Intelligent generation method, device and equipment for distribution network maintenance transfer scheme and storage medium

Abstract

The invention discloses an intelligent generation method and device for a distribution network maintenance transfer scheme, equipment and a storage medium. The method comprises the steps that distribution network multi-source data are acquired, model input data are determined based on the distribution network multi-source data, and the distribution network multi-source data comprise at least two of real-time operation data, equipment state data, environment influence data and maintenance task information; inputting the model input data into a pre-trained wheeling scheme generation model, and determining a candidate distribution network maintenance wheeling scheme based on an output result of the wheeling scheme generation model; and determining a target distribution network maintenance transfer scheme based on the candidate distribution network maintenance transfer scheme. The method can efficiently output the maintenance transfer scheme meeting the actual operation requirements of the distribution network, gives consideration to the generation efficiency, compliance and adaptability of the scheme, and improves the maintenance power supply reliability of the distribution network and the operation decision scientificity.

Description

Technical Field

[0001] This invention relates to the technical field of distribution network safety maintenance, and in particular to a method, apparatus, equipment, and storage medium for intelligently generating distribution network maintenance and transfer schemes. Background Technology

[0002] The power distribution network is a crucial link directly connecting the power system and users, and its safe and reliable operation directly affects the quality of power supply. When equipment failures or planned maintenance occur in the distribution network, uninterrupted power supply must be achieved through transfer operations. The rationality and efficiency of the transfer scheme are the core of ensuring power supply reliability.

[0003] Traditional methods for generating distribution network maintenance and transfer schemes primarily rely on optimal power flow solutions based on the physical characteristics of the power grid, using optimization algorithms to find the optimal transfer path that satisfies constraints. However, distribution networks are characterized by a large number of nodes, complex line branches, dense switching equipment, and dynamically changing load distribution, leading to drawbacks such as low convergence rates and long computation times in traditional methods. Distribution networks can have thousands or even tens of thousands of nodes, requiring traditional optimization algorithms to iteratively solve complex equation systems, resulting in exponentially increasing computational complexity and making it difficult to meet the real-time requirements of scenarios such as fault repair. Summary of the Invention

[0004] This invention provides a method, apparatus, equipment, and storage medium for intelligent generation of distribution network maintenance and power transfer schemes, in order to solve the problems of low convergence rate, long calculation time, and poor adaptability in existing distribution network maintenance and power transfer scheme generation methods.

[0005] According to one aspect of the present invention, a method for intelligently generating distribution network maintenance and power transfer schemes is provided, the method comprising:

[0006] Acquire multi-source data of the distribution network, and determine the model input data based on the multi-source data of the distribution network, wherein the multi-source data of the distribution network includes at least two of the following: real-time operation data, equipment status data, environmental impact data, and maintenance task information;

[0007] The model input data is input into a pre-trained power transfer scheme generation model, and candidate power transfer schemes for distribution network maintenance are determined based on the output results of the power transfer scheme generation model.

[0008] The target distribution network maintenance and power transfer scheme is determined based on the candidate distribution network maintenance and power transfer schemes.

[0009] According to another aspect of the present invention, a smart generation device for power distribution network maintenance and transfer schemes is provided, the device comprising:

[0010] A multi-source data acquisition module is used to acquire multi-source data of the distribution network and determine model input data based on the multi-source data of the distribution network. The multi-source data of the distribution network includes at least two of the following: real-time operation data, equipment status data, environmental impact data, and maintenance task information.

[0011] The candidate solution determination module is used to input the model input data into a pre-trained power transfer solution generation model, and determine the candidate power transfer solution for distribution network maintenance based on the output of the power transfer solution generation model.

[0012] The target scheme determination module is used to determine the target distribution network maintenance and transfer scheme based on the candidate distribution network maintenance and transfer schemes.

[0013] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0014] At least one processor; and

[0015] A memory communicatively connected to the at least one processor; wherein,

[0016] The memory stores a computer program that can be executed by the at least one processor, which is then executed by the at least one processor to enable the at least one processor to execute the intelligent generation method for distribution network maintenance and transfer schemes according to any embodiment of the present invention.

[0017] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions, the computer instructions being configured to cause a processor to execute and implement the intelligent generation method for distribution network maintenance and transfer schemes as described in any embodiment of the present invention.

[0018] The technical solution of this invention acquires multi-source data from the distribution network and determines model input data based on this data. The multi-source data includes at least two of the following: real-time operational data, equipment status data, environmental impact data, and maintenance task information. This avoids model input bias caused by the limited information from a single data source. Then, the model input data is input into a pre-trained power transfer scheme generation model. Based on the output of the model, candidate distribution network maintenance power transfer schemes are determined, improving the efficiency of candidate scheme generation. Finally, a target distribution network maintenance power transfer scheme is determined based on the candidate schemes. This solves the problems of large biases caused by single data sources, low efficiency of manual decision-making, and insufficient security and adaptability of schemes in traditional distribution network maintenance power transfer scheme formulation. It achieves high efficiency in generating maintenance power transfer schemes, improving the efficiency of distribution network maintenance and operation, and enhancing power supply reliability.

[0019] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0021] Figure 1 This is a flowchart of an intelligent generation method for a distribution network maintenance and transfer scheme according to Embodiment 1 of the present invention;

[0022] Figure 2 This is a flowchart of an intelligent generation method for a power distribution network maintenance and transfer scheme according to Embodiment 2 of the present invention;

[0023] Figure 3 This is a schematic diagram of the structure of an intelligent generation device for power distribution network maintenance and transfer scheme provided in Embodiment 3 of the present invention;

[0024] Figure 4 This is a schematic diagram of the structure of an electronic device that implements the intelligent generation method for distribution network maintenance and transfer schemes according to embodiments of the present invention. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0026] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0027] Example 1

[0028] Figure 1 This invention provides a flowchart of a method for intelligently generating distribution network maintenance and transfer schemes, as described in Embodiment 1. This embodiment is applicable to determining distribution network maintenance and transfer scheme situations. This method can be executed by an intelligent distribution network maintenance and transfer scheme generation device, which can be implemented in hardware and / or software and can be configured in electronic equipment. Figure 1 As shown, the method includes:

[0029] S110. Obtain multi-source data of the distribution network, and determine the model input data based on the multi-source data of the distribution network, wherein the multi-source data of the distribution network includes at least two of the following: real-time operation data, equipment status data, environmental impact data, and maintenance task information.

