Power distribution network feeder maintenance planning method and device, equipment and storage medium

Abstract

This invention relates to the field of distribution network operation and maintenance technology, and discloses a method, device, equipment, and storage medium for distribution network feeder maintenance planning, used to achieve flexible decision-making on maintenance schemes. The distribution network feeder maintenance planning method includes: acquiring historical operating data of the distribution network; constructing a probability distribution model reflecting the fluctuation characteristics of reliability indicators and maintenance costs based on the historical operating data; screening dominant outage fault causes and matching maintenance actions according to the probability distribution model and risk preference parameters to generate an initial maintenance scheme; obtaining a Pareto optimal scheme set through multi-criteria optimization algorithm iteration; selecting schemes that balance system and local reliability from the Pareto optimal scheme set and performing sensitivity analysis to obtain the target maintenance planning scheme. This invention ensures the scientific nature of the final decision and overcomes the limitations of existing technologies that only output a single scheme and are difficult to adapt to diverse engineering scenarios, significantly improving the flexibility and robustness of maintenance planning.

Description

Technical Field

[0001] This invention relates to the field of power distribution network operation and maintenance technology, and in particular to a method, device, equipment and storage medium for planning the maintenance of power distribution network feeders. Background Technology

[0002] Existing methods for distribution network feeder maintenance planning typically employ a fuzzy multi-criteria decision-making approach. This approach first processes historical feeder fault records using a fault data statistics module, outputting definite values ​​such as "single feeder fault frequency" and "average outage duration per feeder." Then, maintenance personnel subjectively set the weights for two criteria: "fault impact" and "maintenance cost," and use a ranking method to prioritize all feeders, outputting a maintenance priority list. Finally, a single maintenance plan is output through a maintenance plan output module.

[0003] However, the above methods still have certain limitations in practical applications: the fault data statistics module only outputs the definite values ​​of parameters, without considering the uncertainty of parameters caused by factors such as equipment aging and environmental fluctuations, which makes the decision-making basis deviate significantly from the actual working conditions; in addition, the maintenance plan output module only generates a single maintenance plan, which makes it difficult to clarify the applicable boundaries of the plan under different budgets and risk preferences, thus failing to support the flexible and ever-changing maintenance decision-making needs in actual engineering. Summary of the Invention

[0004] This invention provides a method, apparatus, equipment, and storage medium for distribution network feeder maintenance planning, in order to solve the problem in the prior art that the decision-making basis is detached from the actual working conditions and cannot be adapted to diverse engineering scenarios due to ignoring parameter uncertainties and only outputting a single maintenance scheme.

[0005] The technical solution provided by this invention is a method for maintenance planning of power distribution network feeders, comprising: acquiring historical operating data of the power distribution network, including fault records, outage records, and equipment operation and maintenance data; constructing a probability distribution model reflecting the fluctuation characteristics of reliability indicators and maintenance costs based on the historical operating data; selecting the dominant fault causes of power outages and matching maintenance actions according to the probability distribution model and preset risk preference parameters to generate an initial maintenance plan; obtaining a Pareto optimal solution set through multi-criteria optimization algorithm iteration based on the initial maintenance plan, wherein the iteration process revolves around the objectives of maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs; selecting a solution that balances system and local reliability from the Pareto optimal solution set, and performing sensitivity analysis on the selected solution based on the probability distribution model to obtain a target maintenance planning solution.

[0006] Furthermore, the step of constructing a probability distribution model reflecting the fluctuation characteristics of reliability indicators and maintenance costs based on the historical operating data includes: calculating the time series value of the system's average power outage frequency based on the number of fault occurrences counted by feeder in the fault records and the number of affected users recorded in the outage log; fitting a normal distribution to this time series value using the maximum likelihood estimation method to obtain the mean and standard deviation, and calculating the confidence interval to construct a normal distribution model of the system's average power outage frequency; fitting a Weibull distribution to the fault repair time data classified by fault type in the fault records using the least squares method to obtain the shape parameter and scale parameter, and calculating the confidence interval to construct a Weibull distribution model of the fault repair time; and extracting the minimum, most likely, and maximum costs of various maintenance actions from the single maintenance cost records classified by maintenance type in the equipment operation and maintenance data to construct a triangular distribution model of single maintenance costs.

[0007] Furthermore, the step of selecting the dominant power outage fault causes and matching maintenance actions based on the probability distribution model and preset risk preference parameters to generate an initial maintenance plan includes: Based on the probability distribution model, extracting reliability index interval data for each feeder-fault cause combination, and converting the interval data into deterministic values ​​using preset risk preference parameters; ranking the power outage impact of feeder-fault cause combinations using the interval TOPSIS method; based on the ranking results, applying the Pareto principle to select a list of dominant power outage fault causes, and matching corresponding maintenance actions for each dominant fault cause according to preset matching rules; based on the single maintenance cost distribution model in the probability distribution model, calculating the cost interval of each maintenance action in conjunction with the maintenance length, converting it into deterministic values ​​according to the risk preference parameters, constructing an evaluation matrix, and then ranking the benefit-cost ratio again using the interval TOPSIS method to obtain the initial maintenance plan.

[0008] Furthermore, based on the probability distribution model, reliability index interval data for each feeder-fault cause combination is extracted, and the interval data is transformed into deterministic values ​​by combining preset risk preference parameters. The power outage impact of the feeder-fault cause combination is ranked using the interval TOPSIS method, including: based on the normal distribution model of the system average power outage frequency and the Weibull distribution model of the fault repair time in the probability distribution model, preset confidence level confidence intervals for system-level average power outage frequency, local-level average power outage frequency, system-level average power outage duration, and local-level average power outage duration are extracted respectively, and classified and statistically analyzed according to feeder-fault causes; the interval TOPSIS method is used to rank the classified feeder-fault cause combinations, first performing interval normalization on benefit-type indicators and cost-type indicators respectively, then performing weighted normalization, and then calculating the Euclidean distance from each evaluation object to the ideal solution and the anti-ideal solution; the relative proximity of each feeder-fault cause combination is calculated based on the Euclidean distance, and the combinations are ranked from largest to smallest according to the relative proximity value to obtain the power outage impact ranking result of the feeder-fault cause combination.

[0009] Furthermore, based on the single maintenance cost distribution model in the probability distribution model, the cost range of each maintenance action is calculated in combination with the maintenance length. This range is then converted into a definite value based on the risk preference parameter. After constructing an evaluation matrix, the benefit-cost ratio is ranked again using the interval TOPSIS method to select the initial maintenance plan. This includes: extracting interval data of unit length maintenance cost from the triangular distribution model of single maintenance cost based on the selected dominant fault causes and their matching maintenance actions, and calculating the total cost range of each maintenance action in combination with the maintenance length of each feeder; and calculating the impact of each maintenance action on the system based on the reliability index interval data in the probability distribution model and the preset maintenance effectiveness factor interval. The benefit ranges for system-level reliability and local-level reliability are determined, and the cost range and benefit range are converted into cost-determined values ​​and benefit-determined values ​​respectively by combining preset risk preference parameters. An evaluation matrix is ​​constructed with feeder-maintenance actions as rows and system-level benefits, local-level benefits and costs as columns. The evaluation matrix is ​​sorted by benefit-cost ratio using the interval TOPSIS method, and the relative proximity of each feeder-maintenance action combination is calculated. The combinations are filtered in descending order of relative proximity value, and the cumulative value of the total cost range is used as the basis for judging budget constraints during the filtering process. Under the preset budget constraints or quantity constraints, an initial maintenance plan containing feeder-maintenance action combinations and their cost and benefit-determined values ​​is obtained.

[0010] Furthermore, based on the initial maintenance scheme, a Pareto optimal solution set is obtained through multi-criteria optimization algorithm iteration. The iteration process revolves around maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs. This includes: using the initial maintenance scheme as the current solution, initializing the control parameters of the multi-criteria simulated annealing algorithm, and setting the three objective functions: maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs. The control parameters include initial temperature, cooling coefficient, termination temperature, and maximum number of iterations. A neighborhood perturbation operation is performed on the current solution, generating candidate solutions through three methods: adding maintenance actions, removing maintenance actions, or replacing maintenance actions. The feasibility of the candidate solutions is verified to ensure that their cumulative maintenance cost does not exceed the preset budget limit and covers the preset proportion of dominant failure causes. The cost and benefit ranges of each maintenance action are converted into deterministic values ​​based on preset risk preference parameters. The merits of the candidate solutions compared to the current solution are judged based on Pareto dominance. If the candidate solution dominates the current solution, it is accepted as the new current solution; otherwise, a non-dominated candidate solution is accepted as the new current solution with a certain probability according to the probability acceptance mechanism of simulated annealing. The temperature is updated according to the cooling coefficient, and the iteration continues until the termination condition is met. All accepted feasible solutions are collected during the iteration process, and Pareto optimal solutions are selected through non-dominated sorting, forming a Pareto optimal solution set.

