Distribution network maintenance plan aided decision-making method and system based on multi-target constraint optimization

Abstract

The embodiment of the application provides a distribution network maintenance plan aided decision-making method and system based on multi-target constraint optimization, and belongs to the technical field of power system dispatching automation. The method comprises the following steps: collecting operation state data of a distribution network and to-be-executed maintenance task information in real time; determining multi-target constraint conditions and optimization targets of the operation state data according to a network topology structure of the distribution network, equipment rated parameters and safety operation regulations; generating a plurality of candidate maintenance plan schemes by using a pre-trained large language model; screening a Pareto front solution set and determining an optimal scheme based on a multi-target optimization algorithm; and outputting the optimal maintenance plan scheme and generating an aided decision-making instruction. The system comprises a data acquisition module, a constraint and target configuration module, a large model reasoning module, a multi-target optimization module and an output and decision-making module.

Description

Technical Field

[0001] This invention relates to the field of power system dispatch automation technology, and more specifically to a method and system for auxiliary decision-making of distribution network maintenance plans based on multi-objective constraint optimization. Background Technology

[0002] As a crucial link connecting the transmission network and end users, the reliability of the distribution network directly affects the quality of electricity supply for social production and residential life. With the large-scale integration of distributed power sources, the rapid growth of electric vehicle charging loads, and the increasing demands of users for power supply reliability, the operating environment of the distribution network is becoming increasingly complex. To ensure the health of equipment and the safe operation of the power grid, regular power outages are required for maintenance of distribution lines, transformers, and switching equipment. However, formulating maintenance plans requires comprehensive consideration of multiple factors, including the power grid's operating status, load forecasting, maintenance resources, power supply reliability requirements, and load transfer capabilities. It is a typical complex optimization problem with multiple objectives and constraints.

[0003] Traditional maintenance planning relies heavily on dispatchers' experience, involving repeated trial and error to determine maintenance timing and outage scope. This approach is not only inefficient but also struggles to achieve a globally optimal solution across multiple conflicting objectives (such as minimizing the number of users experiencing power outages, minimizing operational risks, and maximizing plan balance). While researchers have employed intelligent optimization methods like genetic algorithms and particle swarm optimization for maintenance planning in recent years, these methods often require extensive iterative computation and have limited capacity to handle constraints, making it difficult to effectively utilize historical experience and expert knowledge in power grid operation. With the development of artificial intelligence, large language models have demonstrated powerful capabilities in semantic understanding and reasoning generation; however, an effective technical solution remains to be found for organically integrating them with multi-objective constraint optimization of distribution network maintenance plans. Summary of the Invention

[0004] The purpose of this invention is to address the problems of traditional maintenance plan formulation relying on human experience, difficulty in handling multi-objective conflicts, and insufficient utilization of expert knowledge in existing optimization methods. It provides a distribution network maintenance plan auxiliary decision-making method and system based on multi-objective constraint optimization. This method and system combine the reasoning and generation capabilities of large language models with the search capabilities of multi-objective optimization algorithms, enabling automatic generation of maintenance plan schemes that take into account multiple optimization objectives, thus providing auxiliary decision support for dispatchers.

[0005] To achieve the above objectives, one embodiment of the present invention provides a method for auxiliary decision-making of distribution network maintenance plans based on multi-objective constraint optimization, the method comprising: Real-time acquisition of distribution network operation status data and maintenance task information to be executed. The operation status data includes the voltage of each node, the current of each branch, the transformer load rate, the output of distributed power sources and the load forecast value. The maintenance task information includes the maintenance equipment identification, maintenance time window, maintenance priority and the required power outage range. Based on the network topology, equipment rated parameters, and safe operation procedures of the distribution network, the multi-objective constraints and optimization objectives of the operating status data are determined. The multi-objective constraints include at least node voltage constraints, branch current constraints, power supply reliability constraints, maintenance resource constraints, and load transfer capability constraints. The optimization objectives include minimizing the number of users experiencing power outages due to maintenance, minimizing the grid operation risk value, and maximizing the balance of maintenance plan execution. A pre-trained large language model is used to perform semantic understanding and reasoning on the operating status data, multi-objective constraints and optimization objectives to generate several candidate maintenance plans. The large language model is fine-tuned through historical maintenance cases and knowledge of power grid operation experts. Based on a multi-objective optimization algorithm, the candidate maintenance plans are sorted non-dominated and congestion is calculated to select the Pareto front solution set. The optimal maintenance plan is then determined from the Pareto front solution set according to the preset decision preference. Output the optimal maintenance plan and generate auxiliary decision instructions based on the optimal maintenance plan.