[0030] In this context, distribution network multi-source data can be understood as data from various sources involved in the operation of the distribution network. Real-time operational data can be understood as operational status data monitored in real time for the distribution network. Equipment status data can be understood as data reflecting the status of distribution network equipment. Environmental impact data can be understood as external environmental data affecting the operation of the distribution network. Maintenance task information can be understood as information related to distribution network maintenance work. Model input data can be understood as pre-processed (e.g., cleaning, standardization, feature extraction) distribution network multi-source data.

[0031] Specifically, first, various types of data from the power distribution network are collected, including at least two of the four types of data mentioned above; then, these collected multi-source data are processed to finally determine the input data that can be used for the model.

[0032] Optionally, the real-time operating data includes at least one of line current, voltage, power data, switch opening and closing status, user load data, and voltage deviation and harmonic data; the equipment status data includes at least one of transformer, line, equipment health index, service life, historical fault records, and rated parameters; the environmental impact data includes at least one of real-time meteorological data, short-term load forecast results, and distributed power output data; and the maintenance task information includes at least one of maintenance area, maintenance duration, equipment requiring power outage, and priority requirements.

[0033] Among these, line current, voltage, and power data can be understood as the core electrical parameters monitored in real time during the operation of the distribution network. Current reflects the line's current-carrying capacity, voltage reflects the power supply quality at nodes, and power (including active and reactive power) reflects the energy transmission and consumption status, serving as the basic data for judging whether a line is overloaded and whether the power supply is stable. Switch opening and closing status can be understood as the real-time operating status of circuit breakers, disconnectors, and other switching equipment in the distribution network (closed means conducting, open means disconnecting), directly affecting the distribution network topology and load power supply path. Voltage deviation and harmonic data can be understood as follows: voltage deviation refers to the difference between the actual supply voltage and the rated voltage (reflecting the stability of the supply voltage), and harmonic data refers to the non-fundamental frequency components in the grid voltage / current (affecting power quality and potentially damaging equipment); both are key indicators for assessing power quality. Short-term load forecast results can be understood as predictions of future short-term (e.g., 1-24 hours) distribution network load changes based on historical load data, meteorological factors, and user electricity consumption patterns. This allows for advance prediction of load pressure during maintenance periods and assists in planning load transfer schemes. Distributed power generation output data can be understood as the real-time power generation or predicted output value of distributed power sources (such as photovoltaic power generation, small wind power, and energy storage systems) in the distribution network. The access of distributed power sources will affect the power flow distribution of the distribution network, and this data is an important basis for avoiding reverse power flow overload and ensuring power transfer safety. Equipment that needs to be shut down can be understood as distribution network equipment that must be stopped during maintenance (such as transformers, line sections, and switches to be maintained). Clarifying this data can accurately locate the scope of maintenance impact and avoid power outages of unrelated equipment that lead to power waste.

[0034] S120. Input the model input data into the pre-trained power transfer scheme generation model, and determine the candidate power transfer scheme for distribution network maintenance based on the output results of the power transfer scheme generation model.

[0035] The power transfer scheme generation model can be understood as a model trained on a large amount of historical maintenance case data. The candidate power transfer scheme for distribution network maintenance can be understood as a maintenance and power transfer scheme initially output by the model based on the input data, which needs to be further verified.

[0036] Specifically, the input data of the model is imported into the pre-trained power transfer scheme generation model; the model processes the input data through the built-in encoder (capturing distribution network status characteristics and associating power constraint rules), and then generates the corresponding power transfer scheme output result through the decoder; the power transfer scheme output by the model is directly determined as the candidate distribution network maintenance power transfer scheme to be verified, in preparation for subsequent feasibility verification and target scheme selection.

[0037] Optionally, inputting the model input data into a pre-trained transfer scheme generation model includes:

[0038] The model input data is input to the encoder of the model generated by the transfer scheme based on a dual-channel input mechanism.

[0039] The dual-channel input mechanism can be understood as a feature input architecture in the power transfer scheme generation model. It divides the model input data into two independent channels: "static features" and "dynamic features," which are then transmitted separately to the encoder for processing. This simultaneously preserves both the fixed attributes of the distribution network and real-time changing information, avoiding feature confusion caused by single-channel input. The encoder in the power transfer scheme generation model can be understood as the core module responsible for "feature extraction and fusion." Through attention mechanisms, graph convolution, and other techniques, it performs deep processing on the dual-channel input distribution network data, transforming the original features into abstract feature vectors that the model can understand, providing feature support for the subsequent decoder to generate the power transfer scheme.

[0040] Specifically, the pre-processed model input data is classified into static feature channels, such as distribution network topology, equipment rated parameters, and maintenance area locations, which do not change rapidly over time. Dynamic feature channels are classified into real-time dynamic feature channels, such as real-time load fluctuations, voltage and current changes, and environmental data updates. Then, through a dual-channel input mechanism, the two types of features are transmitted to the encoder of the power transfer scheme generation model. The encoder uses an appropriate processing method for the features of different channels (such as strengthening topological correlation for static features and capturing temporal patterns for dynamic features) to complete feature extraction and preliminary fusion.

[0041] Optionally, before feeding the model input data into the pre-trained transfer scheme generation model, the following steps are also included:

[0042] Collect sample distribution network data from historical maintenance cases and corresponding optimal power transfer schemes to construct a dataset, which is then divided into a training set, a validation set, and a test set.

[0043] The pre-built neural network model based on the Transformer encoder-decoder structure is trained based on the training set. The encoder processes the distribution network status features, and the decoder generates the power transfer operation sequence, with a power constraint sensing layer added in between.