[0011] Furthermore, the step of selecting schemes that balance system and local reliability from the Pareto optimal solution set, and performing sensitivity analysis on the selected schemes based on the probability distribution model to obtain the target maintenance planning scheme, includes: assigning preset balance weights to system-level reliability indicators and local-level reliability indicators; constructing a comprehensive reliability evaluation index for each scheme in the Pareto optimal solution set based on the balance weights; and selecting the scheme with the optimal comprehensive reliability evaluation index as a candidate scheme; adjusting the parameters of the reliability indicator range, maintenance cost range, and maintenance effectiveness factor range in the candidate scheme based on the probability distribution model; verifying the change range of the comprehensive reliability evaluation index of the candidate scheme under different parameter fluctuation scenarios; evaluating the adaptability of the candidate scheme to parameter fluctuations based on the verification results under different parameter fluctuation scenarios; and selecting the scheme with the smallest fluctuation range of the comprehensive reliability evaluation index under a preset confidence level as the target maintenance planning scheme output.

[0012] Another technical solution provided by the present invention: a distribution network feeder maintenance planning device, comprising: an acquisition module for acquiring historical operating data of the distribution network, the historical operating data including fault records, outage records, and equipment operation and maintenance data; a construction module for constructing a probability distribution model reflecting the fluctuation characteristics of reliability indicators and maintenance costs based on the historical operating data; a generation module for selecting the dominant fault causes of power outages and matching maintenance actions according to the probability distribution model and preset risk preference parameters, and generating an initial maintenance plan; a processing module for obtaining a Pareto optimal solution set through multi-criteria optimization algorithm iteration based on the initial maintenance plan, wherein the iteration process revolves around the objectives of maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs; and a screening module for selecting solutions that balance system and local reliability from the Pareto optimal solution set, and performing sensitivity analysis on the selected solutions based on the probability distribution model to obtain a target maintenance planning solution.

[0013] Furthermore, the construction module is specifically used for: calculating the time series value of the system's average power outage frequency based on the number of fault occurrences counted by feeder in the fault records and the number of affected users recorded in the power outage ledger; fitting a normal distribution to this time series value using the maximum likelihood estimation method to obtain the mean and standard deviation, and calculating the confidence interval to construct a normal distribution model of the system's average power outage frequency; fitting a Weibull distribution to the fault repair time data classified by fault type in the fault records using the least squares method to obtain the shape and scale parameters, and calculating the confidence interval to construct a Weibull distribution model of the fault repair time; and extracting the minimum, most likely, and maximum costs of various maintenance actions from the single maintenance cost records classified by maintenance type in the equipment operation and maintenance data to construct a triangular distribution model of the single maintenance cost.

[0014] Furthermore, the generation module includes: a sorting unit, used to extract reliability index interval data for each feeder-fault cause combination based on the probability distribution model, and convert the interval data into deterministic values ​​in combination with preset risk preference parameters, and sort the feeder-fault cause combinations by power outage impact using the interval TOPSIS method; a matching unit, used to filter out the list of dominant power outage fault causes based on the sorting results using the Pareto principle, and match corresponding maintenance actions for each dominant fault cause according to preset matching rules; and a filtering unit, used to calculate the cost interval of each maintenance action based on the single maintenance cost distribution model in the probability distribution model, combined with the maintenance length, and convert it into deterministic values ​​according to the risk preference parameters, construct an evaluation matrix, and then sort the benefit-cost ratio again using the interval TOPSIS method to obtain the initial maintenance plan.

[0015] Furthermore, the sorting unit is specifically used for: extracting pre-set confidence intervals for system-level average power outage frequency, local-level average power outage frequency, system-level average power outage duration, and local-level average power outage duration based on the normal distribution model of the system average power outage frequency and the Weibull distribution model of the fault repair duration in the probability distribution model, and classifying and statistically analyzing them according to feeder-fault causes; sorting the classified feeder-fault cause combinations using the interval TOPSIS method, first performing interval normalization on benefit indicators and cost indicators respectively, then performing weighted normalization, and then calculating the Euclidean distance from each evaluation object to the ideal solution and the anti-ideal solution; calculating the relative proximity of each feeder-fault cause combination based on the Euclidean distance, and sorting them according to the relative proximity value from largest to smallest to obtain the power outage impact ranking result of the feeder-fault cause combination.

[0016] Furthermore, the screening unit is specifically used for: extracting interval data of unit length maintenance cost from the triangular distribution model of single maintenance cost based on the screened dominant fault causes and their matching maintenance actions, and calculating the total cost interval of each maintenance action in combination with the maintenance length of each feeder; calculating the benefit interval of each maintenance action on system-level reliability and local-level reliability based on the reliability index interval data in the probability distribution model and the preset maintenance effectiveness factor interval, and converting the cost interval and the benefit interval into cost determination value and benefit determination value respectively in combination with the preset risk preference parameter, constructing an evaluation matrix with feeder-maintenance action as rows and system-level benefit, local-level benefit and cost as columns; sorting the evaluation matrix by benefit-cost ratio using the interval TOPSIS method, calculating the relative closeness of each feeder-maintenance action combination, screening in descending order of relative closeness value, and using the cumulative value of the total cost interval as the basis for judging budget constraints during the screening process, and obtaining an initial maintenance plan containing feeder-maintenance action combinations and their cost and benefit determination values ​​under the preset budget constraint conditions or quantity constraints.

[0017] Furthermore, the processing module is specifically used to: take the initial maintenance scheme as the current solution, initialize the control parameters of the multi-criteria simulated annealing algorithm, and set the three objective functions: maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs. The control parameters include initial temperature, cooling coefficient, termination temperature, and maximum number of iterations. The module performs neighborhood perturbation operations on the current solution, generating candidate solutions through adding, removing, or replacing maintenance actions, and performs feasibility verification on the candidate solutions to ensure that their cumulative total maintenance cost does not exceed a preset budget limit and covers... The system identifies the dominant causes of failures based on a preset proportion; it then transforms the cost and benefit ranges of each maintenance action into deterministic values ​​using preset risk preference parameters. Based on the Pareto dominance relationship, it judges the merits of the candidate solution and the current solution. If the candidate solution dominates the current solution, it accepts the candidate solution as the new current solution; otherwise, it accepts the non-dominated candidate solution as the new current solution with a certain probability according to the probability acceptance mechanism of simulated annealing, and updates the temperature according to the cooling coefficient. The iteration continues until the termination condition is met. All accepted feasible solutions are collected during the iteration process, and Pareto optimal solutions are selected through non-dominated sorting. These solutions are then summarized to form a Pareto optimal solution set.

[0018] Furthermore, the screening module is specifically used for: assigning preset balance weights to system-level reliability indicators and local-level reliability indicators; constructing a comprehensive reliability evaluation index for each scheme in the Pareto optimal solution set based on the balance weights; and selecting the scheme with the optimal comprehensive reliability evaluation index as a candidate scheme; adjusting the parameters of the reliability indicator range, maintenance cost range, and maintenance effectiveness factor range in the candidate scheme based on the probability distribution model; verifying the change range of the comprehensive reliability evaluation index of the candidate scheme under different parameter fluctuation scenarios; evaluating the adaptability of the candidate scheme to parameter fluctuations based on the verification results under different parameter fluctuation scenarios; and selecting the scheme with the smallest fluctuation range of the comprehensive reliability evaluation index under the preset confidence level as the target maintenance planning scheme output.

[0019] Another technical solution provided by the present invention is an electronic device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the electronic device to execute the above-described distribution network feeder maintenance planning method.

[0020] Another technical solution provided by the present invention: a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the aforementioned distribution network feeder maintenance plan.

[0021] The technical solution provided by this invention involves acquiring historical operating data of the power distribution network, including fault records, outage records, and equipment operation and maintenance data; based on the historical operating data, constructing a probability distribution model reflecting the fluctuation characteristics of reliability indicators and maintenance costs; according to the probability distribution model and preset risk preference parameters, identifying the dominant fault causes of power outages and matching maintenance actions to generate an initial maintenance plan; based on the initial maintenance plan, obtaining a Pareto optimal solution set through multi-criteria optimization algorithm iteration, wherein the iteration process revolves around maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs; selecting a solution that balances system and local reliability from the Pareto optimal solution set, and performing sensitivity analysis on the selected solution based on the probability distribution model to obtain a target maintenance planning solution. In this embodiment of the invention, by constructing a probability distribution model that reflects the fluctuation characteristics of reliability indicators and maintenance costs, the uncertainty of parameters caused by factors such as equipment aging and environmental changes is effectively quantified, making the decision-making basis more in line with actual working conditions. On this basis, a Pareto optimal solution set containing multiple equilibrium solutions is generated iteratively through a multi-criteria optimization algorithm, providing decision-makers with a range of feasible choices under different budgets and risk preferences. Finally, the optimal recommended solution that balances system and local reliability is selected from this set, which not only ensures the scientific nature of the final decision, but also overcomes the limitations of existing technologies that only output a single solution and are difficult to adapt to diverse engineering scenarios, significantly improving the flexibility and robustness of maintenance planning. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of one embodiment of the distribution network feeder maintenance planning method in this invention;

[0023] Figure 2 This is a schematic diagram of another embodiment of the distribution network feeder maintenance planning method in this invention;

[0024] Figure 3 This is a schematic diagram of one embodiment of the power distribution network feeder maintenance planning device in this invention;

[0025] Figure 4 This is a schematic diagram of another embodiment of the power distribution network feeder maintenance planning device in this invention;

[0026] Figure 5 This is a schematic diagram of one embodiment of the electronic device in this invention;

[0027] Figure 6 This is a graph showing the maintenance cost range and the number of maintenance actions for each scheme in the embodiments of the present invention. Detailed Implementation

[0028] This invention provides a method, device, equipment, and storage medium for distribution network feeder maintenance planning. By constructing a probability distribution model to quantify parameter uncertainty, generating a Pareto optimal solution set, and screening balanced solutions, it achieves effective adaptation to parameter fluctuations and flexible decision-making on maintenance schemes.

