Power distribution network operation and maintenance optimization method and system based on genetic algorithm

Abstract

The invention discloses a power distribution network operation and maintenance optimization method and system based on a genetic algorithm, and relates to the technical field of power distribution network operation and maintenance, and the method comprises the steps: obtaining real-time power distribution network operation and maintenance data; calculating the maintenance cost and the risk loss according to the data, and establishing an optimization model which takes the minimum weighted value of the maintenance cost and the risk loss after maintenance as a target and is constrained by the operation of resources and a network; the optimization model is converted into two layers of coupling chromosomes, the first layer represents maintenance selection, a time window and a sequence, and the second layer represents resource configuration, parallelism, a transfer scheme and a path; in genetic algorithm iteration, commitment, reconstruction and execution processes are executed in sequence, and feasibility and robustness of solutions are achieved in combination with small-scale accurate sub-problem checking, topology self-adaptive operators and a scene weight gating mechanism; and outputting the differentiated operation and maintenance plan, and triggering rolling re-optimization by executing feedback. According to the invention, the intelligence and the multi-scene adaptability of the operation and maintenance decision of the power distribution network are improved.

Description

Technical Field

[0001] This invention relates to the technical field of distribution network operation and maintenance, and in particular to a distribution network operation and maintenance optimization method and system based on genetic algorithms. Background Technology

[0002] With the continuous expansion of distribution network scale and the increasing complexity of power load characteristics, power grid operation and maintenance tasks are gradually becoming more diversified and dynamic. Traditional operation and maintenance plans mainly rely on manual experience and fixed strategies to arrange maintenance tasks, often making it difficult to balance economy, reliability, and safety. Under current conditions, equipment status is highly uncertain, and environmental factors and load levels have a significant impact on failure probability and risk losses, resulting in significant shortcomings of traditional methods in task allocation, resource coordination, and power outage risk control. Especially under conditions of multiple devices, multiple resources, and multiple constraints, how to balance maintenance costs and risk losses remains a difficult problem in distribution network operation and maintenance optimization.

[0003] On the other hand, most existing intelligent scheduling methods employ single-level optimization modeling, failing to fully reflect the hierarchical relationship between task planning and execution details. Due to a lack of detailed consideration of resource allocation, transfer templates, and path constraints, although these methods can generate overall optimal solutions, resource conflicts, topology infeasibility, or task delays often occur in actual execution, severely impacting the feasibility and practical value of the plan. Furthermore, traditional genetic algorithms are prone to getting trapped in local optima when facing complex and multidimensional power grid operation and maintenance problems, exhibiting slow convergence speeds and insufficient adaptability to various scenarios.

[0004] Therefore, there is an urgent need for an optimization method that can combine equipment status awareness, dynamic adjustment of environmental and load conditions, collaborative modeling of the task and execution layers, and differentiated operation and maintenance plan generation through genetic algorithms. This method should not only reasonably express the balance between maintenance costs and risk losses in the modeling stage, but also consider topological constraints, resource constraints, and dynamic scenario weights in the algorithm implementation, thereby improving the feasibility and robustness of the plan and meeting the actual needs of intelligent distribution network operation and maintenance. Summary of the Invention

[0005] In view of the aforementioned existing problems, the present invention is proposed.

[0006] Therefore, this invention provides a distribution network operation and maintenance optimization method based on genetic algorithms to solve the problem of differentiated operation and maintenance optimization of distribution networks in complex constraints and dynamic environments, achieving a balance between maintenance costs and risk losses.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a distribution network operation and maintenance optimization method based on a genetic algorithm, which includes acquiring real-time distribution network operation and maintenance data, including equipment health and potential risks data, environmental data, load data, and resource data; Based on the aforementioned power distribution network operation and maintenance data, calculate the maintenance costs and risk losses; establish an optimization model with the objective of minimizing the weighted average of maintenance costs and post-maintenance risk losses, and subject to resource and network availability constraints. The objective and constraints of the optimization model are transformed into a two-layer coupled chromosome. The first layer represents: maintenance selection - time window - sequence; the second layer represents: resources - parallelism - transfer - path. The population is initialized with the chromosomes and follows the process of commitment → reconstruction → execution. First, a batch of maintenance is performed and a coarse time window is set up. Then, the feasibility is verified and conflicts are written back with minimal operations. Subsequently, the process is refined by path-time window collaboration and handled by the repairer. During reconstruction and execution, small-scale precise sub-problems are embedded for verification, and smooth convergence is achieved using topological adaptive operators and scene weight gating. Based on the planned output of the iteration results, the execution is fed back to the real-time distribution network operation and maintenance data, triggering rolling re-optimization.

[0008] As a preferred embodiment of the distribution network operation and maintenance optimization method based on genetic algorithm described in this invention, each device is treated as a node, and a model for the maintenance cost and the risk loss is constructed. Based on the health and potential hazards data of the equipment, an n-dimensional equipment feature vector is constructed; using the environmental data and the influence relationship function obtained through pre-training, the parameters of each dimension of the equipment feature vector are adjusted to obtain the adjusted equipment feature vector; The influence relationship function includes the functional relationship between the health and potential risks of the equipment in each dimension and the influence of each type of environmental data; By using a pre-trained Bayesian model and inputting the adjusted device feature vector, the probability distribution of each node's failure is obtained, representing the probability of each type of failure occurring at the node. The maintenance cost is expressed as follows: ; The risk loss is expressed as: ; in, This represents the maintenance cost of the i-th node; This represents the probability that the i-th node will experience a failure of value j. This indicates the risk loss of the i-th node; Indicates the device size at the i-th node Below, the cost of troubleshooting fault j; Indicates the device size at the i-th node The cost of restoring service when fault j occurs; the risk loss after maintenance is: the attenuation rate of the risk loss after maintenance under the current probability distribution learned through prior probability; obtained by multiplying the attenuation rate by the risk loss; The cost of each node during maintenance is a fixed value, and the total resources called by the nodes during maintenance are within the constraints each time maintenance is triggered.