[0006] Optionally, the multi-objective constraint further includes: The minimum and maximum allowable voltages of each node in the distribution network are determined as node voltage constraints. The maximum allowable continuous current carrying capacity of each branch is determined as the branch current constraint condition. Determine the maximum allowable power outage duration and the maximum allowable number of users experiencing power outages in each power supply area as constraints on power supply reliability. The upper limit of the number of maintenance tasks that can be carried out simultaneously within the same time period and the upper limit of the number of spare parts required are determined as maintenance resource constraints. The maximum capacity that the load can be transferred to an adjacent feeder via the tie switch under fault or maintenance conditions is determined as the load transfer capacity constraint.

[0007] Optionally, the generation of several candidate maintenance plan schemes using a pre-trained large language model includes: The running status data, multi-objective constraints, and optimization objectives are constructed as input text according to a preset prompt word template; The input text is fed into the large language model, and the timing arrangement of maintenance tasks, the division of power outage areas, and the selection of load transfer paths are gradually generated through thought chain reasoning. The temperature parameters of the large language model are controlled to generate multiple diverse candidate maintenance plans, each accompanied by its reasoning and expected performance evaluation.

[0008] Optionally, the selection of the Pareto front solution set based on the multi-objective optimization algorithm includes: Using the optimization objective as multiple objective functions, calculate the objective function vector for each candidate maintenance plan scheme; The candidate maintenance plan is divided into different Pareto front levels using a fast non-dominated sorting method. For schemes within the same frontier level, calculate the crowding distance and prioritize retaining schemes with larger crowding distances to ensure the diversity of the solution set; When the number of candidate solutions exceeds a preset threshold, the selection is truncated based on the Pareto front level and the distance to congestion.

[0009] Optionally, determining the optimal maintenance plan from the Pareto front solution set based on preset decision preferences includes: Obtain the weight coefficients or priority ranking of multiple optimization objectives input by the scheduler or preset by the system; Based on the aforementioned weighting coefficients, a multi-attribute decision-making method is used to calculate the comprehensive utility value for each scheme in the Pareto front solution set; The scheme with the highest overall utility value is selected as the optimal maintenance plan.

[0010] Optionally, the method further includes: During the execution of the optimal maintenance plan, the changes in the operating status of the distribution network are monitored in real time. When the operating status deviates from the preset safety boundary or a new emergency occurs, the dynamic rescheduling process is triggered, and the method described above is re-executed to generate an adjusted maintenance plan.

