Operation and maintenance management method and platform of photovoltaic power generation system

Abstract

The application provides an operation and maintenance management method and platform of a photovoltaic power generation system, relates to the technical field of photovoltaic power station operation and maintenance, and comprises the following steps: obtaining a set of operation and maintenance tasks to be processed and a set of idle operation and maintenance personnel; scheduling and optimizing the operation and maintenance tasks by using an improved NSGA-II algorithm, establishing a multi-objective optimization problem of minimizing total operation and maintenance cost, minimizing total task completion time and minimizing response time, using a clustering enhanced coding structure, a hybrid population initialization strategy and a probability weighted genetic operator to solve the multi-objective optimization, and outputting a Pareto optimal solution set; selecting a final scheme from the Pareto optimal solution set and converting the final scheme into an executable format, and outputting a complete operation and maintenance scheduling scheme including a personnel scheduling list, path planning instructions, a material and tool list and a general time schedule. The application can realize multi-objective collaborative optimization of operation and maintenance cost, response time and execution efficiency.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power plant operation and maintenance technology, and in particular to an operation and maintenance management method and platform for photovoltaic power generation systems. Background Technology

[0002] Currently, the optimization of operation and maintenance (O&M) strategies for photovoltaic (PV) power generation systems mainly focuses on two levels: first, data-driven O&M decision optimization, which involves collecting power plant operation data to predict power generation and analyze equipment failures, providing data support for O&M decisions; and second, the intelligent upgrading of O&M equipment, especially the application of unmanned O&M equipment. Existing technologies have made some progress in O&M cost control, response time optimization, and execution efficiency improvement, but technical bottlenecks still exist in multi-objective collaborative optimization, personalized strategy formulation, and dynamic adaptability.

[0003] Chinese invention patent CN111046321A discloses a method and apparatus for optimizing the operation and maintenance strategy of a photovoltaic power plant. This technical solution obtains the information set of execution results of the operation and maintenance system strategy, determines the optimization result information set with the optimal solution in preset dimensions such as operation and maintenance benefits, operation and maintenance accuracy, and response efficiency as the objective, calculates the strategy optimization factor based on the iteration factor, and iteratively optimizes the operation and maintenance strategy accordingly. This method employs a dual-closed-loop iteration mechanism: the outer layer implements iterative optimization of the operation and maintenance strategy, while the inner layer implements self-learning iteration of the strategy optimization factor and the optimal solution algorithm set. The calculation of the strategy optimization factor combines the self-optimization factor of a single power plant in the time dimension and the multi-power plant optimization factor in the spatiotemporal dimension. However, this patent mainly focuses on macro-level optimization at the strategy level, iteratively improving based on historical execution results, and lacks the ability to optimize specific operation and maintenance task allocation and personnel scheduling in real time. Furthermore, although this method considers multiple optimization dimensions, it does not establish a precise multi-objective mathematical model, making it difficult to handle the collaborative optimization problem of multiple conflicting objectives in operation and maintenance scheduling. Summary of the Invention

[0004] In view of this, the present invention provides a method and platform for operation and maintenance management of photovoltaic power generation systems. By adopting a "list of lists" encoding structure enhanced by "dynamic clustering labels" and a triple algorithm improvement strategy, it solves the technical problems in existing photovoltaic operation and maintenance scheduling methods, such as the encoding structure easily generating invalid solutions due to skill mismatch, the completely random initialization of the population leading to slow convergence speed, the excessive randomness of genetic operators lacking directional search capabilities, and the lack of a solution diversity protection mechanism easily getting trapped in local optima. It achieves multi-objective collaborative optimization of operation and maintenance costs, response time and execution efficiency.

[0005] The technical solution of this invention is implemented as follows:

[0006] On the one hand, the present invention provides a method for operation and maintenance management of a photovoltaic power generation system, including:

[0007] S1. Obtain the set of pending operation and maintenance tasks. and a collection of idle maintenance personnel Each of the operation and maintenance tasks Includes geographic location coordinates, urgency weight, estimated operation time, and resource requirements for each operations and maintenance personnel. Includes skill set, starting location coordinates, and hourly labor cost; This represents the total number of maintenance tasks. Total number of maintenance personnel;

[0008] S2. The operation and maintenance tasks are scheduled and optimized by the improved NSGA-II algorithm. A multi-objective optimization problem is established to minimize the total operation and maintenance cost, the total task completion time, and the response time. The multi-objective optimization is solved by using a clustering-enhanced coding structure, a hybrid population initialization strategy, and a probability-weighted genetic operator, and the Pareto optimal solution set is output.