[0009] As a preferred embodiment of the distribution network operation and maintenance optimization method based on genetic algorithm described in this invention, the two-layer coupled chromosome is implemented through collaborative encoding of the task layer and the execution layer. The first layer represents the maintenance plan, including the selection of maintenance equipment nodes, the determination of candidate time windows, and the execution order within the same time window. Each task gene corresponds to a device point and records its commitment status and priority label. The second layer represents the execution details, including the specific resource configuration allocated to the task, the arrangement of parallel time slots, the selected transfer template, and the team's travel path. Each execution gene corresponds one-to-one with the task gene in the first layer.

[0010] As a preferred embodiment of the distribution network operation and maintenance optimization method based on genetic algorithm described in this invention, wherein: during chromosome initialization, an initial batch and time window are generated in the first layer according to the equipment risk-reward ratio, and in the second layer, a feasible work group, transfer template and shortest travel path are assigned to each task; In crossover and mutation operations, the first layer exchanges data on a per-task-gene basis, while the second layer simultaneously adjusts the corresponding resources and paths.

[0011] As a preferred embodiment of the distribution network operation and maintenance optimization method based on genetic algorithm described in this invention, the commitment-reconstruction-execution process includes, in the commitment stage, generating a preliminary maintenance batch based on the node selection and time window results of the first layer chromosome, and concentrating the equipment of the same feeder segment into a coarse-grained candidate time window; During the reconstruction phase, the transfer template of the second-layer chromosome is called to perform network operability verification for each batch, and the power grid topology reconstruction is achieved by prioritizing the switching sequence with the fewest operations. When mutual exclusion switches or power flow over-limit conflicts are detected, the conflict results and correction suggestions are written back to the chromosome library for reference in subsequent generations of crossover and mutation. During the execution phase, based on the feasible batches confirmed by reconstruction, the paths and parallel time slots of the second-layer chromosome are refined. The path-time window coordinator is used to calculate the travel time and arrival order of the shifts. Problems caused by resource overruns or time window conflicts are handled by the repairer, which completes the correction through insertion, postponement, and task exchange to ensure that the final decoded operation and maintenance plan is executable at both the resource and network levels.

[0012] As a preferred embodiment of the distribution network operation and maintenance optimization method based on genetic algorithm described in this invention, the small-scale precise sub-problem verification includes randomly sampling chromosome segments within a local range and using mixed integer programming to calculate the local solution to verify the optimality of the transfer path, resource allocation and power outage range. The verification results are compared with the solutions generated by the genetic algorithm. If significant room for improvement is found in a local area, the corresponding chromosome segment is replaced. At the same time, the feasible transfer and resource scheduling modes obtained from the local optimization are stored in the template library for rapid use in subsequent populations, thereby improving the reliability and convergence speed of the overall solution.

[0013] As a preferred embodiment of the distribution network operation and maintenance optimization method based on genetic algorithm described in this invention, the topology adaptive operator identifies the interlocking relationship between key feeders and mutual exclusion switches in the distribution network structure, and prioritizes keeping these continuous segments intact during crossover and mutation operations to avoid the solution failing during the evolution process. The scenario weight gating dynamically adjusts the cost and risk weights in the objective function based on different environmental and load data, increasing the weight of loss under high-risk or high-load conditions.

[0014] Secondly, this invention provides a distribution network operation and maintenance optimization system based on a genetic algorithm, comprising: a data acquisition unit for acquiring real-time distribution network operation and maintenance data; an analysis unit for calculating maintenance costs and risk losses based on the distribution network operation and maintenance data; establishing an optimization model with the objective of minimizing the weighted average of maintenance costs and post-maintenance risk losses, and subject to resource and network operability constraints; a construction unit for transforming the objective and constraints of the optimization model into a two-layer coupled chromosome, the first layer representing: maintenance selection-time window-sequence; the second layer representing: resources-parallelism-transfer-path; a calculation unit for initializing the population with the chromosome, and following the process of commitment → reconstruction → execution, first generating maintenance batches and coarse time windows, then reconstructing and verifying operability with minimal operations and writing back conflicts, followed by path-time window collaborative refinement and handling through a repairer; an adjustment unit for embedding small-scale precise sub-problems for verification during reconstruction and execution, and smoothing convergence with topology adaptive operators and scene weight gating; and an output unit for outputting the planned results according to the iteration results and feeding back the execution to the real-time distribution network operation and maintenance data to trigger rolling re-optimization.

[0015] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the distribution network operation and maintenance optimization method based on genetic algorithm as described in the first aspect of the present invention.

[0016] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the distribution network operation and maintenance optimization method based on genetic algorithm as described in the first aspect of the present invention.

[0017] The beneficial effects of this invention are as follows: By constructing a two-layer coupled chromosome structure, collaborative optimization between the maintenance task layer and the execution detail layer is achieved, ensuring the organic unity of task selection, time window allocation and resource configuration, and transfer paths. Through the commitment-restructuring-execution process design, the problem of traditional algorithms generating a large number of infeasible solutions is avoided, improving the executability and stability of the optimization results. The embedded small-scale precise sub-problem verification mechanism enables the genetic algorithm to obtain better solutions within a local range, improving the overall solution quality. The introduction of topology adaptive operators maintains the continuity of critical feeders and mutual exclusion switches, avoiding solution failure during the evolution process. The scenario weight gating mechanism dynamically adjusts the weights of cost and risk using environmental and load data, achieving robust optimization under multiple scenario conditions. Combined with the rolling re-optimization mechanism, the model parameters can be continuously updated using execution feedback, enabling the maintenance plan to have continuous optimization capabilities. In summary, this invention can significantly improve the intelligence, dynamism, and differentiation level of distribution network operation and maintenance decision-making. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a distribution network operation and maintenance optimization method based on genetic algorithms. Detailed Implementation

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0021] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0022] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0023] Reference Figure 1 This is one embodiment of the present invention, which provides a distribution network operation and maintenance optimization method based on a genetic algorithm, including the following steps: S1: Obtain real-time power distribution network operation and maintenance data, including equipment health and potential risks, environmental data, load data, and resource data.