[0011] Secondly, the present invention provides a distribution network maintenance planning auxiliary decision-making system based on multi-objective constraint optimization, the system comprising: The data acquisition module is used to collect real-time operating status data of the distribution network and maintenance task information to be executed. The operating status data includes the voltage of each node, the current of each branch, the transformer load rate, the output of distributed power sources, and the load forecast value. The maintenance task information includes the maintenance equipment identification, maintenance time window, maintenance priority, and the required power outage range. The constraint and objective configuration module is used to determine the multi-objective constraints and optimization objectives of the operating status data based on the network topology, equipment rated parameters, and safe operation procedures of the distribution network. The multi-objective constraints include at least node voltage constraints, branch current constraints, power supply reliability constraints, maintenance resource constraints, and load transfer capability constraints. The optimization objectives include minimizing the number of users experiencing power outages due to maintenance, minimizing the grid operation risk value, and maximizing the balance of maintenance plan execution. The large model reasoning module is used to perform semantic understanding and reasoning on the operating status data, multi-objective constraints and optimization objectives using a pre-trained large language model, and generate several candidate maintenance plans. The large language model is fine-tuned through historical maintenance cases and power grid operation expert knowledge. The multi-objective optimization module is used to perform non-dominated sorting and congestion calculation on the several candidate maintenance plans based on the multi-objective optimization algorithm, screen out the Pareto front solution set, and determine the optimal maintenance plan from the Pareto front solution set according to the preset decision preference. The output and decision module is used to output the optimal maintenance plan and generate auxiliary decision instructions based on the optimal maintenance plan.

[0012] Optionally, the large model inference module includes: The prompt word construction unit is used to construct the running status data, multi-objective constraints, and optimization objectives into input text according to a preset prompt word template; The reasoning generation unit is used to send the input text into the large language model and gradually generate the timing arrangement of maintenance tasks, the division of power outage areas, and the selection of load transfer paths through reasoning through thought chain. A diversity control unit is used to control the temperature parameters of the large language model to generate multiple diverse candidate maintenance plans, each accompanied by its reasoning and expected effect evaluation.

[0013] Optionally, the multi-objective optimization module includes: The objective function calculation unit is used to calculate the objective function vector for each candidate maintenance plan scheme, taking the optimization objective as multiple objective functions. Non-dominated sorting units are used to divide candidate maintenance plans into different Pareto front levels using fast non-dominated sorting. The congestion calculation unit is used to calculate the congestion distance of schemes within the same frontier level, and prioritizes retaining schemes with large congestion distances to ensure the diversity of the solution set; The truncation selection unit is used to truncate the selection based on the Pareto front level and the distance to congestion when the number of candidate solutions exceeds a preset threshold.

[0014] Optionally, the system further includes a dynamic rescheduling module, which is used to monitor changes in the operating status of the distribution network in real time during the execution of the optimal maintenance plan. When the operating status deviates from the preset safety boundary or a new emergency occurs, the dynamic rescheduling process is triggered to re-call the data acquisition module, constraint and target configuration module, large model inference module and multi-objective optimization module to generate an adjusted maintenance plan.

[0015] Through the above technical solutions, embodiments of the present invention provide a method and system for auxiliary decision-making in distribution network maintenance planning based on multi-objective constraint optimization. This method and system, by introducing a large language model, can fully utilize historical maintenance cases and power grid operation expert knowledge to achieve intelligent reasoning and generation of candidate maintenance plan schemes from operational data and constraints described in natural language. Simultaneously, by combining a multi-objective optimization algorithm to perform Pareto screening on candidate schemes, it effectively handles multiple conflicting objectives such as the number of power outage users, operational risk values, and plan balance. This method and system can be applied to offline maintenance plan preparation, providing dispatchers with optimized reference schemes, and can also be applied to online dynamic adjustment, quickly generating rescheduling strategies when the power grid state changes abruptly, significantly improving the intelligence level and decision-making efficiency of distribution network maintenance plan formulation.

[0016] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a distribution network maintenance plan auxiliary decision-making method based on multi-objective constraint optimization according to an embodiment of the present invention; Figure 2 This is a flowchart of the steps for generating several candidate maintenance plan schemes using a pre-trained large language model according to an embodiment of the present invention. Figure 3 This is a flowchart of the steps for screening the Pareto front solution set based on a multi-objective optimization algorithm according to an embodiment of the present invention; Figure 4 This is a flowchart illustrating the steps of determining the optimal maintenance plan from the Pareto front solution set according to a preset decision preference, based on one embodiment of the present invention. Figure 5 This is a structural diagram of a distribution network maintenance planning auxiliary decision-making system based on multi-objective constraint optimization according to an embodiment of the present invention. Detailed Implementation