[0009] S3. By selecting a final solution from the Pareto optimal solution set and converting it into an executable format, the output is a complete operation and maintenance scheduling solution including a personnel scheduling list, path planning instructions, a material and tool list, and an overall time plan.

[0010] Preferably, step S2 includes:

[0011] S21. Set population size Maximum number of iterations Crossover probability Probability of mutation Initial temperature parameters Cluster number K and clustering preference threshold ;

[0012] S22. Use the k-means algorithm to cluster the geographic coordinates of all operation and maintenance tasks to obtain K geographic clusters, and then cluster each operation and maintenance task. Assign cluster labels The clustering preference weight vector is calculated based on the Euclidean distance between the starting position of the maintenance personnel and each cluster center. A clustering-enhanced coding scheme is used to define chromosome X as... ,in To be allocated to maintenance personnel Task sequence, For this person The clustering preference weight vector is used, and clustering constraints are set to limit task allocation;

[0013] S23. By establishing a multi-objective optimization function that minimizes total maintenance cost, total task completion time, and response time, the optimization objectives are determined.

[0014] S24. Generate the initial population using a mixed population initialization strategy. 20% of the individuals were generated using a cluster-constrained greedy seed, and 80% of the individuals were generated using a cluster-constrained Latin hypercube sampling method. The number of iterations was set to t=0.

[0015] S25. By calculating the corrected local fitness of each work order in each chromosome. Based on the modified local fitness and dynamic temperature parameters A probability-weighted selection mechanism is used to select parent individuals for reproduction;

[0016] S26. Perform genetic operations to generate a progeny population. Genetic operations include:

[0017] With crossover probability A dedicated crossover operator for performing task block swapping;

[0018] With the probability of mutation Perform a 2-opt local mutation search and apply a rollback mechanism;

[0019] S27, By merging parent populations and offspring population Forming a joint population Crowding distance is calculated for the joint population, and non-dominated ranking is performed to select superior individuals to form the next generation of the population. ;

[0020] S28. Determine if the iteration stopping condition has been met. If yes, output the current Pareto optimal solution set. If no, set the iteration count t = t + 1 and update the dynamic temperature parameter. Then return to step S25 to continue the iteration.

[0021] Preferably, step S22 includes:

[0022] The k-means algorithm was used to locate the task position. Clustering is performed to obtain K geographical clusters, and each task... Assign a cluster label Among them, the cluster center calculation is as follows:

[0023]

[0024] In the formula, Let k be the cluster center of the k-th geographical cluster. Let k be the number of tasks in the k-th geographic cluster. For the task Geographical coordinates;

[0025] For each operations and maintenance personnel Calculate the preference weight vector for each geographic cluster. :

[0026]

[0027] In the formula, For maintenance personnel The preference weight vector for geographical cluster k, Let be the Euclidean distance function. For maintenance personnel The starting position coordinates, To prevent division by zero of small constants;

[0028] The final weight vector is obtained through normalization:

[0029]

[0030] In the formula, K is the total number of clusters. For clustering index;

[0031] Set clustering constraints: Task Only when the clustering preference condition is met Only then can they be assigned to personnel. ,in This is the clustering preference threshold. For the task Clustering labels, Represents the weight vector The Each component, namely .

[0032] Preferably, the multi-objective optimization function includes:

[0033] Minimize total operating costs:

[0034]

[0035] In the formula, Chromosomes The corresponding total operation and maintenance cost, Total number of maintenance personnel For maintenance personnel hourly labor costs To be allocated to maintenance personnel The task set, For the task The estimated working hours For unit distance transportation cost, For maintenance personnel Total path distance;

[0036] Minimize total task completion time:

[0037]

[0038] In the formula, Chromosomes The corresponding total task completion time, This represents the total number of maintenance tasks. For the task Completion time;

[0039] Minimize response time:

[0040]

[0041] In the formula, Chromosomes The corresponding total response time, For the task The urgency level weight.