[0024] Equipment health and hazard data are compiled from online monitoring, inspection records, and historical work orders, covering multiple dimensions such as health index, hazard level, hazard type, failure frequency, and topological criticality. This data is used to determine whether equipment requires maintenance and provides input for failure probability modeling. Environmental data is acquired from meteorological monitoring and geographic information platforms, including natural factors such as temperature, humidity, wind speed, thunderstorms, and freezing, as well as external conditions such as construction and traffic. This data is used to dynamically adjust equipment failure probabilities and define repairable and non-repairable time windows. Load data is provided by the scheduling system and user-side acquisition system, reflecting real-time load curves, peak-valley forecasts, and key user electricity consumption levels. This data is used to assess potential power outage losses during maintenance and guide task prioritization during low-load periods. Resource data is provided by the operation and maintenance management platform and material inventory system, involving the number and skill levels of work teams, tool and vehicle configuration, available working hours, and spare parts inventory. This data is used to constrain the feasible solution range of the genetic algorithm, avoid resource overruns, and affect the calculation of maintenance time and process losses. By integrating the above four types of data, we can provide comprehensive input for modeling subsequent maintenance costs and risk losses, ensuring that the optimization model is both mathematically sound and accurately reflects the actual operating conditions of the power distribution network.

[0025] S2: Based on the power distribution network operation and maintenance data, calculate the maintenance cost and risk loss; establish an optimization model with the objective of minimizing the weighted average of maintenance cost and post-maintenance risk loss, and subject to resource and network availability constraints.

[0026] Each device is treated as a node, and a model is built to represent the maintenance costs and risk losses.

[0027] Based on the health and potential hazards data of the equipment, an n-dimensional equipment feature vector is constructed; using the environmental data and the influence relationship function obtained through pre-training, the parameters of each dimension of the equipment feature vector are adjusted to obtain the adjusted equipment feature vector.

[0028] The influence relationship function includes the functional relationship between the health and hazard data of the device in each dimension and the influence of each type of environmental data.

[0029] By using a pre-trained Bayesian model and inputting the adjusted device feature vector, the probability distribution of each node's failure is obtained, representing the probability of each type of failure occurring at the node.

[0030] The maintenance cost is expressed as follows: .

[0031] The risk loss is expressed as: .

[0032] in, This represents the maintenance cost of the i-th node; This represents the probability that the i-th node will experience a failure of value j. This indicates the risk loss of the i-th node; It is a prefitted function, representing the device size at the i-th node. Below, the cost of troubleshooting fault j; It is a prefitted function, representing the device size at the i-th node. The cost of restoring use when fault j occurs; the risk loss after maintenance is: the attenuation rate of the risk loss after maintenance under the current probability distribution learned through prior probability; obtained by multiplying the attenuation rate by the risk loss.

[0033] The cost of each node during maintenance is a fixed value (a fixed value for each node). The total resources called by the nodes during maintenance each time maintenance is triggered are within the constraints.

[0034] Specifically, firstly, each device is abstracted as a node, and four types of inputs—"status, environment, load, and resources"—are constructed around this node. On the device side, health and potential hazard data are central, forming multi-dimensional features: including health index, hazard level and type, fault / alarm recurrence rate, topological criticality, and maintenance history. To avoid the randomness of a single moment, the features retain both current cross-sectional values ​​and near-to-long-term trends (such as statistics and change rates over the past 24 hours, 7 days, and 90 days), along with data quality labels. Subsequently, a pre-trained "influence relationship function" is introduced, using real-time environmental data (temperature, humidity, wind, rain, ice, lightning, pollution / salt spray, construction disturbances, etc.) to directionally adjust the dimensions of the above features: for example, when high temperature and high load occur simultaneously, the effective weight of the insulation aging-related dimension is increased; strong winds / icing amplify the mechanical stress and conductor galloping-related dimensions. This function is fitted on historical samples, possessing interpretable mappings. During deployment, segmented or scenario-specific parameters are used to avoid extrapolation distortion in extreme environments.

[0035] The environmentally corrected feature vectors serve as input to the Bayesian model, outputting the probability distribution of multiple fault types for each node within a given evaluation window, and providing uncertainty metrics (confidence intervals or variance). To reduce bias caused by small samples or class imbalance, model training employs hierarchical priors categorized by device type / region, and continuously calibrates through feedback. When a sensor malfunctions or features are missing, the system reverts to the group priors of similar devices, ensuring uninterrupted probability output. At this point, the "node-environment-probability" link is closed, leading to two evaluation channels: cost and risk.

[0036] In this embodiment, maintenance costs are not a fixed constant, but rather constructed around the expected amount of "direct costs + maintenance process losses". Direct costs come from quotas and work order ledgers, including labor, materials, permits, arrival / preparation, and necessary switching operations; maintenance process losses characterize the power supply impact caused by planned maintenance, relying on power transfer schemes and time windows, combining "unit time loss" (calculated from the amount of power not supplied and the value of the interruption) with the expected maintenance duration. Maintenance duration is not a rough estimate, but consists of three segments: "baseline man-hours for equipment and faults - resource capacity mapping - non-parallel time on site". Baseline man-hours reflect the workload, resource capacity is mapped from the number of people / levels in the work group, the capacity of special tools and vehicles, and considers the marginal decrease of multiple people working at the same work station; non-parallel time on site comes from switchover and safety isolation processes. To maintain consistency across different equipment sizes and fault types, both direct costs and recovery workload use a pre-fitted function library (layered fitting according to equipment voltage level, capacity, structural type, and typical faults), calibrated by regression testing of historical work orders before deployment, and rolled over by execution feedback after deployment. In addition, each node triggering maintenance will generate a fixed call cost (including scheduling, permission, fence, etc.), which is recorded separately to avoid confusion with direct payment costs.