[0018] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0019] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application all comply with relevant laws and regulations. In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0020] like Figure 1 The diagram shows a flowchart of a power distribution network maintenance plan auxiliary decision-making method based on multi-objective constraint optimization according to an embodiment of the present invention. Figure 1 In this context, the method may include: In step S1, real-time operational status data of the distribution network and information on maintenance tasks to be executed are collected. In one embodiment of the invention, real-time operational data of the 10kV and below distribution network is acquired every 5 minutes through the data acquisition and monitoring control interface of the distribution automation system. The operational status data includes: voltage at each node, current in each branch, transformer load rate, active and reactive power output of distributed generation sources, and load forecasts for the next 24 hours. Simultaneously, information on maintenance tasks to be executed is obtained from the production management information system. Each maintenance task includes: equipment identification for maintenance, suggested maintenance time window, maintenance priority, and required power outage area.

[0021] In step S2, multi-objective constraints and optimization objectives are determined based on the network topology, equipment rated parameters, and safe operation procedures of the distribution network. In one embodiment of the invention, the system first reads the node-branch topology, current carrying capacity limits of each line, rated capacity of each transformer, and allowable voltage range of each node from the power grid model database. Then, constraints are set according to the "Distribution Network Operation Procedures": node voltage constraint is ±7% of rated voltage; branch current constraint is 80% of rated current carrying capacity as a normal operation warning value; power supply reliability constraint requires that the duration of power outage caused by a single maintenance does not exceed 4 hours, and the cumulative number of users affected by power outages on a single feeder does not exceed 500; maintenance resource constraint stipulates that a maximum of 3 maintenance tasks can be carried out simultaneously within the same time period, and the required number of spare parts does not exceed 80% of the warehouse inventory; load transfer capacity constraint is determined based on the thermal stability limit of the tie line where the tie switch is located, for example, the maximum transfer capacity of a tie line is 3MVA. The optimization objectives are set as follows: objective one, minimize the total number of users affected by power outages caused by maintenance; objective two, minimize the power grid operation risk value; objective three, maximize the balance of maintenance plan execution.

[0022] In step S3, several candidate maintenance plans are generated using a pre-trained large language model. In one embodiment of the invention, the LLaMA-3-70B base model is selected, and the model is fine-tuned using 2000 real distribution network maintenance cases from the past three years and 500 power grid operation expert rules. The data obtained in steps S1 and S2 are constructed as input text according to the prompt word template.

[0023] In step S4, a Pareto front solution set is screened and the optimal solution is determined based on a multi-objective optimization algorithm. In a preferred embodiment of the invention, a fast non-dominated sorting genetic algorithm framework with an elitist strategy is used to further optimize 10 candidate solutions. First, the objective vector of each solution is calculated using the number of power outage users, the operational risk value, and the reciprocal of the planned balance as three objective functions. Then, non-dominated sorting is performed: any two solutions A and B are compared. If all objective values ​​of A are not inferior to B and at least one objective is superior to B, then A dominates B. After sorting, the solutions are divided into the first front (Pareto front), the second front, etc. For solutions within the same front, the congestion distance is calculated. After sorting each objective function, the difference between adjacent solutions on that objective is normalized and summed. Solutions with low front levels and high congestion are retained. If the number of solutions exceeds 20, they are truncated according to the aforementioned rules. Next, the weight coefficients set by the dispatcher through the human-machine interface are obtained, and the comprehensive utility value of each Pareto front solution is calculated using a linear weighted method. The solution with the highest comprehensive utility value is selected as the optimal maintenance plan.

[0024] In step S5, the optimal maintenance plan is output and auxiliary decision-making instructions are generated. The system displays the optimal plan in Gantt chart format on the dispatching dashboard, including: the start and end times of each maintenance task, the operation sequence, the expected number of users experiencing power outages, the expected risk value, and the plan balance. Simultaneously, a structured instruction file is generated and sent to the distribution automation system, supporting one-click execution or item-by-item confirmation.