[0042] Preferred strategies for initializing mixed populations include:

[0043] Cluster-constrained greedy seed generation accounts for 20% of the population and includes two types: cost-priority seed and urgency-priority seed. The cost-priority seed assigns each task to the nearest maintenance personnel who meets the cluster constraints, while the urgency-priority seed assigns tasks to idle maintenance personnel who meet the cluster constraints in descending order of urgency.

[0044] Clustering constraint Latin hypercube sampling accounts for 80% of the population. The work order length of each maintenance personnel is used as a sampling dimension, and the total dimension is the number of maintenance personnel m. Latin hypercube sampling is performed within the range of task allocation to generate uniformly distributed work order length combinations, ensuring that the population has orthogonal and uniform coverage in the multidimensional search space.

[0045] The greedy task filling mechanism fills the work order with a corresponding number of specific tasks for each maintenance personnel's work order based on the work order length determined by Latin hypercube sampling. The filling process uses a comprehensive greedy score to select tasks, which comprehensively considers skill matching degree, clustering preference weight and geographical distance factors, and prioritizes the allocation of tasks with the highest comprehensive score that meet the clustering constraints.

[0046] Preferably, in step S25, the probability-weighted selection mechanism is as follows:

[0047] Calculate the corrected local fitness for each work order:

[0048]

[0049] In the formula, For maintenance personnel Correct local fitness, For maintenance personnel Sub-cost, For maintenance personnel Subtask completion time, For maintenance personnel Sub-response time, For entropy weighting coefficients, This is the Shannon entropy regularization term;

[0050] Calculate the selection probability using the softmax function:

[0051]

[0052] In the formula, For maintenance personnel The probability of being selected The dynamic temperature parameter for the current iteration. Index for people.

[0053] Preferably, in step S26:

[0054] The dedicated cross operator uses a task block exchange method to select continuous task segments of operation and maintenance personnel work orders from parent A and insert them into operation and maintenance personnel work orders in parent B that satisfy clustering constraints.

[0055] The 2-opt local search mutation operator optimizes the path of the task sequence and employs a rollback mechanism: when the local fitness is mutated... At that time, based on the rollback probability Rollback to the old state, where For the local fitness after mutation, The local fitness before mutation, This represents the current iteration number. This represents the maximum number of iterations.

[0056] Preferably, the iteration stopping condition is reaching the maximum number of iterations. Or it may reach the convergence condition, where the convergence condition is: when the rate of change of the non-dominated solution set within the continuous observation window is less than the convergence threshold, it is determined to be converged.

[0057] Preferably, in step S3, selecting the final execution scheme from the Pareto optimal solution set includes two modes:

[0058] The automatic selection mode automatically selects the optimal solution from the solution set based on preset preferences.

[0059] The manual selection mode presents the Pareto frontier in a visual way for operations managers to make manual selections.

[0060] On the other hand, the present invention also provides an operation and maintenance management platform for a photovoltaic power generation system, the platform being used to implement the method described in any of the above-mentioned embodiments, the platform comprising:

[0061] The data acquisition module is used to acquire real-time data on the operating status of photovoltaic equipment, fault alarm information, location information of maintenance personnel, and skill resource data.

[0062] The task management module is used to receive operation and maintenance task requests, classify tasks, evaluate priorities, and analyze resource requirements.

[0063] The intelligent scheduling module is used to perform multi-objective optimization calculations based on the improved NSGA-II algorithm and generate Pareto optimal solution sets.

[0064] The decision support module provides both automatic and manual selection modes to determine the final execution plan from the Pareto optimal solution set.

[0065] The execution monitoring module is used to output a complete operation and maintenance execution plan, including a personnel scheduling list, route planning instructions, material and tool list, and time schedule, and to monitor the execution status in real time.

[0066] The data storage module is used to store historical operation and maintenance data, optimization parameter configurations, and scheduling scheme records.

[0067] The present invention has the following advantages over the prior art:

[0068] (1) This invention establishes a photovoltaic power generation system operation and maintenance management method based on the improved NSGA-II algorithm, thereby achieving multi-objective collaborative optimization of operation and maintenance costs, response time and execution efficiency;

[0069] (2) The "list of lists" encoding structure enhanced by "dynamic clustering labels" is adopted. By introducing task labels based on k-means geographic clustering and clustering preference weight vectors of operation and maintenance personnel into the encoding, the feasible domain of task allocation is mathematically constrained.