[0029] The terms "first," "second," "third," "fourth," etc. (if present) 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 described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" or "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.

[0030] It is understood that the executing entity of this invention can be a power distribution network feeder maintenance planning device, a terminal, or a server; no specific limitation is made here. This embodiment of the invention will be described using a server as an example.

[0031] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the distribution network feeder maintenance planning method in this invention includes:

[0032] 101. Obtain historical operating data of the distribution network, including fault records, outage records, and equipment operation and maintenance data;

[0033] Multi-source data accumulated during the historical operation of the power distribution network is collected, including fault records, outage logs, and equipment maintenance records. Fault records cover the occurrence time, repair time, and affected area of ​​various fault events; outage logs detail the start and end times, affected areas, and causes of each outage; equipment maintenance data includes information on maintenance history, material consumption, and manpower costs. This data is then categorized and organized according to feeder number, fault cause, and maintenance action type.

[0034] 102. Based on historical operational data, construct a probability distribution model that reflects the fluctuation characteristics of reliability indicators and maintenance costs;

[0035] Because the operating conditions, failure probabilities, and maintenance costs of a power distribution network vary under different scenarios (such as different seasons, different load levels, and different equipment aging levels), a probability distribution model can be constructed to statistically analyze and simulate relevant indicators and costs under various possible scenarios.

[0036] Differentiated probability distribution models are used to quantify uncertainties for different types of reliability indicators and maintenance cost parameters. For the System Average Outage Frequency (SAIFI) indicator, the number of faults occurring on each feeder within the statistical period is extracted from fault records, and the time series value of SAIFI is calculated by combining it with the number of affected users in the outage ledger. After removing extreme disaster data, a normal distribution is used for fitting, and its probability density function is:

[0037]

[0038] The mean of the distribution is obtained by maximum likelihood estimation. (SAIFI average level) and standard deviation (The degree of volatility of SAIFI), and calculate the 95% confidence interval [ ]

[0039] For fault repair time, the time taken from the occurrence of each fault to the restoration of power is extracted from the fault records, classified according to type such as equipment fault and line fault, and modeled using Weibull distribution, with the probability density function as follows:

[0040]

[0041] Shape parameters are estimated using logarithmic transformation and least squares method. With scale parameters The Kolmogorov-Smirnov test was used to verify the fitting effect, thereby accurately describing the probability distribution of repair time under different fault types.

[0042] For the cost of a single maintenance operation, records of material and labor costs consumed by various maintenance actions are extracted from equipment operation and maintenance data. These costs are categorized by maintenance type, such as condition-based maintenance, tree pruning, and insulation covering. The minimum cost *a*, the most frequent typical cost *b*, and the maximum cost *c* for each maintenance action are statistically analyzed. A triangular distribution model is used, and its probability density function is:

[0043]

[0044] The consistency of the fit was verified by comparing the cumulative frequency curve of the actual cost data with the theoretical cumulative distribution curve.

[0045] Ultimately, this leads to the development of normal distribution models, Weibull distribution models, and triangular distribution models that are adapted to different parameter characteristics.

[0046] 103. Based on the probability distribution model and preset risk preference parameters, identify the main causes of power outages and match them with maintenance actions to generate an initial maintenance plan;

[0047] Setting risk preference parameters can reflect a decision-maker's risk tolerance in different scenarios. For example, during critical events or in scenarios with extremely high power supply reliability requirements, decision-makers may tend to have a lower risk preference and be willing to invest more resources in maintenance to reduce the risk of power outages; while in scenarios where cost is more sensitive and reliability requirements are relatively lower, they may choose a higher risk preference. By considering different risk preference parameters, the solution can generate initial maintenance plans that adapt to the needs of different scenarios.

[0048] The Hurwitz risk preference parameter is introduced to quantify decision-makers' attitudes toward uncertainty. This parameter ranges from zero to one; a higher value indicates a more optimistic preference, emphasizing the upper limit of benefits and the lower limit of costs in subsequent calculations, while a lower value emphasizes the lower limit of benefits and the upper limit of costs. Based on this, confidence intervals for system-level and local-level reliability indicators are extracted from the constructed normal distribution model of system average outage frequency and the Weibull distribution model of fault repair duration. After classification and statistics by feeder number and fault cause, the degree of outage impact of various feeder-fault cause combinations is ranked using the interval approximation ideal solution ranking method. This method uses interval normalization, weighted normalized interval calculation, and Euclidean distance calculation between ideal and anti-ideal solutions to obtain the relative closeness of each combination. Combined with the Pareto principle, a list of dominant fault causes contributing significantly to the outage impact is selected. For the identified primary causes of failure, targeted maintenance actions are assigned according to predefined matching rules, such as matching equipment failures to condition-based maintenance and tree contact to tree pruning. Cost records for each maintenance action are extracted from equipment operation and maintenance data. Combined with the established maintenance cost triangular distribution model and maintenance length parameters, the total cost range for each maintenance action is calculated. Simultaneously, a maintenance effectiveness factor is introduced to quantify the local and system-level reliability benefits after the action implementation. After converting the cost and benefit ranges into deterministic values ​​using Hurwitz risk preference parameters, an evaluation matrix is ​​constructed with feeder and maintenance action combinations as rows and system-level benefits, local-level benefits, and costs as columns. The interval approximation ideal solution ranking method is used again to calculate the relative proximity of each maintenance action and rank them according to the benefit-cost ratio. Finally, under a given budget constraint, the top-ranked maintenance actions are selected sequentially to generate an initial maintenance plan containing specific feeder action combinations and their quantified cost-benefit values.

[0049] The "feeder-failure cause" is sorted using the interval TOPSIS method (system / local indicator weights are set by the decision-maker), and interval normalization is performed first:

[0050] Benefit indicators:

[0051]

[0052]

[0053] in, and For the first The first evaluation object (feeder - cause of failure / maintenance action) The endpoints of the normalized intervals for each indicator; and These are the endpoints of the original index range; This represents the total number of evaluation objects.

[0054] Cost-related indicators:

[0055]

[0056]

[0057] Then perform weighted normalization:

[0058]

[0059]

[0060] in, and These are the endpoints of the weighted interval. For the first The weight of each indicator.

[0061] Next, calculate the separation (Euclidean distance):

[0062]

[0063]

[0064] in, To evaluate the distance from the object to the ideal solution, The distance to the antiideal solution.

[0065] Finally, calculate the relative closeness. :

[0066]

[0067] in, The larger the value, the higher it is ranked, and it is given priority in being included in the Pareto fault list or initial maintenance plan.

[0068] 104. Based on the initial maintenance plan, a Pareto optimal solution set is obtained through multi-criteria optimization algorithm iteration. The iteration process revolves around the objectives of maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs.

[0069] A multi-criteria simulated annealing algorithm adapting to interval uncertainty is employed for global optimization. This algorithm uses the initial maintenance plan as the current solution and initiates an iterative optimization process around three core objectives: maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs. In each iteration, the algorithm generates candidate solutions by adding a maintenance action corresponding to a dominant fault, removing a non-core maintenance action, or replacing a feeder's maintenance action, among other neighborhood perturbations. The feasibility of these candidate solutions is verified to ensure that their cumulative cost does not exceed the budget limit and covers the main causes of dominant faults. Subsequently, the Pareto dominance relation is introduced to determine the merits of the candidate solutions compared to the current solution. If a candidate solution is not inferior to the current solution in all objectives and is strictly superior in at least one objective, it is directly accepted as the new current solution. If the candidate solution fails to dominate the current solution, the acceptance probability is calculated and compared with a random number to determine whether to accept the inferior solution. This probabilistic acceptance mechanism helps the algorithm escape local optima traps. As iterations proceed, the temperature gradually decreases according to a preset cooling coefficient, and the probability of accepting an inferior solution also decreases, leading the algorithm to gradually converge to a stable state. After the iteration terminates, all feasible solutions accepted during the iteration process are sorted in a non-dominated order, and the solution set that is not dominated by any other solution is selected, thus forming a Pareto optimal solution set covering different risk preferences and budget scenarios.

[0070] Cost quantification of maintenance actions: Based on a triangular distribution model of single maintenance cost, combined with maintenance length. (feeder) Implementing actions (length, km), calculate the total cost interval:

[0071]

[0072] in, The unit length cost range.

[0073] Quantifying Benefits: Introducing Maintenance Effectiveness Factors (action For feeder Due to the reason (Fault prevention effectiveness range), calculate the benefit range:

[0074] Local level:

[0075]

[0076]

[0077] System level:

[0078]

[0079]

[0080] A multi-criteria simulated annealing (MCSA) algorithm, adapted to the uncertainty of the interval and the need to balance multiple objectives, is employed for global optimization. This algorithm discovers optimal solutions by simulating the physical annealing process, effectively avoiding local optimum traps. Using the previously generated initial maintenance plan as the initial solution of the algorithm, and inputting quantified reliability indicators, maintenance cost and other uncertainty distribution models, cost and benefit interval data are obtained, along with the Hurwitz risk parameters set by the decision-maker. Focusing on "maximizing system-level reliability benefits" , ), maximizing local-level reliability benefits ( , Minimize total maintenance cost The three core objectives were launched for optimization and iteration.