[0025] Furthermore, in a preferred embodiment of the present invention, the multi-objective constraints may further include determining the minimum and maximum allowable voltages of each node in the distribution network as node voltage constraints, determining the maximum allowable continuous current carrying capacity of each branch as branch current constraints, determining the maximum allowable outage duration and the maximum allowable number of outage users in each power supply area as power supply reliability constraints, determining the upper limit of the number of maintenance tasks that can be carried out simultaneously within the same time period and the upper limit of the number of spare parts required as maintenance resource constraints, and determining the maximum capacity of load that can be transferred to adjacent feeders through tie switches under fault or maintenance conditions as load transfer capacity constraints. All of the above constraints are stored in the constraint library in the form of mathematical inequalities and are verified during the optimization process in steps S3 and S4.

[0026] like Figure 2 The diagram shows a flowchart illustrating the generation of several candidate maintenance plans using a pre-trained large language model, according to one embodiment of the present invention. Figure 2 In this context, the process may include: In step S21, the input text is constructed. The running status data, multi-objective constraints, and optimization objectives are constructed as input text according to a preset prompt word template.

[0027] In step S22, the thought chain reasoning process is performed. The input text is fed into the large language model, and through thought chain reasoning, the timing of maintenance tasks, the division of power outage areas, and the selection of load transfer paths are generated step by step. The model finally outputs a complete maintenance plan.

[0028] In step S23, temperature parameters are controlled to generate diverse solutions. The temperature parameters of the large language model are controlled to generate multiple diverse candidate maintenance plan solutions. Each output solution is accompanied by a text block containing its reasoning basis, making it easier for the dispatcher to understand the solution logic.

[0029] like Figure 3 The diagram shows a flowchart of the steps for selecting the Pareto front solution set based on a multi-objective optimization algorithm according to an embodiment of the present invention. Figure 3 In this context, the process may include: In step S31, the objective function vector is calculated. In one embodiment of the present invention, the candidate solution set is assumed to be... Where N is the number of candidate solutions. For each solution... Calculate three target values: Number of users experiencing power outages; Operating risk value, ,in, As weight, Let j be the risk indicator; The factor is the reciprocal of the balance degree, which is defined as the variance of the number of maintenance tasks in each time period. The larger the variance, the smaller the balance degree. Therefore, minimizing this factor is crucial. This is equivalent to maximizing the balance. The final result is a vector. .

[0030] In step S32, fast non-dominated sorting is performed. The dominance count and dominated set of all solutions are initialized. For each pair of solutions... ,like Each component is no greater than And at least one component is less than Then it is called Dominate After traversing all pairs, the solutions that are not dominated by any other solution are assigned to the first frontier. Then ignore The above process is repeated for the remaining schemes to obtain the solution. wait.

[0031] In step S33, the crowding distance is calculated. For schemes within the same frontier, each objective function value is sorted in ascending order. The crowding of the first and last schemes is set to infinity, and the crowding of the middle scheme is the difference between the two adjacent schemes on that objective divided by the difference between the maximum and minimum values ​​of that objective. The sum of the crowding of the three objectives is the crowding distance of the scheme. Schemes with large crowding distances are retained to maintain the diversity of solutions.

[0032] In step S34, the selection is truncated. When the total number of candidate solutions exceeds a preset threshold, solutions within the preset threshold number are retained as the final Pareto front solution set, according to the principle of Pareto front hierarchy priority and priority of greater crowding distance within the same hierarchy.

[0033] like Figure 4 The diagram shows a flowchart illustrating the steps of determining the optimal maintenance plan from a Pareto front solution set based on preset decision preferences, according to one embodiment of the present invention. Figure 4 In this process, the steps include: In step S41, the weight coefficients or priority rankings of multiple optimization objectives input by the scheduler or preset by the system are obtained.