[0044] The model is trained using a cross-entropy loss function combined with a power constraint penalty term. The hyperparameters are adjusted using a validation set until the model's pass rate on the test set reaches a preset pass rate threshold. The trained model is then used as the power transfer scheme generation model.

[0045] The sample distribution network data can be understood as multi-source distribution network data extracted from historical maintenance cases, corresponding to those cases, including real-time operating data (such as current and voltage), equipment status data (such as equipment health index), and environmental impact data (such as weather conditions at the time). The optimal power transfer scheme can be understood as the final execution scheme that, in historical maintenance cases, has been verified to meet multiple objectives such as safety, efficiency, and cost (such as no overload, shortest outage time, and least resource consumption). It serves as the standard answer for model training, guiding the model to learn the correct power transfer scheme generation logic. The power constraint penalty term can be understood as a constraint reinforcement term added to the cross-entropy loss function. If the power transfer scheme generated by the model violates power constraints (such as line overload), an additional loss value will be added. Through this penalty mechanism, the model is forced to prioritize learning schemes that comply with power regulations during training, improving the compliance of the output scheme. The scheme qualification rate can be understood as a core indicator for evaluating the quality of the model output. The preset qualification rate threshold is a pre-set qualification line for judging whether the model meets the standards. This embodiment does not restrict it based on experience.

[0046] Specifically, a dataset is constructed by collecting "sample distribution network data" (such as current and equipment status at the time) and corresponding optimal power transfer schemes from historical maintenance cases. This data is integrated into a complete dataset and proportionally divided into a training set (for model learning), a validation set (for hyperparameter tuning), and a test set (for final evaluation). Using the training set as input, a Transformer encoder-decoder neural network model is trained. The encoder first extracts key features from the sample distribution network data, incorporates power regulations through a power constraint perception layer, and then the decoder generates a power transfer operation sequence, allowing the model to initially grasp the corresponding logic of "distribution network status → power transfer scheme." The model error is calculated using a "cross-entropy loss function + power constraint penalty term." If the generated scheme differs significantly from the optimal scheme or violates power constraints, a high loss value is generated, and the model adjusts its internal parameters accordingly. Simultaneously, the validation set is used to evaluate model performance, and hyperparameters (such as the learning rate) are adjusted based on the evaluation results to avoid overfitting. Once the model has been trained to a certain extent, its solution qualification rate is tested using a test set. If the qualification rate reaches the preset qualification rate threshold, it means that the model's ability to adapt to unfamiliar data and the compliance of the solution have met the standards. Training is then stopped, and the model at this point is determined as a practically usable transfer solution generation model.

[0047] S130. Determine the target distribution network maintenance and transfer scheme based on the candidate distribution network maintenance and transfer schemes.

[0048] Among them, the target distribution network maintenance and power transfer plan can be understood as the final maintenance and power transfer plan that can be actually implemented.

[0049] Optionally, determining the target distribution network maintenance and power transfer scheme based on the candidate distribution network maintenance and power transfer schemes includes:

[0050] Feasibility verification of the output distribution network maintenance and transfer scheme shall be conducted, including at least one of the following:

[0051] The power distribution network maintenance and transfer scheme is checked against basic electrical constraints using preset rules.

[0052] Verify whether the voltage and current after the implementation of the power supply transfer scheme for distribution network maintenance are within the preset safety range;

[0053] Simulate the execution process of the aforementioned power supply transfer and maintenance scheme, and evaluate its time efficiency and resource consumption;

[0054] If the verification results are satisfactory, the candidate distribution network maintenance and power supply transfer scheme will be determined as the target distribution network maintenance and power supply transfer scheme.

[0055] Among them, the preset rules can be understood as the basic judgment conditions based on the distribution network operation specifications (such as the transfer path must be connected, the switch operation sequence must comply with safety regulations, etc.). The basic electrical constraints can be understood as the core electrical rules that the distribution network operation must meet (such as the line must not be overloaded, the voltage must not exceed the rated range ±5%, etc.).

[0056] The preset safety range can be understood as the voltage and current safety thresholds set according to equipment parameters and operating standards (such as the voltage safety range of a 10kV line, which is usually 9.5-10.5kV). It can be preset based on experience, and this embodiment does not limit it.

[0057] Specifically, feasibility verification is conducted on the candidate distribution network maintenance and power transfer schemes output by the model. Verification methods include at least one of the following: checking whether the scheme meets basic electrical constraints according to preset rules; verifying whether the voltage and current after implementation are within preset safety ranges through power flow calculations; simulating the execution process of the scheme in a simulation environment to evaluate the time efficiency and required manpower, equipment, and other resource consumption for completing the maintenance. If all the above verification results are passed, the candidate scheme is determined as a practically executable target distribution network maintenance and power transfer scheme. If any verification result fails, the candidate distribution network maintenance and power transfer scheme that fails verification is sent to the model to trigger the model to regenerate a new candidate distribution network maintenance and power transfer scheme.

[0058] Optionally, the target distribution network maintenance and transfer scheme includes a switch operation sequence, load transfer path, safety control measures, and scheme evaluation indicators.