[0081] During iteration, the MCSA algorithm parameters are initialized first:

[0082] Initial solution Initial maintenance plan (including "feeder-maintenance action" combination);

[0083] Objective function set:

[0084]

[0085] in, For the overall benefits of system-level reliability, For the overall benefits of local-level reliability, This is a negative target for cost.

[0086] Control parameters =0.3 (adjusting temperature range), cooling coefficient =0.9, minimum temperature Maximum number of iterations .

[0087] Initial maximum temperature The initial exploration range of the algorithm is determined by calculation using the following formula:

[0088]

[0089] in, (Temperature decay coefficient, related to cooling coefficient); (The initial solution's target weighted sum, , (Target weights set for decision-makers) Cost of the initial solution The corresponding target component (characterizing the initial target contribution of cost).

[0090] Initialize the current solution: Let the current solution be... Current temperature .

[0091] For the current solution Perform neighborhood perturbation (adjustment logic to adapt to distribution network maintenance scheme) to generate candidate solutions. Disturbance methods: ① Add one maintenance action corresponding to a dominant fault (e.g., add "Equipment Status Inspection" to a high-load feeder); ② Remove one non-core maintenance action (e.g., remove "Tree Pruning" from a low-fault-risk feeder); ③ Replace a feeder's maintenance action (e.g., replace "Tree Pruning" with "Line Insulation Covering"). Feasibility verification: Ensure Cumulative costs Budget cap, and coverage 70% of the leading causes of failure.

[0092] MCSA introduces the Pareto dominance relationship, combined with the crisp value of the interval parameter (via the Hurwitz risk parameter). (Transformation), judging candidate solutions With the current solution Advantages and disadvantages:

[0093] Scenario 1: If satisfy: And at least one objective is strictly better (e.g. ),but non-dominance Accept directly ,renew .

[0094] Scenario 2: Calculate the probability of acceptance:

[0095]

[0096] in, (The target weighted sum of the current solution and the candidate solutions).

[0097] Generate random numbers ,like Then accept Otherwise, leave it as is. .

[0098] Update temperature according to rules: Repeat the "neighborhood search → acceptance criteria" steps until... Or the number of iterations reaches The iteration is terminated.

[0099] Pareto optimal solution selection: Collect all accepted feasible solutions during the iteration process, and for each solution... Determine if there are other solutions. satisfy" , "And at least one objective is better": if such an objective does not exist. ,but The solution is the Pareto optimal solution; all Pareto optimal solutions are summarized to form the "Pareto optimal solution set".

[0100] 105. Select the scheme that balances system and local reliability from the Pareto optimal scheme set, and perform sensitivity analysis on the selected scheme based on the probability distribution model to obtain the target maintenance planning scheme.

[0101] A balanced weight, designated by decision-makers, is assigned to system-level and local-level reliability indicators. A comprehensive reliability evaluation index is constructed to quantitatively evaluate all schemes in the Pareto optimal solution set, selecting the maintenance scheme with the optimal index. This ensures that the reliability of power supply without users is significantly lower than the system average, thus aligning with the principle of energy equity. After scheme selection, based on the constructed normal distribution model of system average outage frequency, Weibull distribution model of fault repair time, and triangular distribution model of maintenance cost, interval sensitivity analysis is conducted on the selected schemes. Specifically, by adjusting the confidence interval range of reliability indicators, the parameter values ​​of maintenance cost distribution, and the fluctuation range of maintenance effectiveness factors, the impact of uncertainties such as equipment aging and environmental changes on the scheme's effectiveness is simulated, verifying the stability of the comprehensive reliability evaluation index under parameter fluctuation scenarios. If the fluctuation amplitude of the evaluation index of a scheme within a given fluctuation range is controlled within a preset threshold, its strong robustness is confirmed. The final output is a target maintenance planning scheme that balances system and local reliability and has strong adaptability to uncertainty.

[0102] In this embodiment of the invention, by constructing a probability distribution model that reflects the fluctuation characteristics of reliability indicators and maintenance costs, the uncertainty of parameters caused by factors such as equipment aging and environmental changes is effectively quantified, making the decision-making basis more in line with actual working conditions. When generating maintenance plans, this plan does not only output a single plan based on fixed rules, but iteratively generates a Pareto optimal plan set containing multiple equilibrium plans through a multi-criteria optimization algorithm. It clearly presents the feasible selection range and its applicable boundaries under different budgets and risk preferences, and finally selects the optimal recommended plan that balances system and local reliability. This plan provides a global decision-making perspective while retaining the scientific nature of the final recommendation, significantly improving the flexibility and robustness of maintenance planning.

[0103] Implementation example:

[0104] This embodiment takes a medium-voltage power distribution feeder as the research object and conducts a maintenance planning verification considering uncertainty and multi-criteria risk balance. The feeder covers approximately 1 million users, and the main causes of power outages include equipment failure, tree contact, and meteorological factors such as ice / snow / strong winds. Maintenance actions include equipment condition inspection, tree clearing, and phase-to-phase spacer installation.

[0105] According to the method described in this invention, the uncertainty of reliability indicators is first quantified using the Bootstrap method to obtain the 95% confidence intervals of local / system-level SAIDI and SAIFI. Then, the Hurwitz risk preference criterion is introduced to construct five maintenance schemes (P1–P5): risk-averse, risk-seeking, benefit-oriented, local-oriented, and traditional deterministic. Finally, in a scenario without budget constraints, the optimal maintenance scheme set is obtained, and the maintenance cost range and the number of maintenance actions for each scheme are visualized and compared. The results are as follows: Figure 6 As shown.

[0106] The results show significant differences in cost and number of maintenance actions among the five maintenance schemes: Risk-averse scheme P1: cost range [924,300, 1,170,800 yuan], maintenance actions 82, leaning towards conservatism and stability; Risk-seeking scheme P2: cost range [957,100, 1,211,300 yuan], maintenance actions 106, emphasizing high reliability benefits; Traditional deterministic scheme P5: cost range [944,600, 1,195,000 yuan], maintenance actions 94, not considering uncertainty, and less robust than the risk-balanced scheme; Localized scheme P4: cost range [1,558,400, 1,960,100 yuan], maintenance actions 69, focusing more on the power supply quality of local users; Benefit-oriented scheme P3: cost range [1,834,600, 2,318,200 yuan], maintenance actions 132, showing the largest improvement in system / local reliability. The above results fully verify the effectiveness of the patented method in quantifying maintenance planning uncertainty and balancing multi-criteria risk preferences, and can provide distribution companies with robust and flexible feeder maintenance decision support.

[0107] Please see Figure 2 Another embodiment of the distribution network feeder maintenance planning method in this invention includes:

[0108] 201. Obtain historical operating data of the distribution network, including fault records, outage records, and equipment operation and maintenance data;

[0109] Based on a multi-source data platform including distribution network automation systems, production management systems, and geographic information systems, historical operational data is acquired in batches at regular intervals through a unified data interface. Fault records are extracted from the event sequence records of the dispatch automation system, including fault occurrence time, fault type, faulty equipment, fault cause, protection action information, and reclosing status. Power outage records are extracted from the power outage management module of the marketing business application system, including outage start time, outage end time, outage range, outage type (planned outage / fault outage), outage line, and number of affected users. Equipment operation and maintenance data are collected from the equipment ledger and operation and maintenance records of the production management system, including equipment type, commissioning date, maintenance records, test data, defect records, and inspection status. The acquired multi-source heterogeneous data undergoes data cleaning, removing duplicate records, correcting abnormal timestamps, standardizing data formats, and establishing relationships according to feeder number and time dimension to form a structured historical operational dataset.

[0110] 202. Based on historical operational data, construct a probability distribution model that reflects the fluctuation characteristics of reliability indicators and maintenance costs;

[0111] Based on the number of fault occurrences counted by feeder in the fault record and the number of affected users recorded in the outage log, the time series value of the system's average power outage frequency is calculated. The time series value is fitted with a normal distribution using the maximum likelihood estimation method to obtain the mean and standard deviation, and the confidence interval is calculated to construct a normal distribution model of the system's average power outage frequency. Specifically, based on the number of faults recorded by feeder in the fault log and the number of affected users recorded in the outage log, the time series value of the system's average outage frequency is calculated. For example, using a monthly or quarterly statistical period, the total number of faults in each feeder's statistical period is multiplied by the user weight of that feeder, summed, and then divided by the total number of users in the system to obtain the raw value of the system's average outage frequency for each period. This time series value is then fitted to a normal distribution using the maximum likelihood estimation method. Specifically, the preprocessed system average outage frequency data is substituted into the likelihood function of the normal distribution probability density function. By solving for the parameters corresponding to the maximum value of the likelihood function, the estimated values ​​of the mean and standard deviation are obtained. Based on the mean and standard deviation, the confidence interval at the preset confidence level is calculated. At the same time, the chi-square test is used to verify the goodness of fit between the data and the normal distribution. If the test fails, outlier rollback is performed or the log-normal distribution is used for refitting until a normal distribution model that passes the test is obtained.