[0070] (3) By combining the cluster-constrained greedy seed generation and the cluster-constrained Latin hypercube sampling hybrid initialization strategy, the slow convergence speed caused by the completely random initialization of the standard NSGA-II algorithm is solved;

[0071] (4) A special genetic operator suitable for clustering enhancement coding structure was designed. By introducing the Shannon entropy regularization term in the local fitness, the diversified allocation of tasks among different geographical clusters is encouraged, preventing the tasks from being overly concentrated in a few maintenance personnel. At the same time, the rollback mechanism avoids the degradation of solution quality during the mutation process, thus improving the overall quality of the Pareto front. Attached Figure Description

[0072] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0073] Figure 1 This is a flowchart of the method of the present invention.

[0074] Figure 2 This is a flowchart illustrating the improved NSGA-II algorithm of the present invention;

[0075] Figure 3 This is a schematic diagram of the platform of the present invention. Detailed Implementation

[0076] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0077] like Figure 1 As shown, the present invention provides an operation and maintenance management method for a photovoltaic power generation system, comprising:

[0078] S1. Obtain the set of pending operation and maintenance tasks. and a collection of idle maintenance personnel Each of the operation and maintenance tasks Includes geographic location coordinates, urgency weight, estimated operation time, and resource requirements for each operations and maintenance personnel. Includes skill set, starting location coordinates, and hourly labor cost; This represents the total number of maintenance tasks. Total number of maintenance personnel;

[0079] S2. The operation and maintenance tasks are scheduled and optimized by the improved NSGA-II algorithm. A multi-objective optimization problem is established to minimize the total operation and maintenance cost, the total task completion time, and the response time. The multi-objective optimization is solved by using a clustering-enhanced coding structure, a hybrid population initialization strategy, and a probability-weighted genetic operator, and the Pareto optimal solution set is output.

[0080] S3. By selecting a final solution from the Pareto optimal solution set and converting it into an executable format, the output is a complete operation and maintenance scheduling solution including a personnel scheduling list, path planning instructions, a material and tool list, and an overall time plan.

[0081] Specifically, in one embodiment of the present invention, step S1 includes:

[0082] S11. Obtaining and defining the set of operation and maintenance tasks.

[0083] First, obtain the set of pending operation and maintenance tasks. ,in This represents the total number of operation and maintenance tasks. Each operation and maintenance task... Includes the following key information:

[0084] Geographic coordinates: mission Geographical coordinates Represented using two-dimensional coordinates. The coordinates can be GPS-based latitude and longitude coordinates or a plane coordinate system relative to a reference point.

[0085] Urgency weighting: using Indicates task The urgency weight is a quantified numerical parameter. A higher value indicates a more urgent task that needs to be prioritized. This weight value is determined based on factors such as the type of fault, the scope of impact, and power generation losses.

[0086] Estimated work hours: using Indicates completion of task The required standard operating time, in hours. This parameter is determined based on historical maintenance data statistics and the standard time quota for the task type.

[0087] Resource requirements: Indicates task The required resource information includes specific skill requirements, spare parts models, and specialized tools. This information is used to ensure that the assigned maintenance personnel have the necessary skills and resources to perform the task.

[0088] S12. Obtaining and defining the set of operation and maintenance personnel.

[0089] Simultaneously acquire a set of available operations and maintenance personnel. ,in This represents the total number of operations and maintenance personnel. Each operations and maintenance personnel... It includes the following attribute information, among which A unique identifier for operations and maintenance personnel:

[0090] Skill set: using Indicates maintenance personnel A set of skills used to meet the resource requirements of a task. Matching is performed. The skills set includes professional skills identifiers such as electrical repair, mechanical maintenance, cleaning operations, and equipment inspection.

[0091] Starting position coordinates: using Indicates maintenance personnel The starting coordinates are set using the same coordinate system as the task location coordinates. This location represents the current location of the operations and maintenance personnel or the designated starting point, and is used to calculate the travel distance and time to each task location.

[0092] Hourly labor cost: Indicates maintenance personnel The hourly labor cost includes the time-allocated value of labor cost elements such as basic wages and welfare expenses.

[0093] S13. Data integrity verification and preprocessing.

[0094] After obtaining the task set and personnel set, perform data integrity verification. For geographic coordinates, verify the validity of the coordinate values ​​and standardize the coordinate system; for urgency weights, normalize the weight values ​​to ensure they are within a reasonable range; for estimated work hours, verify their reasonableness based on task type and historical data.