[0059] The switch operation sequence can be understood as a list of the order in which various switches (such as circuit breakers and disconnectors) within the distribution network are operated during distribution network maintenance and power transfer. Distribution network switches are key equipment controlling the connection and disconnection of lines and enabling load transfer. The operation sequence must strictly adhere to power safety regulations (e.g., disconnecting the load-side switch first, then the power-side switch, or closing the standby power switch first, then switching the load switch) to avoid errors in the operation sequence that could lead to short circuits, equipment damage, or personnel safety accidents. For example, during the maintenance of a certain line, the operation sequence might be: "First disconnect the circuit breakers at both ends of the line under maintenance → open the disconnectors on both sides → close the tie circuit breaker of the standby line → transfer the original load to the standby line." The load transfer path can be understood as the specific channel through which the user loads (such as residential electricity and enterprise electricity) in the distribution network area that needs to be de-energized during maintenance are transferred from the original power supply line to other normally operating lines. For example, when a 10kV Line A is under maintenance, the loads of residential area A and factory B on Line A can be transferred to 10kV Line B or 10kV Line C via a tie line. This path, "Line A → Tie Line → Line B / C," is the load transfer path. Safety control measures can be understood as targeted protective measures designed to ensure the safety of personnel, equipment, and the stable operation of the distribution network during the implementation of the distribution network maintenance and transfer scheme. These include two aspects: first, personnel safety measures (such as workers wearing insulated protective equipment, setting up safety warning signs, and implementing power outage testing and grounding procedures); second, equipment and system safety measures (such as real-time monitoring of the current and voltage of the lines after transfer, setting overload protection settings, preemptively disconnecting some non-critical loads to avoid line overcapacity, and preventing voltage spikes and drops during the transfer process). For example, before load transfer, insulation testing of the lines along the transfer path is required to ensure there are no potential faults. Scheme evaluation indicators can be understood as parameters used to quantitatively evaluate the merits of the target distribution network maintenance and transfer scheme. Power supply reliability indicators can be understood as outage duration (the time users experience power outages during the transfer process) and outage scope (the number of affected users). Better indicators indicate less impact on user electricity consumption. Safety compliance indicators can be understood as whether the maximum line current exceeds the rated value, whether the voltage deviation is within the allowable range, and whether the switch operation complies with regulations. Meeting these indicators is the foundation for a feasible solution. Economic efficiency indicators can be understood as the equipment wear and tear costs, labor costs, and resource (such as emergency generator) input costs during the transfer process. Better indicators indicate better economic efficiency. Operational complexity indicators can be understood as the number of switch operations and the number of load transfer steps. Better indicators indicate lower implementation difficulty and reduced risk of human error.

[0060] The technical solution of this invention acquires multi-source data from the distribution network and determines model input data based on this data. The multi-source data includes at least two of the following: real-time operational data, equipment status data, environmental impact data, and maintenance task information. This avoids model input bias caused by the limited information from a single data source. Then, the model input data is input into a pre-trained power transfer scheme generation model. Based on the output of the model, candidate distribution network maintenance power transfer schemes are determined, improving the efficiency of candidate scheme generation. Finally, a target distribution network maintenance power transfer scheme is determined based on the candidate schemes. This solves the problems of large biases caused by single data sources, low efficiency of manual decision-making, and insufficient security and adaptability of schemes in traditional distribution network maintenance power transfer scheme formulation. It achieves high efficiency in generating maintenance power transfer schemes, improving the efficiency of distribution network maintenance and operation, and enhancing power supply reliability.

[0061] Example 2

[0062] Figure 2 This is a flowchart of an intelligent generation method for distribution network maintenance and transfer schemes provided in Embodiment 2 of the present invention. This embodiment is a further refinement of the above embodiment. Optionally, determining the model input data based on the multi-source data of the distribution network includes: preprocessing the multi-source data of the distribution network, and determining the preprocessed multi-source data of the distribution network as the model input data, wherein the preprocessing includes data cleaning, standardization, and feature extraction.

[0063] like Figure 2 As shown, the method includes:

[0064] S210. Obtain multi-source data of the distribution network, preprocess the multi-source data of the distribution network, and determine the preprocessed multi-source data of the distribution network as the input data of the model. The preprocessing includes data cleaning, standardization and feature extraction.

[0065] Specifically, the system scans multi-source data from the distribution network to identify "dirty data" such as missing values ​​and outliers. For a small number of missing values, the average of the values ​​before and after the missing value is used to fill the gap (e.g., if the current at 10:05 is missing for a line, the average current at 10:00 and 10:10 is used to replace it). For outliers, they are removed or corrected by determining whether they exceed the rated range of the equipment (e.g., the current exceeds the maximum allowable value of the line). Duplicate values ​​are deleted directly, and incorrect formats are uniformly adjusted (e.g., all dates are changed to the "year-month-day hour:minute:second" format). Finally, accurate and interference-free basic data is obtained. For different types of data after cleaning, a preset mathematical method is used to adjust data of different magnitudes, such as current, voltage, and load rate, to a uniform range. From the standardized basic data, core features that are meaningful for generating the power transfer scheme are selected. For example, "real-time load rate of the line" and "voltage deviation value" are extracted from equipment monitoring data, "number of backup power supply paths for the line to be inspected" are extracted from topology data, and "total load of affected users" are extracted from user load data. At the same time, irrelevant information (such as equipment manufacturing numbers that are not related to the operation of the distribution network and historical weather data of non-maintenance areas) are removed, and finally structured model input data is formed.

[0066] S220. Input the model input data into the pre-trained power transfer scheme generation model, and determine the candidate power transfer scheme for distribution network maintenance based on the output results of the power transfer scheme generation model.

[0067] S230. Determine the target distribution network maintenance and transfer scheme based on the candidate distribution network maintenance and transfer schemes.

[0068] The technical solution of this invention preprocesses the multi-source data of the distribution network, using the preprocessed data as model input data. The preprocessing includes data cleaning, standardization, and feature extraction. This lays a high-quality data foundation for accurate model reasoning and the output of compliant and effective maintenance and power transfer solutions for power transfer schemes.

[0069] As an optional example of Embodiment 1 of the present invention, the intelligent method for power distribution network maintenance and transfer scheme of this embodiment specifically includes the following steps:

[0070] Step S1: Integrate multi-source heterogeneous data from the distribution network to construct a training dataset covering various maintenance scenarios.

[0071] In step S1, the training dataset consists of two parts: input features and output labels. The input features cover three core dimensions: network topology features, electrical parameter features, and maintenance information features. Network topology features include information such as the number of nodes, line connection relationships, switch status, equipment type, bus configuration, and protection configuration. Electrical parameter features include parameters such as load power of each node, node voltage amplitude, node voltage phase angle, line impedance, line resistance, line reactance, line current carrying capacity limit, transformer capacity, transformer ratio, and reactive power compensation device parameters. Maintenance information features include information such as maintenance area range, maintenance time window, maintenance duration, load range to be transferred, maintenance equipment type, and maintenance priority. The output labels are the optimal power transfer scheme generated by the commercial solver, including the power transfer path sequence, switch operation sequence, load transfer allocation, voltage regulation measures, reactive power compensation scheme, and safety margin assessment results.