[0037] Risk loss measures the expected consequences within an assessment window under "no maintenance" conditions. Its generation process runs parallel to maintenance costs: First, based on scenario-specific environments and loads (normal conditions, high temperatures, typhoons / icing, external construction, etc.), the consequences of a failure are assessed, including user interruption losses (calculated based on time-of-use interruption value and critical user weights), system operation losses (restoration costs for power flow / voltage exceedances, marginal impact of reliability indicators), and equipment asset losses (replacement / overhaul increments due to minor faults evolving into major faults). Then, the failure probability distribution obtained in the previous step is used for weighting to obtain the expected loss within the scenario. This is then robustly aggregated across multiple scenarios using quantile or tail-weighted methods to suppress estimation bias under extreme environments. To avoid double-counting with "maintenance process losses," risk loss is strictly limited to the consequences of "no maintenance – accident occurrence – recovery," excluding power outage costs during planned maintenance.

[0038] Post-maintenance risk loss is characterized by an "attenuation rate," which is the expected reduction in risk loss at a given node after implementing a certain type of maintenance, under current prior / sample conditions. The attenuation rate originates from prior learning and execution feedback: combining equipment health improvements, re-inspection results, and lifespan models (such as state transitions or survival analysis), it lowers the probability of failure and the severity of consequences over a subsequent period, and precipitates these parameters as reusable parameters based on equipment type / failure type. To prevent over-optimism, the attenuation rate includes a confidence interval and converges dynamically under scenario weight gating: at high uncertainties, the attenuation rate is reduced; after the review task is completed, the attenuation bonus is released. In this way, the closed loop of "risk—maintenance—re-risk" is digitized and linked together.

[0039] Resource and network availability constraints determine feasibility. On the resource side, time-by-time verification is performed: within the same time window, the total occupancy of committed tasks for work teams, qualifications, vehicles, specialized tools, and spare parts inventory must not exceed the upper limit; task binding to resources must meet skill matching and work permit requirements; cross-site operations must include travel and preparation time to ensure arrival time and service duration fall within the candidate time window; for splittable operations, the system allows segmented execution without violating safety boundaries, but does not allow crossing prohibited maintenance windows. On the network side, a closed-loop "reconstruction-verification-writeback" process is implemented: within the designated window, the transfer template with the fewest operations is invoked, maintaining the radial or controlled ring network state of the distribution network; the thermal stability limits of lines and transformers, node voltage boundaries, and phase imbalance are verified one by one; the operation interlocking of mutual exclusion switches, protection configuration, and cold load recovery constraints are verified; and, if necessary, the impact of short-circuit current and reverse power flow from connected distributed power sources is checked. If any check fails, conflict evidence (conflict edge, bottleneck window, minimum operation suggestion) is immediately generated and written back to the feasible domain and genetic operator mask to prevent similar combinations from recurring in subsequent generations.

[0040] S3: The objective and constraints of the optimization model are transformed into a two-layer coupled chromosome. The first layer represents: maintenance selection - time window - sequence; the second layer represents: resources - parallelism - transfer - path.

[0041] The two-layer coupled chromosome is implemented through the collaborative coding of the task layer and the execution layer: the first layer is used to represent the maintenance plan, including the selection of maintenance equipment nodes, the determination of candidate time windows, and the execution order within the same time window. Each task gene corresponds to a device point and records its commitment status and priority label. The second layer is used to represent the execution details, including the specific resource configuration allocated to the task, the arrangement of parallel time slots, the selected transfer template, and the movement path of the shift. Each execution gene corresponds one-to-one with the task gene in the first layer.

[0042] In step S3, the objective and constraints of the optimization model are transformed into a two-layer coupled chromosome structure to ensure that the maintenance plan and execution details are expressed simultaneously during the evolution of the genetic algorithm. This chromosome consists of a task layer and an execution layer; the task layer corresponds to the maintenance plan, and the execution layer corresponds to the implementation method. The encoding of the task layer includes three elements: maintenance selection, time window, and sequence. Maintenance selection indicates whether a certain equipment node is included in the current plan, marked using a commitment status field; the time window indicates the time range within which the equipment can be scheduled for maintenance, encoded as a time window number or interval identifier; the sequence determines the execution order of multiple tasks within the same time window, encoded as a sequence label or serial number value. Through the combination of these three elements, the task layer can clearly express "which equipment to repair, when to repair, and the execution order within that time window."

[0043] The execution layer further refines the implementation methods for each committed task, with its encoding corresponding one-to-one with the task layer, ensuring that each task gene has a matching execution gene. The execution layer includes four parts: resources, parallelism, power transfer, and path. The resource field is used to allocate specific maintenance teams, personnel, and tools to ensure that the task has the conditions for implementation; the parallelism field is used to mark whether the task is arranged in parallel with other tasks within its current time window, and the corresponding parallel time slot number; the power transfer field is used to indicate the power supply alternative when a node is out of service, that is, to switch the current to other lines to maintain continuous power supply to the user side by operating the switch, and the value of this field corresponds to the predefined power transfer template; the path field describes the spatial travel route of the maintenance team, including the optimal travel sequence from the base to each maintenance point, which affects the arrival time and operation time. Unlike the sequence label in the task layer, the former is the physical movement path, while the latter is the logical task sequence. The two interact but do not overlap.

[0044] The task layer and execution layer of the two-layer chromosome are bound together by device node IDs, ensuring a one-to-one correspondence between "maintenance selection-time window-sequence" and "resource-parallelism-transfer-path". During decoding, if the execution layer finds that resources are exceeded or transfer is not feasible, it will write the conflict information back to the task layer, thereby guiding the genetic algorithm to adjust task allocation and execution strategies in subsequent iterations. This cooperative encoding method enables the genetic algorithm to ensure the rationality of the global plan at the task level and the feasibility under resource and network constraints at the execution level, thus gradually converging to an operation and maintenance solution that meets both the target optimization and is feasible for implementation.

[0045] In the two-layer coupled chromosome design, although the task layer and the execution layer have different focuses in terms of encoding content, they must be jointly constrained by the constraints during the evolution process to ensure that the decoded repair scheme can not only achieve the optimality of the objective function, but also be implemented in the actual environment.