[0034] In step S42, the comprehensive utility value is calculated for each scheme in the Pareto front solution set according to formula (1). After obtaining the weights, for each scheme s in the Pareto front solution set, the three objective values ​​are first normalized to obtain... Then, calculate the overall utility value according to formula (1).

[0035] in, This represents the overall utility value.

[0036] In step S43, the scheme with the highest overall utility value is selected as the optimal maintenance plan. If multiple schemes have the same utility value, their congestion distance is further compared, and the scheme with the larger congestion distance is selected.

[0037] Furthermore, in a preferred embodiment of the present invention, changes in the operating status of the distribution network are monitored in real time during the execution of the optimal maintenance plan. When the operating status deviates from the preset safety boundary or a new emergency occurs, a dynamic rescheduling process is triggered to re-execute the above method to generate an adjusted maintenance plan.

[0038] Secondly, this invention also provides a distribution network maintenance planning auxiliary decision-making system based on multi-objective constraint optimization. For example... Figure 5 The diagram shown is a structural diagram of a distribution network maintenance planning auxiliary decision-making system based on multi-objective constraint optimization according to an embodiment of the present invention. Figure 5The system may include a data acquisition module, a constraint and objective configuration module, a large-scale model inference module, a multi-objective optimization module, and an output and decision module. The data acquisition module collects real-time operational status data of the distribution network and information on maintenance tasks to be performed. Operational status data includes node voltage, branch current, transformer load rate, distributed generation output, and load forecast values. Maintenance task information includes maintenance equipment identification, maintenance time window, maintenance priority, and required outage range. The constraint and objective configuration module determines multi-objective constraints and optimization objectives based on the distribution network topology, equipment rated parameters, and safe operation procedures. Multi-objective constraints include at least node voltage constraints, branch current constraints, power supply reliability constraints, maintenance resource constraints, and load transfer capability constraints. Optimization objectives include minimizing the number of users experiencing power outages due to maintenance, minimizing the grid operation risk value, and maximizing the balance of maintenance plan execution. The large-scale model inference module utilizes a pre-trained large-scale language model to perform semantic understanding and inference on operational status data, multi-objective constraints, and optimization objectives, generating several candidate maintenance plans. The large-scale language model is fine-tuned using historical maintenance cases and power grid operation expert knowledge. The multi-objective optimization module performs non-dominated ranking and congestion calculation on the candidate maintenance plans based on a multi-objective optimization algorithm, selecting a Pareto front solution set and determining the optimal maintenance plan from the Pareto front solution set according to preset decision preferences. The output and decision module outputs the optimal maintenance plan and generates auxiliary decision-making instructions based on it.

[0039] Specifically, the large-scale model's inference module includes a prompt word construction unit, an inference generation unit, and a diversity control unit. The prompt word construction unit constructs input text from operational status data, multi-objective constraints, and optimization objectives according to a preset prompt word template. The inference generation unit feeds the input text into the large language model, which then uses thought chain reasoning to progressively generate the timing arrangement of maintenance tasks, the division of power outage areas, and the selection of load transfer paths. The diversity control unit controls the temperature parameters of the large language model to generate multiple diverse candidate maintenance plans, each accompanied by its reasoning basis and expected effect evaluation.

[0040] Furthermore, this multi-objective optimization module includes an objective function calculation unit, a non-dominated sorting unit, a congestion calculation unit, and a truncation selection unit. The objective function calculation unit calculates the objective function vector for each candidate maintenance plan, using the optimization objective as multiple objective functions. The non-dominated sorting unit uses fast non-dominated sorting to divide the candidate maintenance plans into different Pareto front levels. The congestion calculation unit calculates the congestion distance between plans within the same front level, prioritizing plans with larger congestion distances to ensure solution diversity. The truncation selection unit performs truncation selection based on the Pareto front level and congestion distance when the number of candidate plans exceeds a preset threshold.