[0095] In addition, a skills matching matrix was established to compare the skill sets of each operations and maintenance personnel. Resource requirements for each task Feasible personnel-task allocation combinations are pre-selected to provide constraints for subsequent optimization algorithms and avoid generating invalid solutions with skill mismatches.

[0096] Specifically, such as Figure 2 As shown, in one embodiment of the present invention, step S2 includes:

[0097] S21. Set population size Maximum number of iterations Crossover probability Probability of mutation Initial temperature parameters Cluster number K and clustering preference threshold .

[0098] Specifically, in this embodiment, Set to an even number between 100 and 200, the specific value of which can be determined by the product of the task size and the number of maintenance personnel, for example... , where m is the total number of maintenance personnel and n is the total number of tasks. Set it to between 200 and 500 as one of the termination conditions of the algorithm. The value is set between 0.8 and 0.9 to control the probability of crossover operations among individuals in the population. Because this invention uses a dedicated task block exchange crossover operator, a relatively high crossover probability is beneficial for promoting the propagation of superior gene fragments while maintaining the integrity of the coding structure. The value is set between 0.1 and 0.2 to control the probability of an individual undergoing mutation. The mutation probability setting needs to consider that the 2-opt local search mutation operator used in this invention has strong local optimization capabilities; therefore, a moderate mutation probability can maintain population diversity while avoiding excessive perturbation. Set to 10.0 for the softmax function in the probability-weighted selection mechanism. A higher initial temperature helps maintain greater selection randomness in the early stages of the algorithm, preventing premature convergence. The temperature parameter will be dynamically adjusted based on the chaotic mapping during iteration. K is determined based on the geographical distribution of the tasks and the number of maintenance personnel, and is set to... The optimal number of clusters can be determined using the elbow method. Setting it to 0.3 controls the strength of geographical constraints for task assignment.

[0099] S22. Use the k-means algorithm to cluster the geographic coordinates of all operation and maintenance tasks to obtain K geographic clusters, and then cluster each operation and maintenance task. Assign cluster labels The clustering preference weight vector is calculated based on the Euclidean distance between the starting position of the maintenance personnel and each cluster center. A clustering-enhanced coding scheme is used to define chromosome X as... ,in To be allocated to maintenance personnel Task sequence, For this person The clustering preference weight vector is determined, and clustering constraints are set to limit task allocation.

[0100] Specifically, in this embodiment, step S22 includes:

[0101] The k-means algorithm was used to locate the task position. Clustering is performed to obtain K geographical clusters, and each task... Assign a cluster label The specific execution process of k-means clustering includes: first, randomly initializing K cluster centers; then iteratively executing task allocation and center update steps until convergence. The formula for calculating cluster centers is:

[0102]

[0103] In the formula, Let k be the cluster center of the k-th geographical cluster. Let k be the number of tasks in the k-th geographic cluster. For the task The geographic coordinates; the clustering algorithm uses Euclidean distance as a similarity measure, and the iteration termination condition is that the change of cluster center is less than a preset threshold (such as 0.01) or the maximum number of iterations (such as 50 times).

[0104] For each operations and maintenance personnel Calculate the preference weight vector for each geographic cluster. The weights are calculated based on the Euclidean distance between the starting position of the operations and maintenance personnel and each cluster center:

[0105]

[0106] In the formula, For maintenance personnel The preference weight vector for geographical cluster k, Let be the Euclidean distance function. For maintenance personnel The starting position coordinates, To prevent division by zero of small constants;

[0107] The final weight vector is obtained through normalization:

[0108]

[0109] In the formula, K is the total number of clusters. This is a clustering index; the normalized weight vector has the property of probability distribution, and the higher the weight value, the stronger the preference of the operator for the corresponding geographical cluster.

[0110] The scheduling scheme is represented using a cluster-enhanced "list of lists" encoding structure. Chromosome X is defined as: ,in To be allocated to maintenance personnel Task sequence, For this person The clustering preference weight vector, where m is the number of operations and maintenance personnel.

[0111] This encoding structure ensures and (when (Time). Among them, This means that the union of the task lists assigned to all operations and maintenance personnel equals the total set of tasks. (when (When) indicates that there is no overlap between the task lists of different operation and maintenance personnel, i.e., an empty set, and b is the personnel index.