[0046] The constraints at the task layer mainly revolve around three aspects: maintenance selection, time windows, and sequence. First, regarding maintenance selection, the priority of the selected nodes must be consistent with the risk assessment results; that is, high-risk equipment should be prioritized for the candidate set, while low-risk equipment can be postponed to avoid unnecessary resource consumption. Second, regarding time window constraints, each device can only be assigned to an allowed maintenance window, and crossing prohibited maintenance periods, such as peak load periods, periods of extreme weather, or windows restricted by regulatory authorities, is prohibited. Third, regarding sequence constraints, tasks within the same time window must conform to the topology and operational logic, such as the order of switching operations and the standardization of isolation and recovery sequences, to avoid logical conflicts.

[0047] The constraints at the execution layer are more focused on resource and network operability. Resource constraints require that the total resource consumption of tasks within the same time window must not exceed the number of work teams, tool capacity, and working hours limits. Each task must be executed by a work team with the corresponding qualifications, and the parallel relationship between tasks must conform to the scheduling logic of manpower and tools. Power transfer constraints require that when a node is out of service, there must be a legal power transfer path to ensure power supply to the remaining users, and the power transfer operation must not trigger mutual exclusion switch conflicts or power flow exceeding limits. Path constraints require that the movement routes of work teams are spatially feasible, and that the order of task allocation and physical paths do not contradict each other. For example, the same work team cannot cross maintenance points that are far apart in a short period of time.

[0048] Through the combined effect of the above constraints, the task layer ensures the rationality and standardization of the planning, while the execution layer ensures resource feasibility and power grid security during implementation. The genes in the two layers maintain a corresponding relationship during encoding and decoding. Once a conflict or infeasibility occurs in the execution layer, the information will be written back to the task layer, driving the genetic algorithm to readjust, ultimately ensuring that the solution achieves a balance between the optimal objective and the feasible constraints.

[0049] S4: Initialize the population with the chromosomes, and follow the process of commitment → reconstruction → execution. First, produce a batch of maintenance and a coarse time window, then reconstruct and verify the feasibility with minimal operations and write back the conflicts. Then, refine the process through path-time window collaboration and handle it through the repairer.

[0050] The commitment-reconstruction-execution process includes, in the commitment phase, generating a preliminary maintenance batch based on the node selection and time window results of the first-layer chromosome, and concentrating the equipment of the same feeder segment into a coarse-grained candidate time window. During the reconstruction phase, the transfer template of the second-layer chromosome is called to perform network operability verification for each batch, and the power grid topology reconstruction is achieved by prioritizing the switching sequence with the fewest operations. When mutual exclusion switches or power flow over-limit conflicts are detected, the conflict results and correction suggestions are written back to the chromosome library for reference in subsequent generations of crossover and mutation.

[0051] During the execution phase, based on the feasible batches confirmed by reconstruction, the paths and parallel time slots of the second-layer chromosome are refined. The path-time window coordinator is used to calculate the travel time and arrival order of the shifts. Problems caused by resource overruns or time window conflicts are handled by the repairer, which completes the correction through insertion, postponement, and task exchange to ensure that the final decoded operation and maintenance plan is executable at both the resource and network levels.

[0052] It's important to note that by the end of S3, each device's genes are already defined: whether it's committed (commitment field = 1 / 0), fine-grained time window ID, order within the same window, and corresponding execution layer pointers (resources, parallelism, transfer, path placeholders). The first step of S4 is simply reading the set of genes with "commitment = 1." The commitment flags and fine-grained time windows of these genes are not changed; they are the "defined objects" for all subsequent operations. A fine-grained time window is a specific time slice at the device level (e.g., 09:00–10:30, 10:30–12:00). Coarse-grained candidate time windows are the higher-level merging labels of these fine-grained time windows (e.g., "morning," "afternoon," "night," "low load period of the day"). The merging rules are fixed in the system configuration: each fine-grained time window has a unique higher-level coarse-grained window label. In S4, only each task with "commitment = 1" is labeled with its coarse-grained window label, without changing its original fine-grained time window. This is done to group tasks within the same feeder segment and falling under the same coarse window label into batch containers, so as to uniformly perform supply transfer refactoring and subsequent path-time window collaborative refinement.

[0053] The system first reads the commitment field (1 / 0 encoding) of the first-layer chromosome, only identifying device nodes with a commitment value of "1" as maintenance targets. These nodes already carry fine-grained time window information and sequence labels, indicating when and in what order the equipment should be maintained. To reduce the complexity of subsequent topology calculations, the system assigns a corresponding coarse-grained time window label (e.g., "morning low-load segment," "night window") to each committed task, and groups the tasks using "feeder segment ID + coarse window label" as the key, forming preliminary maintenance batches. The coarse-grained time window is merely a batch container and does not change the original fine-grained time window of the task. Its purpose is to group tasks on the same feeder segment together, facilitating unified topology switching and conflict checking. This approach reduces repetitive switching operations while maintaining the flexibility of the fine-grained time window.

[0054] During the reconfiguration phase, the system uses the coarse-grained batches generated in the commitment phase as objects, and performs network operability checks on the execution details in the second-layer chromosome for each batch. Each task gene is bound to a transfer template at the execution layer. This template records the switching sequence with the fewest switching operations during maintenance, used to switch the load to the backup line or adjacent feeder when the target equipment is out of service, ensuring uninterrupted power supply to users. The system loads these transfer templates task by task and simulates the switching one by one in the distribution network topology model. Subsequently, it performs fast power flow calculations and safety checks, checking whether the current carrying capacity of lines and transformers exceeds the limit, whether the node voltage exceeds the limit, whether mutual exclusion switches are operated simultaneously, whether the protection and tripping logic is broken, and whether the reverse power flow is controllable when there are distributed power sources.