[0112] Based on fault repair time data categorized by fault type from fault records, a Weibull distribution is fitted using the least squares method to obtain shape and scale parameters, and confidence intervals are calculated to construct a Weibull distribution model for fault repair time. Specifically, based on fault repair time data categorized by fault type from fault records, a Weibull distribution is fitted using the least squares method to obtain shape and scale parameters, and confidence intervals are calculated. Specifically, the repair time data for each type of fault are arranged in ascending order, the empirical value of the cumulative distribution function is calculated, and a double log-linear transformation is performed on the Weibull distribution function. Linear regression is then performed with the logarithm of the repair time as the independent variable and the transformed cumulative distribution function as the dependent variable. The regression coefficients are solved using the least squares method, and the estimated values ​​of the shape and scale parameters are derived. Confidence intervals at a pre-set confidence level are calculated based on the asymptotic normality of the parameter estimates. The Kolmogorov-Smirnov test is used to verify the goodness of fit between the data and the Weibull distribution. If the test fails, the data is either log-transformed or reclassified according to finer-grained fault causes before refitting.

[0113] Based on the single maintenance cost records categorized by maintenance type in the equipment operation and maintenance data, the minimum, most likely, and maximum costs of each maintenance action are extracted to construct a triangular distribution model for single maintenance costs. Specifically, based on the single maintenance cost records categorized by maintenance type in the equipment operation and maintenance data, the minimum, most likely, and maximum costs of each maintenance action are extracted to construct a triangular distribution model for single maintenance costs. This involves sorting the historical cost data for each type of maintenance action, taking the minimum value as the lower limit parameter of the triangular distribution, and taking the maximum value as the upper limit parameter. The most likely value is determined using the mode or through analysis of the data distribution pattern. If the data shows significant skewness, a modified Pearson empirical formula can be used to estimate the most likely value. The minimum, most likely, and maximum values ​​are substituted into the probability density function of the triangular distribution, and the cumulative frequency curve of the actual cost data is compared with the cumulative distribution function curve of the triangular distribution for verification. If the deviation exceeds a preset threshold, the percentile method is used again to determine the three core parameters: the lower quintile is taken as the minimum value, the upper quintile as the maximum value, and the median or mode as the most likely value, ensuring that the triangular distribution model can accurately reflect the distribution characteristics of historical costs.

[0114] When constructing a normal distribution model for the average power outage frequency, if the chi-square test shows that the data does not fit the normal distribution well, outlier rollback is first performed. Outliers are identified and removed using box plots or the Laida criterion. The remaining data are then re-evaluated for maximum likelihood and subjected to the chi-square test. If the results still fail, a log-normal or Weibull distribution is used for refitting until the optimal distribution model that passes the test is found. When constructing a Weibull distribution model for fault repair time, if the Kolmogorov-Smirnov test fails, a logarithmic transformation is performed on the original data. Alternatively, a secondary classification based on finer-grained fault causes can be performed and the model refitted. If multiple attempts fail to pass the test, a nonparametric probability distribution model can be constructed using kernel density estimation as an alternative. When constructing a triangular distribution model for single maintenance costs, if the cumulative frequency curve of the actual cost data deviates from the cumulative distribution function curve of the triangular distribution by more than a preset threshold, the minimum, most likely, and maximum values ​​are re-selected, and the three core parameters are determined using the percentile method. Alternatively, a beta distribution or log-normal distribution can be used for fitting to ensure that each probability distribution model accurately reflects the statistical regularity of historical data.

[0115] 203. Based on the probability distribution model, extract the reliability index interval data of each feeder-fault cause combination, and combine it with the preset risk preference parameter to transform the interval data into a definite value. Then, sort the power outage impact of the feeder-fault cause combination by the interval TOPSIS method.

[0116] Based on the normal distribution model of system average outage frequency and the Weibull distribution model of fault repair time in the probability distribution model, pre-set confidence intervals for system-level average outage frequency, local-level average outage frequency, system-level average outage duration, and local-level average outage duration are extracted, and statistically classified according to feeder-fault causes. The interval TOPSIS method is used to rank the classified feeder-fault cause combinations. First, the benefit indicators and cost indicators are subjected to interval normalization, and then weighted normalization is performed. Then, the Euclidean distance from each evaluation object to the ideal solution and the anti-ideal solution is calculated. The relative proximity of each feeder-fault cause combination is calculated based on the Euclidean distance, and the combinations are ranked from largest to smallest according to the relative proximity value to obtain the ranking results of the outage impact of feeder-fault cause combinations.

[0117] Based on the normal distribution model of the system's average outage frequency, for each fault cause combination on each feeder, confidence intervals for the system-level and local-level average outage frequency at a preset confidence level are extracted. Simultaneously, based on the Weibull distribution model of fault repair time, confidence intervals for the corresponding combinations of system-level and local-level average outage durations are extracted. All indicator data are then categorized and organized according to feeder number and fault cause. Subsequently, the interval approximation ideal solution ranking method is used to rank these combinations using multi-criteria decision-making. Specifically, all evaluation indicators are first clearly distinguished into benefit indicators and cost indicators, and the interval data for these two types of indicators are normalized to eliminate... In addition to the influence of different dimensions, the normalized interval data is weighted by pre-set index weights to construct a weighted normalized interval decision matrix. Based on this, the positive ideal solution composed of the optimal value and the negative ideal solution composed of the worst value are determined for each index interval. The Euclidean distance between each feeder-fault cause combination and the positive and negative ideal solutions is calculated, and the relative proximity of each combination is calculated based on these two distance values. Finally, all combinations are sorted from largest to smallest relative proximity. The larger the proximity, the more severe the overall power outage impact of the combination, thus obtaining the quantified priority ranking result of the power outage impact of the feeder-fault cause combination.

[0118] 204. Based on the sorting results, apply the Pareto principle to filter out the list of fault causes that dominate the power outage, and match the corresponding maintenance actions for each dominant fault cause according to the preset matching rules.

[0119] All combinations are arranged from highest to lowest relative proximity, and the cumulative contribution of each combination's relative proximity to the total is calculated. The Pareto principle is applied to select the top few combinations with significant cumulative contributions as the list of dominant fault causes for power outages, ensuring that the focus is on the key factors that contribute the most to the power outage impact. The significance level can be a preset percentage of cumulative contribution, such as 80%. Subsequently, based on a preset maintenance action matching rule library, which establishes a mapping relationship based on fault cause type, equipment type, fault phenomenon, and historical maintenance experience, the corresponding standard maintenance action is automatically matched for each selected dominant fault cause, thus forming a "feeder-dominant fault cause-matching maintenance action" correspondence table with feeders as the unit.

[0120] 205. Based on the single maintenance cost distribution model in the probability distribution model, the cost range of each maintenance action is calculated in combination with the maintenance length, and then converted into a definite value according to the risk preference parameter. After constructing the evaluation matrix, the benefit-cost ratio is sorted again by the interval TOPSIS method to obtain the initial maintenance plan.

[0121] Based on the identified primary causes of failure and their corresponding maintenance actions, interval data of unit length maintenance costs are extracted from the triangular distribution model of single maintenance costs. Combined with the maintenance length of each feeder, the total cost interval for each maintenance action is calculated. Based on the reliability index interval data in the probability distribution model and the preset maintenance effectiveness factor interval, the benefit intervals of each maintenance action on system-level reliability and local-level reliability are calculated. Combined with preset risk preference parameters, the cost intervals and benefit intervals are transformed into cost determinants and benefit determinants, respectively. An evaluation matrix is ​​constructed with feeder-maintenance actions as rows and system-level benefits, local-level benefits, and costs as columns. The interval TOPSIS method is used to sort the evaluation matrix by benefit-cost ratio, calculate the relative proximity of each feeder-maintenance action combination, and filter according to the relative proximity value from largest to smallest. During the filtering process, the cumulative value of the total cost interval is used as the basis for judging budget constraints. Under preset budget constraints or quantity constraints, an initial maintenance plan containing feeder-maintenance action combinations and their cost and benefit determinants is obtained.

[0122] For the identified dominant fault causes and their corresponding maintenance actions, firstly, confidence interval data of the unit length maintenance cost are extracted from the triangular distribution model of single maintenance costs, based on the maintenance action type. Then, combined with the actual length of the planned maintenance action on each feeder, the total cost interval for each feeder-maintenance action combination is calculated using interval multiplication. Simultaneously, based on the established normal distribution model of system average outage frequency and Weibull distribution model of fault repair duration, the reliability index interval for each feeder-fault cause combination before maintenance is extracted. This is then combined with a maintenance effectiveness factor interval pre-set through historical fault data statistical analysis or expert experience (this interval quantifies the preventative effect of different maintenance actions on different fault causes; its value is determined based on historical failure rate changes before and after similar maintenance or the Delphi method). Interval calculations are then used to calculate the improvement in system-level average outage frequency and local-level average outage frequency after maintenance, thus obtaining the system-level reliability benefit brought by each maintenance action. The system first identifies the interval and local-level reliability benefit intervals. Then, it introduces preset risk preference parameters and uses the quantile method to transform the cost and benefit intervals of each combination into single cost and benefit determinants, respectively. An evaluation matrix is ​​constructed with feeder-maintenance actions as rows and system-level benefit determinants, local-level benefit determinants, and cost determinants as columns. The evaluation matrix is ​​then sorted by benefit-cost ratio using the interval approximation ideal solution ranking method. The relative proximity of each feeder-maintenance action combination is calculated, and all combinations are arranged in descending order of relative proximity. Combinations are then selected one by one in this order to enter the initial maintenance plan. During the selection process, the upper limit of the total cost interval of the selected combinations is accumulated in real time as a basis for budget occupancy judgment, ensuring that the accumulated cost does not exceed the preset budget upper limit. The number of combinations to be selected can also be set as needed. Finally, under the condition of meeting budget constraints or quantity constraints, an initial maintenance plan is obtained consisting of the selected feeder-maintenance action combinations and their corresponding cost and benefit determinants.