[0072] Specifically, the integration of multi-source heterogeneous data includes four types of data sources: real-time operational data, including telemetry data from SCADA systems, status information from distribution automation systems, load data from electricity consumption information collection systems, and data from power quality monitoring devices; equipment status data, including equipment health index, equipment operating years, historical maintenance records, equipment rated parameters, equipment fault statistics, and equipment performance degradation curves; environmental impact data, including weather forecast information, temperature and humidity data, wind speed and direction data, load forecast results, distributed power generation output forecasts, and new energy power generation forecasts; and protection information data, including relay protection configuration parameters, protection setting values, protection action records, protection coordination relationships, and fault waveform data.

[0073] Optionally, the training dataset construction process includes complete data preprocessing and feature engineering steps: the data cleaning stage includes outlier detection, missing value imputation, noise filtering, and data consistency verification to ensure data quality and integrity; the data standardization stage includes Z-score standardization, maximum-minimum normalization, and quantile standardization to eliminate differences in units and inconsistencies in numerical ranges; the feature extraction stage includes temporal feature extraction, topological feature extraction, electrical feature extraction, and statistical feature extraction to construct multi-dimensional feature representations; the feature selection stage uses methods such as mutual information, Pearson correlation coefficient, and analysis of variance to screen key features, reduce dimensionality, and improve model training efficiency.

[0074] Step S2: Construct a power transfer scheme generation model based on power constraint perception and graph-enhanced Transformer architecture.

[0075] In step S2, the power constraint sensing attention mechanism introduces power system physical constraints to correct the standard attention calculation, enabling the model to sense power grid operation constraints. The calculation formula is as follows:

[0076]

[0077] Among them, C power Let be the power constraint matrix, α, γ1, γ2, γ3 be learnable parameters, and d be the power constraint matrix. k For attention head dimension.

[0078] The power constraint matrix C power It consists of three types of core constraints, constructed through a weighted combination:

[0079]

[0080] The specific definitions of each constraint term in the power constraint matrix are as follows:

[0081] Voltage constraint terms:

[0082]

[0083] Capacity constraints:

[0084]

[0085] Power flow constraints:

[0086]

[0087] Among them, V i V j Let S be the voltage magnitude at nodes i and j. ij , δ represents the apparent power and rated capacity of line ij. i δ j Let P be the voltage phase angle at nodes i and j. ij , Let represent the active power and maximum transmission power of line ij, and β be a learnable parameter.

[0088] Optionally, in step S2, the graph-enhanced Transformer hybrid architecture achieves accurate modeling of the distribution network topology through deep fusion of graph neural networks and Transformers. This hybrid architecture includes two parallel processing branches: graph paths and sequence paths. The graph-enhanced attention mechanism incorporates graph structure information on top of standard attention, and the calculation formula is as follows:

[0089]

[0090] Among them, G structure β is the graph structure perception matrix, and β is the graph structure weight parameter. This mechanism can simultaneously consider sequence relationships and topological relationships, making full use of the spatial structure information of the distribution network.

[0091] The graph structure perception matrix G structureBased on the definition of graph distance between nodes, the connectivity characteristics of the power grid topology are reflected:

[0092]

[0093] Among them, w direct w 2hop w 3hop d is a learnable distance weight parameter. graph(i,j) Let E be the shortest graph distance between nodes i and j, and let E be the set of edges. The distance attenuation factor; the dual-path parallel processing architecture processes topological relationships through graph paths and temporal decision relationships through sequence paths, with the fusion formula being:

[0094] H fused =γ graph ·GCN(X,A)+γ seq Transformer(X) seq )+CrossAttention(H graph H seq )

[0095] Among them, H fused For the fused feature representation, γ graph γ seq Here, X represents the weight parameters for the graph path and the sequence path, GCN is a graph convolutional network, X is the node feature matrix, A is the adjacency matrix, and Transformer is the sequence transformer. seq For serialized input features, H graph H represents the topological features output by the graph convolutional network. seq The output sequence features are represented by Transformer, and CrossAttention is a cross-attention fusion mechanism.

[0096] The cross-attention fusion mechanism adopts a multi-head attention structure and achieves bidirectional information interaction by querying graph features and key-value sequence features. The specific calculation is as follows:

[0097] CrossAttention(H graph H seq ) = MultiHead(H graph H seq H seq )+MultiHead(H seq H graph H graph )

[0098] This mechanism ensures that graph structure information and sequence decision information can be fully integrated.

[0099] Optionally, in step S2, a hierarchical topology coding algorithm is used to process the multi-level structure of the distribution network, including three levels: main transformer layer, feeder layer, and distribution transformer layer. The hierarchical awareness graph embedding method is as follows:

[0100]

[0101] Where l is the hierarchical number (1 = main transformer, 2 = line, 3 = distribution transformer), GCN is the graph convolutional network operation function, and A is the hierarchical adjacency matrix. l Consider electrical distance weighting:

[0102]

[0103] Electrical distance is defined as:

[0104]

[0105] Among them, Z ij R is the line impedance. ij X ij P represents the line resistance and reactance. i P j For node load, σ and λ are adjustment parameters, and E l Let be the set of edges in the l-th layer.

[0106] Optionally, in step S2, the sequence generation and decoding algorithm adopts a constraint-aware autoregressive decoding strategy to ensure that the generated power transfer operation sequence meets power system security constraints and timing logic constraints.

[0107]

[0108] Constraint modification items:

[0109]

[0110] Timing correction items:

[0111]

[0112] Where p(y) t |y <t (x) represents the input x and the historical sequence y. <t Predict the probability of the operation at time t under the given conditions, y t For the transfer operation at time t, y <t Let x be the sequence of historical operations before time t, x be the input feature, MLP be a multilayer perceptron network, and h be the input feature. t Let be the hidden state representation at time t. ψ is the constraint correction weight parameter, C is the electrical constraint correction term, T is the timing constraint correction term, and F is the weight parameter. elecFor electrical constraint checking functions, F time is the timing constraint checking function, and softmax is the normalization function.