[0055] When a batch passes the overall verification, it is marked as a "feasible batch" and enters the execution phase. When the verification fails, the system generates conflict evidence, including the device node that triggered the conflict, the affected feeder segment, the time window in which the conflict occurred, and recommended minimum correction suggestions (such as replacing the suboptimal template, reducing parallelism, splitting the batch, or fine-shifting the time window). This conflict evidence is written back to the chromosome database and attached to the corresponding gene combination in the form of a blacklist or penalty tag, so that subsequent generations of genetic inheritance automatically reduce the generation of the same infeasible solutions during crossover and mutation operations, while prioritizing the inheritance of verified feasible fragments.

[0056] In genetic evolution, crossover, mutation, and reconstruction are closely coupled. Crossover operations use task blocks with "same feeder + same coarse window" as units. In the first layer, the selection of tasks and the order of time windows are exchanged, while in the second layer, corresponding resource configurations, transfer templates, and path segments are exchanged synchronously, ensuring consistency between the upper and lower layers of the chromosome. Mutations in the first layer typically manifest as flipping commitment bits (0↔1) or minor shifts in the neighborhood of time windows; once such mutations occur, they inevitably trigger synchronous changes in the second-layer execution layer, such as reassigning resources, selecting available transfer templates, or recalculating paths. Mutations in the second layer, such as template replacement, parallel gap adjustment, or local path optimization, can be performed independently without changing the commitment selection in the first layer, used to improve the feasibility and efficiency of the execution layer. Regardless of whether it's crossover or mutation, the results need to undergo an immediate and rapid topology pre-check; if clearly infeasible is detected, a conflict evidence write-back mechanism blacklists the combination to reduce the unnecessary overhead of the search.

[0057] Through this reconfiguration phase design, the system continuously performs a "generation-verification-feedback-correction" cycle during the genetic algorithm search process, enabling individuals to not only gradually optimize the objective function, but also maintain feasibility under the actual power grid topology and resource constraints, ensuring that the scheduling refinement in the subsequent execution phase can proceed smoothly.

[0058] After confirming feasible batches, the system refines the path and parallel time slot fields of the second-level chromosome and calls the path-time window coordinator to generate the actual scheduling plan. The coordinator combines the task's fine time window label and geographical coordinates to calculate the team's travel route and travel time, aligning it with the task's original time window. If the team's arrival time is earlier than the task's time window, a wait is inserted; if it is later, it attempts to make the least-cost sequence fine-tuning or parallel slot adjustment within the same coarse window. For high-density tasks within the same time window, the system controls the maximum number of simultaneous jobs through the parallel field to avoid resource overruns. If conflicts still exist or resources are insufficient, the repairer is called to intervene. The repairer adjusts through three types of operations: inserting new time slots to accommodate conflicting tasks; deferring low-priority tasks to adjacent fine or coarse windows; or exchanging task order within the same team or feeder to shorten paths and reduce peak loads. If the conflicts are complex, the repairer calls a small-scale precise solver to perform a one-time optimization on a subset of local tasks. Once the repair is complete, the results will be written back to both chromosome layers to ensure consistency between the task layer and the execution layer. The commitment field will only be temporarily revoked if all repairs fail.

[0059] Through screening and batching in the commitment phase, topology verification and conflict write-back in the reconstruction phase, and path-time window coordination and repair in the execution phase, this step achieves the complete implementation from genetic individual to actual plan. The optimization objective remains consistent, namely, minimizing the weighted sum of "maintenance costs and post-maintenance risk losses" under resource and network constraints. This design ensures that the generated operation and maintenance plan not only converges at the algorithm level, but also possesses executability and stability in actual distribution network operation.

[0060] S5: During reconstruction and execution, small-scale precise sub-problem verification is embedded, and smooth convergence is achieved by topological adaptive operators and scene weight gating.

[0061] The small-scale precise subproblem verification includes randomly selecting chromosome segments within a local area and using mixed integer programming to calculate the local solutions to verify the optimality of the transfer path, resource allocation, and power outage range. The verification results are compared with the solutions generated by the genetic algorithm. If significant improvement is found locally, the corresponding chromosome segments are replaced. Simultaneously, feasible transfer patterns and resource scheduling patterns obtained from local optimization are stored in a template library for rapid retrieval in subsequent populations, thereby improving the reliability and convergence speed of the overall solution.

[0062] It should be noted that, in order to avoid the genetic algorithm getting stuck in local convergence and improve the engineering feasibility of the results during the reconstruction and execution process, the system embeds small-scale precise sub-problem verification, and combines the mechanism of topological adaptive operator and scene weight gating to achieve a balanced convergence between feasibility and optimization goal in the evolution process.

[0063] The solutions generated by genetic algorithms in the global search are often near-optimal. To improve local quality, the system randomly selects a subset of chromosome segments as verification objects in each generation of evolution. These segments typically correspond to a set of tasks or a continuous set of devices on a single feeder within the same coarse time window. For these local subsets, the system calls the Mixed Integer Programming (MIP) method to perform precise optimization calculations on the transfer path, resource allocation, and power outage range. The results obtained from the precise solution are compared with the solutions from the genetic algorithm. If significant room for improvement is found locally (e.g., reduced operation steps, more balanced parallel scheduling, or smaller power outage range), the corresponding chromosome segment is replaced with the precise solution. Simultaneously, these feasible transfer patterns and resource scheduling patterns obtained from local optimization are deposited into a template library, forming reusable pattern units that can be directly called in subsequent generations, reducing redundant calculations. Through this approach of "genetic search + precise verification + template deposit," the feasibility and convergence speed of the overall solution are significantly improved.

[0064] The topology adaptive operator identifies the interlocking relationships between critical feeders and mutual exclusion switches in the distribution network structure, prioritizing the preservation of these continuous segments during crossover and mutation operations to prevent solution failure during evolution. Distribution network structures exhibit significant interlocking relationships, such as power flow coupling between critical feeder segments and the protection logic of mutual exclusion switches. Arbitrarily breaking these segments during crossover or mutation can easily lead to solution infeasibility. Therefore, the system introduces a topology adaptive operator in the genetic operations: first, based on the distribution network structure analysis, it identifies interlocking combinations of critical nodes, critical feeders, and mutual exclusion switches, marking them as "topology protected segments"; during crossover operations, the entire segment is inherited as a whole without interruption; during mutation, only minor perturbations are allowed at segment boundaries, such as template replacement or task order shifting, without disrupting the logical continuity within the segment. This ensures genetic diversity while avoiding frequent generation of invalid solutions due to disrupted topology logic, thereby improving evolutionary efficiency.