[0041] Furthermore, in a preferred embodiment of the present invention, the auxiliary decision-making system may further include a dynamic rescheduling module, which is used to monitor changes in the operating status of the distribution network in real time during the execution of the optimal maintenance plan. When the operating status deviates from the preset safety boundary or a new emergency occurs, the dynamic rescheduling process is triggered to re-call the data acquisition module, constraint and target configuration module, large model inference module and multi-objective optimization module to generate an adjusted maintenance plan.

[0042] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0043] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0044] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0045] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0046] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0047] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0048] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0049] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0050] 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. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for auxiliary decision-making in distribution network maintenance planning based on multi-objective constraint optimization, characterized in that, The method includes: Real-time acquisition of distribution network operation status data and maintenance task information to be executed. The operation status data includes the voltage of each node, the current of each branch, the transformer load rate, the output of distributed power sources and the load forecast value. The maintenance task information includes the maintenance equipment identification, maintenance time window, maintenance priority and the required power outage range. Based on the network topology, equipment rated parameters, and safe operation procedures of the distribution network, the multi-objective constraints and optimization objectives of the operating status data are determined. The multi-objective constraints include at least node voltage constraints, branch current constraints, power supply reliability constraints, maintenance resource constraints, and load transfer capability constraints. The optimization objectives include minimizing the number of users experiencing power outages due to maintenance, minimizing the grid operation risk value, and maximizing the balance of maintenance plan execution. A pre-trained large language model is used to perform semantic understanding and reasoning on the operating status data, multi-objective constraints and optimization objectives to generate several candidate maintenance plans. The large language model is fine-tuned through historical maintenance cases and knowledge of power grid operation experts. Based on a multi-objective optimization algorithm, the candidate maintenance plans are sorted non-dominated and congestion is calculated to select the Pareto front solution set. The optimal maintenance plan is then determined from the Pareto front solution set according to the preset decision preference. Output the optimal maintenance plan and generate auxiliary decision instructions based on the optimal maintenance plan.

2. The method according to claim 1, characterized in that, The multi-objective constraints also include: The minimum and maximum allowable voltages of each node in the distribution network are determined as node voltage constraints. The maximum allowable continuous current carrying capacity of each branch is determined as the branch current constraint condition. Determine the maximum allowable power outage duration and the maximum allowable number of users experiencing power outages in each power supply area as constraints on power supply reliability. The upper limit of the number of maintenance tasks that can be carried out simultaneously within the same time period and the upper limit of the number of spare parts required are determined as maintenance resource constraints. The maximum capacity that the load can be transferred to an adjacent feeder via the tie switch under fault or maintenance conditions is determined as the load transfer capacity constraint.

3. The method according to claim 1, characterized in that, The method of generating several candidate maintenance plans using a pre-trained large language model includes: The running status data, multi-objective constraints, and optimization objectives are constructed as input text according to a preset prompt word template; The input text is fed into the large language model, and the timing arrangement of maintenance tasks, the division of power outage areas, and the selection of load transfer paths are gradually generated through thought chain reasoning. The temperature parameters of the large language model are controlled to generate multiple diverse candidate maintenance plans, each accompanied by its reasoning and expected performance evaluation.

4. The method according to claim 1, characterized in that, The method for selecting the Pareto front solution set based on multi-objective optimization algorithm includes: Using the optimization objective as multiple objective functions, calculate the objective function vector for each candidate maintenance plan scheme; The candidate maintenance plan is divided into different Pareto front levels using a fast non-dominated sorting method. For schemes within the same frontier level, calculate the crowding distance and prioritize retaining schemes with larger crowding distances to ensure the diversity of the solution set; When the number of candidate solutions exceeds a preset threshold, the selection is truncated based on the Pareto front level and the distance to congestion.