[0113] Optionally, in step S2, the model training employs a multi-objective optimization loss function, comprehensively considering prediction accuracy, constraint satisfaction, safety margin, and economic performance.

[0114] L total =L seq +λ1·L constraint +λ2·L safety +λ3·L efficiency

[0115] The sequence prediction loss uses cross-entropy loss:

[0116]

[0117] Where T is the sequence length. For real operation, This represents the actual historical sequence of operations.

[0118] Constraint violation penalty:

[0119]

[0120] Where N is the batch size, K is the number of constraint types, and g k Let ξ be the k-th constraint function. k ε k To constrain weights and tolerances.

[0121] Security loss:

[0122]

[0123] Where C is the set of key nodes, w c SM represents the importance weight of node c. min To meet minimum safety margin requirements, SM c This is the c-th type of safety margin, which includes voltage safety margin, power safety margin, and stability safety margin.

[0124] Economic losses:

[0125] L efficiency =α loss network loss +α switching ·switching cost

[0126] Among them, network loss For network loss, switchingcost For the cost of switching operation, α loss α switching The weights for each cost item.

[0127] Optionally, the specific calculation of each loss in the multi-objective loss function includes: sequence prediction loss using cross-entropy loss to measure the difference between the predicted sequence and the actual sequence; constraint violation loss penalizing the prediction results that violate power system operation constraints through a combination of soft and hard constraints; safety loss considering the system's N-1 safety criterion and stability margin requirements; economic loss integrating factors such as network losses, switching operation costs, and reliability costs; and scheme diversity loss calculating the differences between different schemes through cosine similarity to avoid the model generating a single solution.

[0128] Optionally, in step S2, the model architecture adopts a deep separable attention mechanism, with the encoder and decoder each containing 8 Transformer layers, each containing 16 attention heads, a hidden layer dimension of 768, and a feedforward network dimension of 3072; the Adam optimizer is used for parameter updates, with an initial learning rate of 2e-4, a cosine annealing learning rate scheduling strategy, a batch size of 64, and 200 training epochs; gradient clipping and early stopping mechanisms are used to prevent overfitting, while regularization techniques such as Dropout and LayerNorm are introduced to improve the model's generalization ability.

[0129] Optionally, the model training process adopts a course learning strategy, arranging training samples in order of increasing difficulty: in the initial stage, basic training is carried out using simple topologies and scenarios with a small number of faults to establish basic transfer logic; in the middle stage, complex topologies and multi-fault scenarios are introduced to improve the model's ability to handle complex situations; in the later stage, complex factors such as extreme weather and equipment aging are added to enhance the robustness of the model; at the same time, data augmentation techniques are used, including topology transformation, load disturbance, parameter noise injection and other methods to increase the diversity of training samples.

[0130] Optionally, in step S2, a transfer learning and continuous learning mechanism is established: in the pre-training stage, a basic model is trained using large-scale general distribution network data to learn general power system knowledge and transfer strategies; in the fine-tuning stage, domain adaptation is performed based on the equipment characteristics and operating rules of the distribution network in a specific area; in the online learning stage, the model parameters are continuously optimized based on actual operation feedback, key samples are stored using an experience replay buffer, and knowledge sharing among multiple distribution networks is achieved through federated learning to ensure that the model can continuously improve and adapt to environmental changes.

[0131] Step S3: Input real-time data into the training model and output end-to-end maintenance and supply transfer solutions.

[0132] Optionally, in step S3, the real-time inference system adopts a layered architecture design, including four core modules: a data access layer, a feature processing layer, a model inference layer, and a result output layer. The data access layer accesses data from systems such as SCADA, EMS, and distribution automation through standardized interfaces such as OPC, IEC 61850, CIM, and RESTAPI. The feature processing layer extracts and standardizes input features in real time and processes time-series data using sliding window technology. The model inference layer loads the trained model for forward computation. The result output layer generates standardized transfer schemes and verifies their feasibility.

[0133] Specifically, the output of the scheme includes three levels of prediction tasks: switch operation prediction outputs the specific operation instructions for each switch device, including operation type (close / open / no operation), operation sequence, and operation priority; line selection prediction determines the set of lines participating in power transfer and the load allocation ratio; system regulation prediction provides auxiliary measures such as voltage regulation, reactive power compensation, and load control; all prediction results include confidence assessment and uncertainty quantification.

[0134] Optionally, in step S3, a multi-level feasibility verification mechanism is established: rapid pre-verification uses heuristic rules to check obviously infeasible solutions; detailed verification verifies electrical constraints through simplified power flow calculations; precise verification calls a complete power flow analysis tool for comprehensive verification; solutions that fail verification trigger a regeneration mechanism, using a combination of bundle search and genetic algorithms to select suboptimal feasible solutions from candidate solutions.

[0135] Step S4: Establish a self-healing system for distribution network faults to automate the entire fault handling process.

[0136] Optionally, in step S4, the distribution network fault self-healing and intelligent recovery system adopts a layered distributed architecture, including a perception layer, a decision layer, an execution layer, and an evaluation layer. The perception layer establishes a fault location mechanism that integrates multi-source information, such as SCADA alarm information, relay protection action information, fault waveform data, customer complaint information, and line inspection information. An improved Bayesian network is used for information fusion, and the fault location and fault type are determined through a confidence propagation algorithm.

[0137] Optionally, in step S4, the decision-making layer designs an intelligent load transfer capacity assessment algorithm, comprehensively considering the target feeder's carrying capacity and transmission conditions:

[0138] transfer capability(i,j) =min(capacity) remaining(j) ,path limit(i,j) voltage margin(j) ,reactive balance(j) )

[0139] Among them, capacity remaining For the remaining capacity of the target feeder, path limit For path transmission limitations, voltage margin For voltage regulation margin, reactive balance To ensure reactive power balancing capability, and considering dynamic factors such as load forecasting uncertainty, equipment health status, weather impact, and system operation mode, a load transfer risk assessment model is established.