[0065] The core idea of ​​scene weight gating is to no longer use a fixed cost-risk weight ratio in the fitness evaluation stage of the genetic algorithm, but to dynamically adjust the ratio between the two based on real-time environmental data and load data.

[0066] First, the system collects environmental data including meteorological conditions (temperature, humidity, strong winds, lightning, heavy rain, etc.), the climatic characteristics of geographical areas, and special events such as typhoons and ice storms. If extreme weather is detected or equipment is in a harsh environment, the scenario is marked as a "high-risk environment." Simultaneously, the system monitors and predicts the load data of the distribution network, including daily load curves, peak-to-valley differences, and power shortage levels during critical periods. When the load level approaches or exceeds a safe threshold, the scenario is marked as "high load pressure."

[0067] In the objective function calculation, the scenario weight gating module automatically adjusts the weight allocation based on the aforementioned markers. In high-risk or high-load scenarios, the weight of post-maintenance risk losses automatically increases, and the algorithm is more inclined to select solutions that significantly reduce potential risks, prioritizing the safety and reliability of the power grid operation even at the cost of increased maintenance. Conversely, in low-load or favorable environmental conditions, the system increases the weight of maintenance costs, making the optimization results more economical and avoiding excessive resource investment.

[0068] The specific implementation is as follows: Before each generation of the genetic algorithm evolves, the weight gating module calculates a scenario factor based on the latest environment-load monitoring results and applies it to the weighting coefficients of the objective function. This factor can be set as a tiered threshold type (such as three levels: "normal," "early warning," and "emergency") or a continuous function type (such as increasing the risk weight proportionally as the load rate increases). In this way, the genetic algorithm adaptively adjusts its convergence direction under different operating scenarios, ensuring that the final solution not only meets the current optimization objective but also possesses scenario adaptability.

[0069] S6: Based on the planned output of the iteration results, the execution is fed back to the real-time distribution network operation and maintenance data, triggering rolling re-optimization.

[0070] This embodiment also provides a distribution network operation and maintenance optimization system based on a genetic algorithm, including: a data acquisition unit for acquiring real-time distribution network operation and maintenance data; an analysis unit for calculating maintenance costs and risk losses based on the distribution network operation and maintenance data; establishing an optimization model with the objective of minimizing the weighted average of maintenance costs and post-maintenance risk losses, and subject to resource and network operability constraints; a construction unit for transforming the objective and constraints of the optimization model into a two-layer coupled chromosome, where the first layer represents: maintenance selection-time window-sequence; and the second layer represents: resources-parallelism-transfer-path; a calculation unit for initializing the population with the chromosome, and following the process of commitment → reconstruction → execution, first generating maintenance batches and coarse time windows, then reconstructing and verifying operability with minimal operations and writing back conflicts, followed by path-time window collaborative refinement and handling through a repairer; an adjustment unit for embedding small-scale precise sub-problems for verification during reconstruction and execution, and smoothing convergence with topology adaptive operators and scenario weight gating; and an output unit for outputting the planned results according to the iteration, and feeding back the execution to the real-time distribution network operation and maintenance data to trigger rolling re-optimization.

[0071] This embodiment also provides a computer device applicable to the distribution network operation and maintenance optimization method based on genetic algorithms, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the distribution network operation and maintenance optimization method based on genetic algorithms proposed in the above embodiment.

[0072] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0073] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the distribution network operation and maintenance optimization method based on genetic algorithms as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0074] In summary, this invention achieves the following: First, it acquires real-time operation and maintenance data on equipment health and potential hazards, environment, load, and resources in the distribution network. Second, it establishes an optimization model with the goal of minimizing the weighted average of maintenance costs and post-maintenance risk losses. Third, it maps the optimization model to a two-layer coupled chromosome structure. The first layer performs maintenance task selection, time window allocation, and sequential arrangement. The second layer performs resource allocation, parallel control, power transfer schemes, and path planning. Fourth, it iterates using a genetic algorithm, sequentially executing commitment, reconstruction, and execution processes. It also introduces small-scale precise sub-problem verification, topology adaptive operators, and scenario weight gating mechanisms to achieve solution repair, optimization, and multi-scenario adaptation. Finally, it outputs differentiated operation and maintenance plans and utilizes execution feedback to drive rolling re-optimization, thereby improving the feasibility, economy, and robustness of the distribution network operation and maintenance plan.

[0075] It should be noted that the above 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 preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A distribution network operation and maintenance optimization method based on genetic algorithm, characterized in that: This includes acquiring real-time power distribution network operation and maintenance data; This includes equipment health and potential risks data, environmental data, load data, and resource data; Based on the aforementioned power distribution network operation and maintenance data, calculate the maintenance costs and risk losses; establish an optimization model with the objective of minimizing the weighted average of maintenance costs and post-maintenance risk losses, and subject to resource and network availability constraints. The objective and constraints of the optimization model are transformed into a two-layer coupled chromosome. The first layer represents: maintenance selection - time window - sequence. The second layer represents: Resources-Parallelism-Support-Path; The population is initialized with the chromosomes and follows the process of commitment → reconstruction → execution. First, a batch of maintenance is performed and a coarse time window is set up. Then, the feasibility is verified and conflicts are written back with minimal operations. Subsequently, the process is refined by path-time window collaboration and handled by the repairer. During reconstruction and execution, small-scale precise sub-problems are embedded for verification, and smooth convergence is achieved using topological adaptive operators and scene weight gating. Based on the planned output of the iteration results, the execution is fed back to the real-time distribution network operation and maintenance data, triggering rolling re-optimization.

