the penetration rate of distributed generation, the distribution network topology is divided into three categories: single-radial basic type, ring network enhanced type, and microgrid autonomous type, and a dynamic model is established, specifically including: The dynamic load of nodes is determined based on seasonal coefficients, time-period coefficients, and baseline loads, providing dynamic load parameters for power flow calculations and voltage threshold settings in multi-objective optimization. The output range of distributed power sources is determined based on the ultra-short-term predicted output and fluctuation coefficient, providing boundary basis for the constraint conditions and reconfiguration correction triggering judgment of the multi-objective optimization algorithm. By combining topology types, a distribution network model incorporating the dynamic characteristics of loads and distributed power sources is constructed, providing complete model support for the construction and reconstruction of the adjacency matrix predicted by graph neural networks and the verification of the execution effect.

5. The high-resilience adaptive topology reconfiguration method for rural distribution networks integrating real-time monitoring and dynamic correction as described in claim 1, characterized in that, The improved genetic algorithm uses integer encoding to represent the switch state and optimizes the population evolution process through elite retention selection, connectivity constraint crossover, and dynamic mutation probability.

6. The high-resilience adaptive topology reconfiguration method for rural distribution networks integrating real-time monitoring and dynamic correction as described in claim 1, characterized in that, The step of using an improved genetic algorithm to solve the fitness function and determine the optimal topology switch state specifically includes: Using the topology switch state as a genetic variable, a fitness function is constructed that includes network loss optimization terms, voltage qualification rate optimization terms, and critical load reliability optimization terms. Initialize the population and set the iteration parameters, then calculate the fitness value of the initial population; Select high-quality individuals using an elite retention strategy, perform crossover operations on the selected individuals, and then perform connectivity checks. Dynamic mutation probability is used to perform mutation operations on individuals to avoid premature convergence of the algorithm. Iteratively update the individual optimal values ​​and the global optimal value of the population until the convergence condition is met or the maximum number of iterations is reached; The switch state corresponding to the global optimal solution is taken as the optimal topology switch state.

7. The rural distribution network high-resilience adaptive topology reconfiguration method integrating real-time monitoring and dynamic correction as described in claim 6, characterized in that, In the fitness function, the weights of the network loss optimization term, the voltage qualification rate optimization term, and the critical load reliability optimization term are 0.4, 0.3, and 0.3, respectively.

8. The high-resilience adaptive topology reconfiguration method for rural distribution networks integrating real-time monitoring and dynamic correction as described in claim 6, characterized in that, The dynamic mutation probability ranges from 0.01 to 0.05, and gradually decreases as the number of iterations increases.

9. The high-resilience adaptive topology reconfiguration method for rural distribution networks integrating real-time monitoring and dynamic correction as described in claim 1, characterized in that, The graph neural network model takes the distribution network topology coupling model, dynamic load sequence, historical meteorological data, and distributed generation output data as inputs. It extracts spatial correlation features through graph convolutional layers and outputs ultra-short-term predicted output of distributed generation within 15 minutes, with a prediction error of ≤8%. It provides a basis for the dynamic correction of distributed generation output fluctuations in subsequent topology reconfiguration strategies. The topology and load parameters provided by the dynamic model are used to enhance the graph neural network's ability to capture the coupling relationship between 'topology-load-distributed generation'.

10. The method for highly resilient adaptive topology reconfiguration of rural distribution networks integrating real-time monitoring and dynamic correction as described in claim 1, characterized in that, The distribution network automation terminal includes feeder terminal units (FTU) and distribution terminal units (DTU). The reconfiguration strategy execution delay is ≤2 minutes. The critical loads include rural irrigation equipment and livestock and poultry breeding equipment. The resilience indicators include voltage qualification rate ≥98%, critical load reliability 100%, and network loss reduction rate ≥8%.

Images