[0140] Optionally, in step S4, the execution layer establishes a tiered and time-based recovery strategy, formulating differentiated recovery sequences based on load importance and power supply time requirements: Level 1 loads include important government agencies, hospital operating rooms, data center UPS, communication hubs, etc., requiring power restoration within 15 seconds; Level 1 loads include hospitals, schools, financial institutions, rail transit, important industrial users, etc., requiring power restoration within 2 minutes; Level 2 loads include commercial complexes, residential communities, general industrial facilities, etc., requiring power restoration within 10 minutes; Level 3 loads include agriculture, landscape lighting, non-critical facilities, etc., requiring power restoration within 30 minutes.

[0141] Optionally, in step S4, the execution layer also establishes a dynamic network reconfiguration optimization mechanism. After completing fault isolation and load transfer, it automatically optimizes the network operation mode to minimize operating losses. The reconfiguration optimization comprehensively considers multiple objectives such as network losses, voltage deviation, current imbalance, and switching operation costs, and uses an improved genetic algorithm to solve them. The constraints include power balance constraints, voltage range constraints, line capacity constraints, radial constraints, switching state constraints, and N-1 safety constraints.

[0142] Optionally, in step S4, the evaluation layer establishes a self-healing effect evaluation mechanism, including a continuous learning and improvement mechanism: establishing a self-healing case knowledge base to record the complete process of each fault handling in detail; using reinforcement learning algorithms to extract experience rules and best practices from historical cases; designing an adaptive threshold adjustment algorithm to dynamically optimize various decision parameters based on historical performance and environmental changes; and establishing a performance monitoring mechanism to monitor the system's operating status in real time.

[0143] The technical solution of this invention lays a data foundation by integrating multi-source heterogeneous data and constructing a high-quality training dataset. It relies on power constraint perception and graph augmentation Transformer architecture to achieve accurate modeling and compliant solution generation. It combines real-time inference and multi-level verification to ensure the feasibility of the solution. It is equipped with a hierarchical distributed distribution network fault self-healing system to achieve full-process automation. At the same time, it optimizes model performance through transfer learning and course learning. Finally, it efficiently outputs a safe, compliant, and economically adaptable maintenance and power supply solution, which greatly improves the efficiency of distribution network maintenance and operation and the reliability of power supply.

[0144] Example 3

[0145] Figure 3 This is a schematic diagram of the intelligent generation device for power distribution network maintenance and transfer schemes provided in Embodiment 3 of the present invention. Figure 3 As shown, the device includes: a multi-source data acquisition module 310, a candidate scheme determination module 320, and a target scheme determination module 330.

[0146] The multi-source data acquisition module 310 is used to acquire multi-source data of the distribution network and determine model input data based on the multi-source data of the distribution network. The multi-source data of the distribution network includes at least two of the following: real-time operation data, equipment status data, environmental impact data, and maintenance task information. The candidate scheme determination module 320 is used to input the model input data into a pre-trained power transfer scheme generation model and determine candidate distribution network maintenance power transfer schemes based on the output results of the power transfer scheme generation model. The target scheme determination module 330 is used to determine a target distribution network maintenance power transfer scheme based on the candidate distribution network maintenance power transfer schemes.

[0147] The technical solution of this invention acquires multi-source data from the distribution network and determines model input data based on this data. The multi-source data includes at least two of the following: real-time operational data, equipment status data, environmental impact data, and maintenance task information. This avoids model input bias caused by the limited information from a single data source. Then, the model input data is input into a pre-trained power transfer scheme generation model. Based on the output of the model, candidate distribution network maintenance power transfer schemes are determined, improving the efficiency of candidate scheme generation. Finally, a target distribution network maintenance power transfer scheme is determined based on the candidate schemes. This solves the problems of large biases caused by single data sources, low efficiency of manual decision-making, and insufficient security and adaptability of schemes in traditional distribution network maintenance power transfer scheme formulation. It achieves high efficiency in generating maintenance power transfer schemes, improving the efficiency of distribution network maintenance and operation, and enhancing power supply reliability.

[0148] Optionally, the candidate solution determination module is specifically used for:

[0149] The model input data is input to the encoder of the model generated by the transfer scheme based on a dual-channel input mechanism.

[0150] Optionally, the device further includes:

[0151] The dataset construction module is used to collect sample distribution network data of historical maintenance cases and corresponding optimal power transfer schemes before inputting the model input data into the pre-trained power transfer scheme generation model, so as to construct a dataset and divide the dataset into training set, validation set and test set.

[0152] The model training module is used to train a pre-built neural network model based on the Transformer encoder-decoder structure based on the training set. The encoder processes the distribution network state features, and the decoder generates the power transfer operation sequence, with an electric power constraint sensing layer added in between.

[0153] The model determination module is used to train the model based on the cross-entropy loss function combined with the power constraint penalty term. The hyperparameters are adjusted through the validation set until the model's pass rate on the test set reaches the preset pass rate threshold. Then the training ends and the trained model is used as the model for the power transfer scheme.

[0154] Optionally, the base target scheme determination module is specifically used for:

[0155] The feasibility of the candidate power distribution network maintenance and transfer schemes shall be verified, including at least one of the following:

[0156] The candidate power distribution network maintenance and transfer schemes are checked against basic electrical constraints using preset rules.

[0157] Verify whether the voltage and current after the implementation of the candidate distribution network maintenance and power transfer scheme are within the preset safety range;

[0158] Simulate the execution process of the candidate distribution network maintenance and power transfer schemes to evaluate time efficiency and resource consumption;

[0159] If the verification results are satisfactory, the candidate distribution network maintenance and power supply transfer scheme will be determined as the target distribution network maintenance and power supply transfer scheme.