2. The distribution network operation and maintenance optimization method based on genetic algorithm as described in claim 1, characterized in that: Each device is treated as a node, and a model is built to represent the maintenance costs and risk losses. Based on the health and potential hazards data of the equipment, an n-dimensional equipment feature vector is constructed; using the environmental data and the influence relationship function obtained through pre-training, the parameters of each dimension of the equipment feature vector are adjusted to obtain the adjusted equipment feature vector; The influence relationship function includes the functional relationship between the health and potential risks of the equipment in each dimension and the influence of each type of environmental data; By using a pre-trained Bayesian model and inputting the adjusted device feature vector, the probability distribution of each node's failure is obtained, representing the probability of each type of failure occurring at the node. The maintenance cost is expressed as follows: ; The risk loss is expressed as: ; in, This represents the maintenance cost of the i-th node; This represents the probability that the i-th node will experience a failure of value j. This indicates the risk loss of the i-th node; Indicates the device size at the i-th node Below, the cost of troubleshooting fault j; Indicates the device size at the i-th node The cost of restoring service when fault j occurs; the risk loss after maintenance is: the attenuation rate of the risk loss after maintenance under the current probability distribution learned through prior probability; obtained by multiplying the attenuation rate by the risk loss; The cost of each node during maintenance is a fixed value, and the total resources called by the nodes during maintenance are within the constraints each time maintenance is triggered.

3. The distribution network operation and maintenance optimization method based on genetic algorithm as described in claim 2, characterized in that: The two-layer coupled chromosome is implemented through collaborative encoding of the task layer and the execution layer: The first layer is used to represent the maintenance plan, including the selection of maintenance equipment nodes, the determination of candidate time windows, and the execution order within the same time window. Each task gene corresponds to a device point and records its commitment status and priority label. The second layer is used to represent execution details, including the specific resource configuration allocated to the task, the arrangement of parallel time slots, the selected transfer template, and the movement path of the shift team. Each execution gene corresponds one-to-one with the task gene in the first layer.

4. The distribution network operation and maintenance optimization method based on genetic algorithm as described in claim 3, characterized in that: During chromosome initialization, initial batches and time windows are generated at the first level based on the equipment risk-benefit ratio, and feasible work teams, transfer templates, and shortest travel paths are assigned to each task at the second level. In crossover and mutation operations, the first layer exchanges data on a per-task-gene basis, while the second layer simultaneously adjusts the corresponding resources and paths.

5. The distribution network operation and maintenance optimization method based on genetic algorithm as described in claim 4, characterized in that: The commitment-reconstruction-execution process includes, in the commitment phase, generating a preliminary maintenance batch based on the node selection and time window results of the first layer of chromosomes, and concentrating the equipment of the same feeder segment into a coarse-grained candidate time window; During the reconstruction phase, the transfer template of the second-layer chromosome is called to perform network operability verification for each batch, and the power grid topology reconstruction is achieved by prioritizing the switching sequence with the fewest operations. When mutual exclusion switches or power flow over-limit conflicts are detected, the conflict results and correction suggestions are written back to the chromosome library for reference in subsequent generations of crossover and mutation. During the execution phase, based on the feasible batches confirmed by reconstruction, the paths and parallel time slots of the second-layer chromosome are refined. The path-time window coordinator is used to calculate the travel time and arrival order of the shifts. Problems caused by resource overruns or time window conflicts are handled by the repairer, which completes the correction through insertion, postponement, and task exchange to ensure that the final decoded operation and maintenance plan is executable at both the resource and network levels.

6. The distribution network operation and maintenance optimization method based on genetic algorithm as described in claim 5, characterized in that: The small-scale precise subproblem verification includes randomly sampling chromosome segments within a local range and using mixed integer programming to calculate the local solutions to verify the optimality of the transfer path, resource allocation, and power outage range. The verification results are compared with the solutions generated by the genetic algorithm. If significant room for improvement is found in a local area, the corresponding chromosome segment is replaced. At the same time, the feasible transfer and resource scheduling modes obtained from the local optimization are stored in the template library for rapid use in subsequent populations, thereby improving the reliability and convergence speed of the overall solution.

7. The distribution network operation and maintenance optimization method based on genetic algorithm as described in claim 6, characterized in that: The topology adaptive operator identifies the interlocking relationships between critical feeders and mutual exclusion switches in the distribution network structure, and prioritizes keeping these continuous segments intact during crossover and mutation operations to avoid the solution becoming invalid during the evolution process. The scenario weight gating dynamically adjusts the cost and risk weights in the objective function based on different environmental and load data, increasing the weight of loss under high-risk or high-load conditions.

8. A distribution network operation and maintenance optimization system based on genetic algorithms, based on the distribution network operation and maintenance optimization method based on genetic algorithms according to any one of claims 1 to 7, characterized in that: This includes a data acquisition unit to obtain real-time power distribution network operation and maintenance data; The analysis unit calculates maintenance costs and risk losses based on the power distribution network operation and maintenance data; it establishes an optimization model with the objective of minimizing the weighted average of maintenance costs and post-maintenance risk losses, subject to resource and network availability constraints; the construction unit transforms the objective and constraints of the optimization model into a two-layer coupled chromosome, where the first layer represents: maintenance selection - time window - sequence. The second layer represents: Resources-Parallelism-Support-Path; The computation unit initializes the population with the chromosome and follows the process of commitment → reconstruction → execution. First, it generates maintenance batches and coarse time windows, then reconstructs and verifies the runnability with minimal operations and writes back the conflicts. Subsequently, it refines the process through path-time window collaboration and handles them through the repairer. The adjustment unit embeds small-scale precise sub-problem verification in the reconstruction and execution process and uses topology adaptive operators and scene weight gating to achieve smooth convergence. The output unit outputs the plan based on the iteration results and feeds back the execution to the real-time distribution network operation and maintenance data, triggering rolling re-optimization.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the distribution network operation and maintenance optimization method based on genetic algorithm as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the distribution network operation and maintenance optimization method based on genetic algorithm as described in any one of claims 1 to 7.

Images