5. The method according to claim 1, characterized in that, The process of determining the optimal maintenance plan from the Pareto front solution set based on preset decision preferences includes: Obtain the weight coefficients or priority ranking of multiple optimization objectives input by the scheduler or preset by the system; Based on the aforementioned weighting coefficients, a multi-attribute decision-making method is used to calculate the comprehensive utility value for each scheme in the Pareto front solution set; The scheme with the highest overall utility value is selected as the optimal maintenance plan.

6. The method according to claim 1, characterized in that, The method further includes: During the execution of the optimal maintenance plan, the changes in the operating status of the distribution network are monitored in real time. When the operating status is detected to deviate from the preset safety boundary or a new emergency occurs, the dynamic rescheduling process is triggered to re-execute the method described in claim 1 to generate an adjusted maintenance plan.

7. A distribution network maintenance planning auxiliary decision-making system based on multi-objective constraint optimization, characterized in that, The system includes: The data acquisition module is used to collect real-time operating status data of the distribution network and maintenance task information to be executed. The operating status data includes the voltage of each node, the current of each branch, the transformer load rate, the output of distributed power sources, and the load forecast value. The maintenance task information includes the maintenance equipment identification, maintenance time window, maintenance priority, and the required power outage range. The constraint and objective configuration module is used to determine the multi-objective constraints and optimization objectives of the operating status data based on the network topology, equipment rated parameters, and safe operation procedures of the distribution network. The multi-objective constraints include at least node voltage constraints, branch current constraints, power supply reliability constraints, maintenance resource constraints, and load transfer capability constraints. The optimization objectives include minimizing the number of users experiencing power outages due to maintenance, minimizing the grid operation risk value, and maximizing the balance of maintenance plan execution. The large model reasoning module is used to perform semantic understanding and reasoning on the operating status data, multi-objective constraints and optimization objectives using a pre-trained large language model, and generate several candidate maintenance plans. The large language model is fine-tuned through historical maintenance cases and power grid operation expert knowledge. The multi-objective optimization module is used to perform non-dominated sorting and congestion calculation on the several candidate maintenance plans based on the multi-objective optimization algorithm, screen out the Pareto front solution set, and determine the optimal maintenance plan from the Pareto front solution set according to the preset decision preference. The output and decision module is used to output the optimal maintenance plan and generate auxiliary decision instructions based on the optimal maintenance plan.

8. The system according to claim 7, characterized in that, The large model inference module includes: The prompt word construction unit is used to construct the running status data, multi-objective constraints, and optimization objectives into input text according to a preset prompt word template; The reasoning generation unit is used to send the input text into the large language model and gradually generate the timing arrangement of maintenance tasks, the division of power outage areas, and the selection of load transfer paths through reasoning through thought chain. A diversity control unit is used to control the temperature parameters of the large language model to generate multiple diverse candidate maintenance plans, each accompanied by its reasoning and expected effect evaluation.

9. The system according to claim 7, characterized in that, The multi-objective optimization module includes: The objective function calculation unit is used to calculate the objective function vector for each candidate maintenance plan scheme, taking the optimization objective as multiple objective functions. Non-dominated sorting units are used to divide candidate maintenance plans into different Pareto front levels using fast non-dominated sorting. The congestion calculation unit is used to calculate the congestion distance of schemes within the same frontier level, and prioritizes retaining schemes with large congestion distances to ensure the diversity of the solution set; The truncation selection unit is used to truncate the selection based on the Pareto front level and the distance to congestion when the number of candidate solutions exceeds a preset threshold.

10. The system according to claim 7, characterized in that, The system also includes a dynamic rescheduling module, which is used to monitor changes in the operating status of the distribution network in real time during the execution of the optimal maintenance plan. When the operating status deviates from the preset safety boundary or a new emergency occurs, the dynamic rescheduling process is triggered, and the data acquisition module, constraint and target configuration module, large model inference module and multi-objective optimization module are called back to generate an adjusted maintenance plan.

Images