[0160] Optionally, the real-time operating data includes at least one of line current, voltage, power data, switch opening and closing status, user load data, and voltage deviation and harmonic data; the equipment status data includes at least one of transformer, line, equipment health index, service life, historical fault records, and rated parameters; the environmental impact data includes at least one of real-time meteorological data, short-term load forecast results, and distributed power output data; and the maintenance task information includes at least one of maintenance area, maintenance duration, equipment requiring power outage, and priority requirements.

[0161] Optionally, the target distribution network maintenance and transfer scheme includes a switch operation sequence, load transfer path, safety control measures, and scheme evaluation indicators.

[0162] Optionally, the multi-source data acquisition module is specifically used for:

[0163] The multi-source data of the distribution network is preprocessed, and the preprocessed multi-source data of the distribution network is determined as the input data of the model. The preprocessing includes data cleaning, standardization, and feature extraction.

[0164] The intelligent generation device for distribution network maintenance and power transfer schemes provided in this embodiment of the invention can execute the intelligent generation method for distribution network maintenance and power transfer schemes provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0165] Example 4

[0166] Figure 4 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0167] like Figure 4 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0168] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0169] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, central processing unit (CPU), graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as the intelligent generation of power distribution network maintenance and transfer schemes.

[0170] In some embodiments, the intelligent generation of the distribution network maintenance and power supply transfer scheme can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the intelligent generation of the distribution network maintenance and power supply transfer scheme described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform the intelligent generation of the distribution network maintenance and power supply transfer scheme by any other suitable means (e.g., by means of firmware).

[0171] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0172] Computer programs used to implement the methods of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs can be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0173] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0174] To provide interaction with a service recipient, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the service recipient; and a keyboard and pointing device (e.g., a mouse or trackball) through which the service recipient can provide input to the electronic device. Other types of devices can also be used to provide interaction with the service recipient; for example, feedback provided to the service recipient can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the service recipient can be received in any form (including voice input, speech input, or tactile input).

[0175] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., service-seeking computers with a graphical service-seeking interface or a web browser through which the service-seeking party can interact with the implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0176] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through a communication network. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0177] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and no limitation is imposed herein.

[0178] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for intelligently generating power distribution network maintenance and transfer schemes, characterized in that, include: Acquire multi-source data of the distribution network, and determine the model input data based on the multi-source data of the distribution network, wherein the multi-source data of the distribution network includes at least two of the following: real-time operation data, equipment status data, environmental impact data, and maintenance task information; The model input data is input into a pre-trained power transfer scheme generation model, and candidate power transfer schemes for distribution network maintenance are determined based on the output results of the power transfer scheme generation model. The target distribution network maintenance and power transfer scheme is determined based on the candidate distribution network maintenance and power transfer schemes.

2. The method according to claim 1, characterized in that, The model input data is fed into a pre-trained transfer scheme to generate a model, including: The model input data is input to the encoder of the model generated by the transfer scheme based on a dual-channel input mechanism.

3. The method according to claim 1, characterized in that, Before inputting the model input data into the pre-trained transfer scheme generation model, the following steps are also included: Collect sample distribution network data from historical maintenance cases and corresponding optimal power transfer schemes to construct a dataset, which is then divided into a training set, a validation set, and a test set. The pre-built neural network model based on the Transformer encoder-decoder structure is trained based on the training set. The encoder processes the distribution network status features, and the decoder generates the power transfer operation sequence, with a power constraint sensing layer added in between. The model is trained using a cross-entropy loss function combined with a power constraint penalty term. The hyperparameters are adjusted using a validation set until the model's pass rate on the test set reaches a preset pass rate threshold. The trained model is then used as the power transfer scheme generation model.

4. The method according to claim 1, characterized in that, The process of determining the target distribution network maintenance and power supply transfer scheme based on the candidate distribution network maintenance and power supply transfer schemes includes: The feasibility of the candidate power distribution network maintenance and transfer schemes shall be verified, including at least one of the following: The candidate power distribution network maintenance and transfer schemes are checked against basic electrical constraints using preset rules. Verify whether the voltage and current after the implementation of the candidate distribution network maintenance and power transfer scheme are within the preset safety range; Simulate the execution process of the candidate distribution network maintenance and power transfer schemes to evaluate time efficiency and resource consumption; If the verification results are satisfactory, the candidate distribution network maintenance and power supply transfer scheme will be determined as the target distribution network maintenance and power supply transfer scheme.

5. The method according to claim 1, characterized in that, The real-time operating data includes line current, voltage, power data, switch opening and closing status, user load data, and voltage deviation and harmonic data; the equipment status data includes transformer, line, equipment health index, service life, historical fault records, and rated parameters; the environmental impact data includes real-time meteorological data, short-term load forecast results, and distributed power output data; the maintenance task information includes maintenance area, maintenance duration, equipment to be shut down, and priority requirements.

6. The method according to claim 1, characterized in that, The target distribution network maintenance and transfer scheme includes switch operation sequence, load transfer path, safety control measures, and scheme evaluation indicators.

7. The method according to claim 1, characterized in that, The determination of model input data based on the multi-source data of the distribution network includes: The multi-source data of the distribution network is preprocessed, and the preprocessed multi-source data of the distribution network is determined as the input data of the model. The preprocessing includes data cleaning, standardization, and feature extraction.

8. A smart generation device for power distribution network maintenance and transfer schemes, characterized in that, include: A multi-source data acquisition module is used to acquire multi-source data of the distribution network and determine model input data based on the multi-source data of the distribution network. The multi-source data of the distribution network includes at least two of the following: real-time operation data, equipment status data, environmental impact data, and maintenance task information. The candidate solution determination module is used to input the model input data into a pre-trained power transfer solution generation model, and determine the candidate power transfer solution for distribution network maintenance based on the output of the power transfer solution generation model. The target scheme determination module is used to determine the target distribution network maintenance and transfer scheme based on the candidate distribution network maintenance and transfer schemes.

9. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to execute the intelligent generation method for distribution network maintenance and transfer schemes as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that are used to cause a processor to execute the intelligent generation method for distribution network maintenance and transfer schemes as described in any one of claims 1-7.

Images