[0123] 206. Based on the initial maintenance plan, a Pareto optimal solution set is obtained through multi-criteria optimization algorithm iteration, wherein the iteration process revolves around the objectives of maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs;

[0124] Using the initial maintenance plan as the current solution, the control parameters of the multi-criteria simulated annealing algorithm are initialized, and three objective functions are set: maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs. The control parameters include initial temperature, cooling coefficient, termination temperature, and maximum number of iterations. Neighborhood perturbation operations are performed on the current solution, generating candidate solutions through adding, removing, or replacing maintenance actions. The feasibility of candidate solutions is verified to ensure that their cumulative total maintenance cost does not exceed a preset budget limit and covers a preset proportion of dominant failure causes. The cost and benefit ranges of each maintenance action are transformed into deterministic values ​​based on preset risk preference parameters. The merits of candidate solutions compared to the current solution are judged based on Pareto dominance. If a candidate solution dominates the current solution, it is accepted as the new current solution; otherwise, according to the probability acceptance mechanism of simulated annealing, a non-dominated candidate solution is accepted as the new current solution with a certain probability, and the temperature is updated according to the cooling coefficient. Iteration continues until the termination condition is met. All accepted feasible solutions are collected during the iteration process, and Pareto optimal solutions are selected through non-dominated sorting, forming a Pareto optimal solution set.

[0125] After obtaining the initial maintenance plan, given that it has been pre-screened for local benefit-cost ratio using the interval TOPSIS method and is a high-quality feasible solution that meets the basic budget constraints, compared to a randomly generated initial solution, it can effectively narrow the search range of subsequent global optimization and improve convergence efficiency. Therefore, it is used as the initial current solution for the multi-criteria simulated annealing algorithm, and the algorithm's control parameters are initialized, including setting the initial temperature, cooling coefficient, termination temperature, and maximum number of iterations. Three optimization objectives are defined: maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs. In each iteration, candidate solutions are generated by performing neighborhood perturbation operations on the current solution. Perturbation methods include randomly adding an unselected feeder-maintenance action combination, randomly removing a selected combination, or randomly replacing a selected combination with an unselected combination. The feasibility of the generated candidate solutions is verified to ensure that the cumulative total maintenance cost of all maintenance actions does not exceed the preset budget limit, and that the proportion of the number of dominant failure causes covered by the candidate solutions reaches a certain threshold. The pre-defined requirements are as follows: After successful verification, the cost and benefit ranges of each maintenance action in the candidate solution are transformed into definite cost and benefit values ​​based on the pre-defined risk preference parameters. These values ​​are then accumulated to obtain the three objective function values ​​of the candidate solution: system-level benefit, local-level benefit, and total cost. Based on the Pareto dominance relationship, the candidate solution is compared with the current solution. If the candidate solution is not inferior to the current solution in all three objectives and is superior in at least one objective, it is determined that the candidate solution dominates the current solution and is directly accepted as the new current solution. Otherwise, the acceptance probability is calculated according to the probability acceptance mechanism of simulated annealing, and a candidate solution that is not dominated or inferior to the current solution is accepted as the new current solution with a certain probability, thus avoiding getting trapped in local optima. After the acceptance judgment is completed, the current temperature is updated according to the cooling coefficient, and the next iteration begins until the termination temperature or the maximum number of iterations is reached. During the iteration process, all feasible solutions that have passed the feasibility verification and been accepted are recorded. After the algorithm terminates, all collected feasible solutions are quickly sorted for non-dominated solutions, and Pareto optimal solutions that are not dominated by any other solution are selected. These solutions are then summarized to form a Pareto optimal solution set.

[0126] 207. Select schemes that balance system and local reliability from the Pareto optimal scheme set, and perform sensitivity analysis on the selected schemes based on the probability distribution model to obtain the target maintenance planning scheme.

[0127] Pre-defined balancing weights are assigned to system-level and local-level reliability indicators. Based on these weights, a comprehensive reliability evaluation index is constructed for each scheme in the Pareto optimal solution set, and the scheme with the optimal comprehensive reliability evaluation index is selected as a candidate scheme. Based on a probability distribution model, parameter fluctuations are adjusted for the reliability indicator range, maintenance cost range, and maintenance effectiveness factor range in the candidate schemes. The magnitude of change in the comprehensive reliability evaluation index of the candidate schemes is verified under different parameter fluctuation scenarios. Based on the verification results under different parameter fluctuation scenarios, the adaptability of the candidate schemes to parameter fluctuations is evaluated, and the scheme with the smallest fluctuation magnitude of the comprehensive reliability evaluation index under a pre-set confidence level is selected as the target maintenance planning scheme output.

[0128] Based on the requirements of distribution network operation and management, preset balance weights are assigned to system-level reliability indicators and local-level reliability indicators. Based on this, a comprehensive reliability evaluation index is constructed for each scheme in the scheme set. This involves merging the two reliability benefit target values ​​of the scheme into a single comprehensive evaluation value through weighted summation, and selecting the scheme with the highest comprehensive reliability evaluation index as the initial candidate scheme. Subsequently, sensitivity analysis is performed on this candidate scheme. Based on the constructed probability distribution model, parameter fluctuation adjustments are made to the reliability indicator range, maintenance cost range, and maintenance effectiveness factor range involved in each maintenance action. Specifically, multiple sets of parameter fluctuation scenarios are generated within the preset confidence intervals of each distribution model using a certain step size or random sampling. The fluctuation scenario represents a possible uncertainty. The comprehensive reliability evaluation index of the candidate scheme is recalculated under different parameter fluctuation scenarios, and the changes in the index under all scenarios are recorded. Then, its fluctuation amplitude, such as standard deviation or range, is calculated. Finally, based on the verification results under the above different parameter fluctuation scenarios, the adaptability of the candidate scheme to parameter fluctuations is comprehensively evaluated. If its fluctuation amplitude is within the preset acceptable range, it is directly output as the target maintenance planning scheme. Otherwise, the scheme with the second-best comprehensive reliability evaluation index is selected from the Pareto optimal scheme set, and the above sensitivity analysis process is repeated until the scheme with the smallest comprehensive reliability evaluation index fluctuation amplitude and the index value meeting the requirements under the preset confidence level is selected as the final target maintenance planning scheme output.

[0129] In this embodiment of the invention, by constructing a probability distribution model reflecting the fluctuation characteristics of reliability indicators and maintenance costs, the uncertainty of parameters caused by factors such as equipment aging and environmental changes is effectively quantified, making the decision-making basis more closely aligned with actual working conditions. By combining the interval TOPSIS method with risk preference parameters twice, the dominant fault causes are accurately screened and an initial maintenance plan is generated, improving the targeting and economy of maintenance actions. On this basis, a Pareto optimal solution set containing multiple equilibrium solutions is generated iteratively through a multi-criteria optimization algorithm, providing decision-makers with a flexible selection range under different budgets and risk preferences. Finally, a solution that balances system and local reliability is selected and sensitivity analysis is performed, ensuring the robustness and power supply fairness of the final solution under uncertain environments, significantly improving the scientificity, flexibility, and adaptability of distribution network maintenance planning.

[0130] The above describes the distribution network feeder maintenance planning method in the embodiments of the present invention. The following describes the distribution network feeder maintenance planning device in the embodiments of the present invention. Please refer to [link / reference]. Figure 3 One embodiment of the distribution network feeder maintenance planning device in this invention includes:

[0131] The acquisition module 301 is used to acquire historical operating data of the power distribution network, including fault records, outage records and equipment operation and maintenance data.

[0132] Module 302 is used to construct a probability distribution model that reflects the fluctuation characteristics of reliability indicators and maintenance costs based on historical operating data.

[0133] The generation module 303 is used to filter out the main causes of power outages and match maintenance actions based on the probability distribution model and preset risk preference parameters, and generate an initial maintenance plan.

[0134] The processing module 304 is used to obtain a Pareto optimal solution set through a multi-criteria optimization algorithm based on the initial maintenance plan. The iterative process revolves around the objectives of maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs.

[0135] The screening module 305 is used to screen out schemes that balance system and local reliability from the Pareto optimal scheme set, and to perform sensitivity analysis on the screened schemes based on the probability distribution model to obtain the target maintenance planning scheme.