[0112] Set clustering constraints: Task Only when the clustering preference condition is met Only then can they be assigned to personnel. ,in This is the clustering preference threshold. For the task Clustering labels, Represents the weight vector The Each component, namely This constraint mathematically restricts the feasible domain of task allocation, avoiding allocation schemes with overly dispersed geographical locations, effectively reducing the solution space containing "mathematically feasible but practically ineffective" solutions, and improving the algorithm's search efficiency and the practicality of the solutions.

[0113] It should be noted that, It is a task The cluster label is an integer value, such as 1, 2, 3..., which represents the task. Which geographic cluster does it belong to? Where K is the total number of clusters, and k is the cluster index, which is also an integer, used to represent the k-th cluster. If the task... Clustering tags If this task belongs to the third cluster, then for The explanation is that it involves taking the weight vector. The Each component, for example ,but .

[0114] S23. By establishing a multi-objective optimization function that minimizes total maintenance cost, total task completion time, and response time, the optimization objectives are determined.

[0115] Specifically, in this embodiment, the multi-objective optimization function includes:

[0116] The objective function for minimizing total maintenance costs It includes two components: labor costs and transportation costs.

[0117]

[0118] In the formula, Chromosomes The corresponding total operation and maintenance cost, Total number of maintenance personnel For maintenance personnel hourly labor costs To be allocated to maintenance personnel The task set, For the task The estimated working hours For unit distance transportation cost, For maintenance personnel Total path distance;

[0119] The labor cost section calculates the total labor cost required for all maintenance personnel to complete their assigned tasks, while the transportation cost section calculates the total transportation cost incurred by all maintenance personnel during task execution. Path distance. It is calculated based on the positions of adjacent tasks in the task sequence and the starting position.

[0120] The objective function to minimize the total task completion time Also known as "Makespan", it represents the total time required to complete all tasks:

[0121]

[0122] In the formula, Chromosomes The corresponding total task completion time, This represents the total number of maintenance tasks. For the task Completion time;

[0123] Task completion time The calculation needs to consider the departure time of the operations and maintenance personnel performing the task, their travel time to the task location, and the time spent by the personnel performing the task. The cumulative work time and travel time for all other tasks previously completed.

[0124] Response time minimization objective function Consider the urgency level of tasks to ensure that high-urgency tasks receive priority responses:

[0125]

[0126] In the formula, Chromosomes The corresponding total response time, For the task The urgency level weight.

[0127] This objective function effectively penalizes scheduling schemes that cause high-urgency tasks to be delayed by multiplying the completion time of each task by its urgency weight and then summing the results. The higher the urgency weight of a task, the greater the negative impact of its delayed completion on the objective function value, thus guiding the algorithm to prioritize the timely completion of urgent tasks.

[0128] The three objective functions constitute a multi-objective optimization problem:

[0129]

[0130] The solution to this multi-objective optimization problem is a Pareto optimal set, where each solution represents a different trade-off between cost, efficiency, and responsiveness. Since the three objectives are conflicting, there is no single solution that simultaneously optimizes all objectives; therefore, the NSGA-II algorithm is needed to find multiple non-dominated solutions on the Pareto front.

[0131] S24. Generate the initial population using a mixed population initialization strategy. 20% of the individuals were generated using a cluster-constrained greedy seed, and 80% of the individuals were generated using a cluster-constrained Latin hypercube sampling method. The number of iterations was set to t=0.

[0132] Specifically, in this embodiment, the mixed population initialization strategy includes:

[0133] Cluster-constrained greedy seed generation accounts for 20% of the population, including two types: cost-priority seeds and urgency-priority seeds, each accounting for 10%.

[0134] The cost-priority seed assigns each task to the nearest operations and maintenance personnel who meet the clustering constraints. The specific process is as follows: 1. Create a candidate task list, initialized to the set of all tasks. 2. For each unassigned task Calculate the allocation cost among all operations and maintenance personnel: ,in, For maintenance personnel Current position (initially) (After task allocation, update to the position of the last assigned task). 3. Under the condition of satisfying clustering constraints. Skill matching conditions Under the premise of minimizing allocation costs, select the personnel-task pairing for assignment. 4. Update the current location of the operations and maintenance personnel and the candidate task list, and repeat the above process until all tasks are assigned.