[0136] In this embodiment of the invention, by constructing a probability distribution model that reflects the fluctuation characteristics of reliability indicators and maintenance costs, the uncertainty of parameters caused by factors such as equipment aging and environmental changes is effectively quantified, making the decision-making basis more in line with actual working conditions. When generating maintenance plans, this plan does not only output a single plan based on fixed rules, but iteratively generates a Pareto optimal plan set containing multiple equilibrium plans through a multi-criteria optimization algorithm. It clearly presents the feasible selection range and its applicable boundaries under different budgets and risk preferences, and finally selects the optimal recommended plan that balances system and local reliability. This plan provides a global decision-making perspective while retaining the scientific nature of the final recommendation, significantly improving the flexibility and robustness of maintenance planning.

[0137] Please see Figure 4 Another embodiment of the distribution network feeder maintenance planning device in this invention includes:

[0138] The acquisition module 301 is used to acquire historical operating data of the power distribution network, including fault records, outage records and equipment operation and maintenance data.

[0139] Module 302 is used to construct a probability distribution model that reflects the fluctuation characteristics of reliability indicators and maintenance costs based on historical operating data.

[0140] The generation module 303 is used to filter out the main causes of power outages and match maintenance actions based on the probability distribution model and preset risk preference parameters, and generate an initial maintenance plan.

[0141] The processing module 304 is used to obtain a Pareto optimal solution set through a multi-criteria optimization algorithm based on the initial maintenance plan. The iterative process revolves around the objectives of maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs.

[0142] The screening module 305 is used to screen out schemes that balance system and local reliability from the Pareto optimal scheme set, and to perform sensitivity analysis on the screened schemes based on the probability distribution model to obtain the target maintenance planning scheme.

[0143] Optionally, building module 302 can be specifically used for:

[0144] Based on the number of fault occurrences by feeder in the fault records and the number of affected users recorded in the outage log, the time series value of the system's average power outage frequency is calculated. A normal distribution is fitted to this time series value using the maximum likelihood estimation method to obtain the mean and standard deviation, and confidence intervals are calculated to construct a normal distribution model of the system's average power outage frequency. Based on the fault repair time data categorized by fault type in the fault records, a Weibull distribution is fitted using the least squares method to obtain the shape and scale parameters, and confidence intervals are calculated to construct a Weibull distribution model of the fault repair time. Based on the single maintenance cost records categorized by maintenance type in the equipment operation and maintenance data, the minimum, most likely, and maximum costs of various maintenance actions are extracted to construct a triangular distribution model of single maintenance costs.

[0145] Optionally, the generation module 303 includes:

[0146] The sorting unit 3031 is used to extract the reliability index interval data of each feeder-fault cause combination based on the probability distribution model, and convert the interval data into a definite value by combining the preset risk preference parameter, and sort the power outage impact of the feeder-fault cause combination by the interval TOPSIS method.

[0147] The matching unit 3032 is used to filter out the list of dominant power outage fault causes based on the sorting results and the Pareto principle, and to match the corresponding maintenance actions for each dominant fault cause according to the preset matching rules.

[0148] The screening unit 3033 is used to calculate the cost range of each maintenance action based on the single maintenance cost distribution model in the probability distribution model and the maintenance length, and convert it into a definite value according to the risk preference parameter. After constructing the evaluation matrix, the benefit-cost ratio is sorted again by the interval TOPSIS method to screen and obtain the initial maintenance plan.

[0149] Optionally, the sorting unit 3031 can be specifically used for:

[0150] Based on the normal distribution model of system average outage frequency and the Weibull distribution model of fault repair time in the probability distribution model, pre-set confidence intervals for system-level average outage frequency, local-level average outage frequency, system-level average outage duration, and local-level average outage duration are extracted, and statistically classified according to feeder-fault causes. The interval TOPSIS method is used to rank the classified feeder-fault cause combinations. First, the benefit indicators and cost indicators are subjected to interval normalization, and then weighted normalization is performed. Then, the Euclidean distance from each evaluation object to the ideal solution and the anti-ideal solution is calculated. The relative proximity of each feeder-fault cause combination is calculated based on the Euclidean distance, and the combinations are ranked from largest to smallest according to the relative proximity value to obtain the ranking results of the outage impact of feeder-fault cause combinations.

[0151] Optionally, the screening unit 3033 can be specifically used for: extracting interval data of unit length maintenance cost from the triangular distribution model of single maintenance cost based on the screened dominant fault causes and their matching maintenance actions, and calculating the total cost interval of each maintenance action in combination with the maintenance length of each feeder; calculating the benefit interval of each maintenance action on system-level reliability and local-level reliability based on the reliability index interval data in the probability distribution model and the preset maintenance effectiveness factor interval, and converting the cost interval and benefit interval into cost determination value and benefit determination value respectively in combination with the preset risk preference parameter, constructing an evaluation matrix with feeder-maintenance action as the row and system-level benefit, local-level benefit and cost as the column; sorting the evaluation matrix by benefit-cost ratio using the interval TOPSIS method, calculating the relative closeness of each feeder-maintenance action combination, screening in descending order of relative closeness value, and using the cumulative value of the total cost interval as the judgment basis for budget constraints during the screening process, and obtaining an initial maintenance plan containing feeder-maintenance action combinations and their cost and benefit determination values ​​under the preset budget constraints or quantity constraints.

[0152] Optionally, the processing module 304 can be specifically used for:

[0153] Using the initial maintenance plan as the current solution, the control parameters of the multi-criteria simulated annealing algorithm are initialized, and three objective functions are set: maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs. The control parameters include initial temperature, cooling coefficient, termination temperature, and maximum number of iterations. Neighborhood perturbation operations are performed on the current solution, generating candidate solutions through adding, removing, or replacing maintenance actions. The feasibility of candidate solutions is verified to ensure that their cumulative total maintenance cost does not exceed a preset budget limit and covers a preset proportion of dominant failure causes. The cost and benefit ranges of each maintenance action are transformed into deterministic values ​​based on preset risk preference parameters. The merits of candidate solutions compared to the current solution are judged based on Pareto dominance. If a candidate solution dominates the current solution, it is accepted as the new current solution; otherwise, according to the probability acceptance mechanism of simulated annealing, a non-dominated candidate solution is accepted as the new current solution with a certain probability, and the temperature is updated according to the cooling coefficient. Iteration continues until the termination condition is met. All accepted feasible solutions are collected during the iteration process, and Pareto optimal solutions are selected through non-dominated sorting, forming a Pareto optimal solution set.

[0154] Optionally, the filtering module 305 can be specifically used for:

[0155] Pre-defined balancing weights are assigned to system-level and local-level reliability indicators. Based on these weights, a comprehensive reliability evaluation index is constructed for each scheme in the Pareto optimal solution set, and the scheme with the optimal comprehensive reliability evaluation index is selected as a candidate scheme. Based on a probability distribution model, parameter fluctuations are adjusted for the reliability indicator range, maintenance cost range, and maintenance effectiveness factor range in the candidate schemes. The magnitude of change in the comprehensive reliability evaluation index of the candidate schemes is verified under different parameter fluctuation scenarios. Based on the verification results under different parameter fluctuation scenarios, the adaptability of the candidate schemes to parameter fluctuations is evaluated, and the scheme with the smallest fluctuation magnitude of the comprehensive reliability evaluation index under a pre-set confidence level is selected as the target maintenance planning scheme output.

[0156] In this embodiment of the invention, by constructing a probability distribution model reflecting the fluctuation characteristics of reliability indicators and maintenance costs, the uncertainty of parameters caused by factors such as equipment aging and environmental changes is effectively quantified, making the decision-making basis more closely aligned with actual working conditions. By combining the interval TOPSIS method with risk preference parameters twice, the dominant fault causes are accurately screened and an initial maintenance plan is generated, improving the targeting and economy of maintenance actions. On this basis, a Pareto optimal solution set containing multiple equilibrium solutions is generated iteratively through a multi-criteria optimization algorithm, providing decision-makers with a flexible selection range under different budgets and risk preferences. Finally, a solution that balances system and local reliability is selected and sensitivity analysis is performed, ensuring the robustness and power supply fairness of the final solution under uncertain environments, significantly improving the scientificity, flexibility, and adaptability of distribution network maintenance planning.

[0157] above Figure 3 and Figure 4 The distribution network feeder maintenance planning device in this embodiment of the invention is described in detail from the perspective of modular functional entities. The electronic equipment in this embodiment of the invention is described in detail from the perspective of hardware processing.

[0158] See Figure 5 As shown, the electronic device includes a processor 500 and a memory 501. The memory 501 stores machine-executable instructions that can be executed by the processor 500. The processor 500 executes the machine-executable instructions to implement the above-described power distribution network feeder maintenance planning method.

[0159] Furthermore, Figure 5 The electronic device shown also includes a bus 502 and a communication interface 503. The processor 500, the communication interface 503 and the memory 501 are connected via the bus 502.

[0160] The memory 501 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 503 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 502 may be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0161] The processor 500 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 500 or by instructions in software form. The processor 500 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this disclosure. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this disclosure can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 501. The processor 500 reads the information in memory 501 and, in conjunction with its hardware, completes the method steps of the aforementioned embodiment.

[0162] The present invention also provides an electronic device, the computer device including a memory and a processor, the memory storing computer-readable instructions, which, when executed by the processor, cause the processor to perform the steps of the power distribution network feeder maintenance planning method in the above embodiments.

[0163] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of the distribution network feeder maintenance planning method.

[0164] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0165] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0166] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for planning the maintenance of distribution network feeders, characterized in that, include: Acquire historical operating data of the power distribution network, including fault records, outage records, and equipment operation and maintenance data; Based on the historical operating data, a probability distribution model reflecting the fluctuation characteristics of reliability indicators and maintenance costs is constructed. Based on the probability distribution model and preset risk preference parameters, the main causes of power outages are screened out and maintenance actions are matched to generate an initial maintenance plan. Based on the initial maintenance plan, a Pareto optimal solution set is obtained through multi-criteria optimization algorithm iteration. The iteration process revolves around the objectives of maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs. From the Pareto optimal solution set, a solution that balances system and local reliability is selected, and a sensitivity analysis based on the probability distribution model is performed on the selected solution to obtain the target maintenance planning solution.

2. The distribution network feeder maintenance planning method according to claim 1, characterized in that, The construction of a probability distribution model reflecting the fluctuation characteristics of reliability indicators and maintenance costs based on the historical operational data includes: Based on the number of fault occurrences counted by feeder in the fault record and the number of affected users recorded in the power outage account, the time series value of the system's average power outage frequency is calculated; the time series value is fitted with a normal distribution using the maximum likelihood estimation method to obtain the mean and standard deviation, and the confidence interval is calculated to construct a normal distribution model of the system's average power outage frequency. Based on the fault repair time data classified by fault type in the fault records, the Weibull distribution is fitted by the least squares method to obtain the shape parameters and scale parameters, and the confidence interval is calculated to construct a Weibull distribution model of fault repair time. Based on the single maintenance cost records categorized by maintenance type in the equipment operation and maintenance data, the minimum, most likely, and maximum costs of each type of maintenance action are extracted to construct a triangular distribution model of single maintenance costs.

3. The distribution network feeder maintenance planning method according to claim 1, characterized in that, The step of filtering out the dominant power outage causes and matching maintenance actions based on the probability distribution model and preset risk preference parameters, and generating an initial maintenance plan, includes: Based on the probability distribution model, reliability index interval data for each feeder-fault cause combination is extracted, and the interval data is transformed into deterministic values ​​by combining preset risk preference parameters. The power outage impact of the feeder-fault cause combination is ranked by the interval TOPSIS method. Based on the sorting results, the Pareto principle is applied to filter out the list of fault causes that dominate the power outage, and corresponding maintenance actions are matched for each dominant fault cause according to the preset matching rules. Based on the single maintenance cost distribution model in the probability distribution model, the cost range of each maintenance action is calculated in combination with the maintenance length, and then converted into a definite value according to the risk preference parameter. After constructing the evaluation matrix, the benefit-cost ratio is sorted again by the interval TOPSIS method to select the initial maintenance plan.

4. The distribution network feeder maintenance planning method according to claim 3, characterized in that, Based on the probability distribution model, reliability index interval data for each feeder-fault cause combination is extracted, and the interval data is transformed into deterministic values ​​by combining preset risk preference parameters. The power outage impact of the feeder-fault cause combination is then ranked using the interval TOPSIS method, including: Based on the normal distribution model of the system average power outage frequency and the Weibull distribution model of the fault repair time in the probability distribution model, the preset confidence levels of the system-level average power outage frequency, local-level average power outage frequency, system-level average power outage time, and local-level average power outage time are extracted respectively, and classified and statistically analyzed according to feeder-fault cause. The interval TOPSIS method is used to sort the feeder-failure cause combinations after classification. First, the benefit indicators and cost indicators are subjected to interval normalization, then weighted normalization is performed, and then the Euclidean distance from each evaluation object to the ideal solution and the anti-ideal solution is calculated. The relative proximity of each feeder-fault cause combination is calculated based on the Euclidean distance, and the combinations are sorted from largest to smallest relative proximity value to obtain the power outage impact ranking result of the feeder-fault cause combination.

5. The distribution network feeder maintenance planning method according to claim 3, characterized in that, The single maintenance cost distribution model based on the probability distribution model, combined with the maintenance length, calculates the cost range of each maintenance action, transforms it into a definite value according to the risk preference parameter, constructs an evaluation matrix, and then uses the interval TOPSIS method to sort the benefit-cost ratio to obtain the initial maintenance plan, including: Based on the identified primary causes of failure and their corresponding maintenance actions, the interval data of maintenance cost per unit length is extracted from the triangular distribution model of single maintenance cost. Combined with the maintenance length of each feeder, the total cost interval of each maintenance action is calculated. Based on the reliability index interval data in the probability distribution model and the preset maintenance effectiveness factor interval, the benefit interval of each maintenance action on system-level reliability and local-level reliability is calculated. Combined with the preset risk preference parameter, the cost interval and the benefit interval are converted into cost determination value and benefit determination value, respectively. An evaluation matrix is ​​constructed with feeder-maintenance action as the row and system-level benefit, local-level benefit and cost as the column. The evaluation matrix is ​​sorted by benefit-cost ratio using the interval TOPSIS method. The relative proximity of each feeder-maintenance action combination is calculated. The combinations are then filtered in descending order of relative proximity value. During the filtering process, the cumulative value of the total cost interval is used as the basis for judging budget constraints. Under the condition of satisfying the preset budget constraints or quantity constraints, an initial maintenance plan containing feeder-maintenance action combinations and their determined cost and benefit values ​​is obtained.

6. The distribution network feeder maintenance planning method according to claim 1, characterized in that, The process involves iteratively obtaining a Pareto optimal solution set based on the initial maintenance plan using a multi-criteria optimization algorithm. The iterative process revolves around the objectives of maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs. This includes: Using the initial maintenance scheme as the current solution, the control parameters of the multi-criteria simulated annealing algorithm are initialized, and the three objective functions of maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs are set. The control parameters include initial temperature, cooling coefficient, termination temperature, and maximum number of iterations. The current solution is subjected to a neighborhood perturbation operation. Candidate solutions are generated by adding maintenance actions, removing maintenance actions, or replacing maintenance actions. The feasibility of the candidate solutions is verified to ensure that their total cumulative maintenance cost does not exceed the preset budget limit and covers a preset proportion of the dominant fault causes. The cost and benefit ranges of each maintenance action are converted into deterministic values ​​by combining preset risk preference parameters. The merits of the candidate solution and the current solution are judged based on the Pareto dominance relationship. If the candidate solution dominates the current solution, the candidate solution is accepted as the new current solution. Otherwise, the non-dominated candidate solution is accepted as the new current solution with a certain probability according to the probability acceptance mechanism of simulated annealing. The temperature is updated according to the cooling coefficient, and the iteration continues until the termination condition is reached. Collect all accepted feasible solutions during the iteration process, select Pareto optimal solutions through non-dominated sorting, and summarize them to form a Pareto optimal solution set.

7. The distribution network feeder maintenance planning method according to claim 1, characterized in that, The process of selecting a scheme that balances system and local reliability from the Pareto optimal solution set, and performing sensitivity analysis on the selected scheme based on the probability distribution model to obtain the target maintenance planning scheme, includes: A preset balance weight is assigned to the system-level reliability index and the local-level reliability index. Based on the balance weight, a comprehensive reliability evaluation index is constructed for each scheme in the Pareto optimal scheme set, and the scheme with the best comprehensive reliability evaluation index is selected as the candidate scheme. Based on the probability distribution model, the reliability index range, maintenance cost range, and maintenance effectiveness factor range in the candidate scheme are adjusted for parameter fluctuations, and the change range of the comprehensive reliability evaluation index of the candidate scheme is verified under different parameter fluctuation scenarios. Based on the verification results under different parameter fluctuation scenarios, the adaptability of the candidate schemes to parameter fluctuations is evaluated, and the scheme with the smallest fluctuation amplitude of the comprehensive reliability evaluation index under the preset confidence level is selected as the target maintenance planning scheme output.

8. A distribution network feeder maintenance planning device, characterized in that, The power distribution network feeder maintenance planning device includes: The acquisition module is used to acquire historical operating data of the power distribution network, including fault records, outage records, and equipment operation and maintenance data. The construction module is used to construct a probability distribution model that reflects the fluctuation characteristics of reliability indicators and maintenance costs based on the historical operating data. The generation module is used to filter out the main causes of power outages and match maintenance actions based on the probability distribution model and preset risk preference parameters, and generate an initial maintenance plan. The processing module is used to obtain a set of Pareto optimal solutions through a multi-criteria optimization algorithm based on the initial maintenance plan. The iterative process revolves around the objectives of maximizing system-level reliability benefits, maximizing local-level reliability benefits, and minimizing total maintenance costs. The screening module is used to select schemes that balance system and local reliability from the Pareto optimal scheme set, and to perform sensitivity analysis on the selected schemes based on the probability distribution model to obtain the target maintenance planning scheme.

9. An electronic device, characterized in that, include: A memory and at least one processor, wherein the memory stores instructions; The at least one processor invokes the instructions in the memory to cause the electronic device to execute the power distribution network feeder maintenance planning method as described in any one of claims 1-7.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When the instruction is executed by the processor, it implements the power distribution network feeder maintenance planning method as described in any one of claims 1-7.

Images