[0135] The urgency-priority seeds are allocated to available maintenance personnel who meet the clustering constraints, in descending order of task urgency. The specific process is as follows: 1. Based on the weight of task urgency. 1. Sort all tasks from highest to lowest. 2. For each task, select those that satisfy the clustering constraints. 3. If multiple qualified maintenance personnel exist, select the one with the lightest current workload for assignment. 4. Update the personnel's workload and proceed to the next task.

[0136] Clustering constraints are applied using Latin hypercube sampling, which accounts for 80% of the population. The work order length for each operations and maintenance (O&M) is used as one sampling dimension, with the total dimension being the number of O&M personnel (m). Latin hypercube sampling is performed within the task allocation range to generate uniformly distributed work order length combinations, ensuring orthogonal and uniform coverage of the population in the multidimensional search space. The sampling range for each dimension is set to... Where n is the total number of tasks and m is the total number of operations and maintenance personnel. Upper bound This indicates the maximum number of tasks that each person can perform under an average distribution.

[0137] For the s-th sample individual, the work order length for the j-th maintenance personnel is calculated using the following formula:

[0138]

[0139] in, For the first A permutation of random numbers of dimension, satisfying And for fixed ,all constitute An arrangement. The number of individuals generated by Latin hypercube sampling, i.e. .

[0140] The greedy task filling mechanism uses the work order length determined by Latin hypercube sampling. The system fills in a corresponding number of specific tasks for each maintenance worker's work order. The filling process uses a comprehensive greedy score to select tasks. This score takes into account skill matching, clustering preference weight, and geographical distance factors, and prioritizes the tasks with the highest comprehensive score that meet the clustering constraints for allocation.

[0141] Comprehensive Greedy Score Calculation: For Operations and Maintenance Personnel and candidate tasks The comprehensive greedy score is defined as follows:

[0142]

[0143] in, Let be the skill matching function, when Returns 1 if the condition is met, otherwise returns 0. For maintenance personnel For the task The preference weight of the geographical cluster to which it belongs; The Euclidean distance between the current location of the operations and maintenance personnel and the task location; These are the weighting coefficients; To prevent small constants from being divided by zero.

[0144] The task population process must satisfy the following constraints: only if the following constraints are satisfied Only tasks can be assigned to operations and maintenance personnel. Skill constraint: The resource requirements of a task must match the skill set of the operations and maintenance personnel. Uniqueness constraint: Each task can only be assigned to one operations and maintenance personnel.

[0145] Filling algorithm flow: 1. Initialize the set of unassigned tasks 2. For each operations and maintenance personnel Repeat the following steps Next: Calculate the overall greedy score for the person across all unassigned tasks; select the task with the highest score that satisfies the constraints. ;Will Assigned to ,from Removed from the list; updated the current location of the operations and maintenance personnel. 3. If there are unassigned tasks, randomly select available personnel to assign them.

[0146] The orthogonality of Latin hypercube sampling ensures that the generated individuals have good distribution characteristics in the multidimensional search space. Specifically, the covariance matrix of the initial population is more stable, and the propagation index of the initial Pareto front is improved.

[0147] S25. By calculating the corrected local fitness of each work order in each chromosome. Based on the modified local fitness and dynamic temperature parameters A probability-weighted selection mechanism is used to select parent individuals for reproduction.

[0148] Specifically, in this embodiment, to achieve probability-weighted selection, the global objective function first needs to be decomposed into the local contribution of each operations and maintenance personnel's work order. For operations and maintenance personnel... Its local fitness consists of three components:

[0149] Sub-cost Maintenance personnel The contribution to total operating costs includes labor costs and transportation costs:

[0150]

[0151] in, To be allocated to maintenance personnel The task set, The total path distance for this person to perform all assigned tasks. Path distance calculation needs to consider the distance from the starting position. The distance to the first mission location, and the distance traveled between missions.

[0152] Subtask completion time Maintenance personnel Total time required to complete all assigned tasks:

[0153]

[0154] in, For maintenance personnel Complete the task The time taken from the starting position or the previous task position to the task The travel time and the operation time for carrying out the task .

[0155] Sub-response time Maintenance personnel Weighted response time for task allocation:

[0156]

[0157] in, For the task The urgency level weight.

[0158] To prevent tasks from being overly concentrated on a small number of operations personnel, a Shannon entropy regularization term based on geographic clustering is introduced. This regularization term encourages diverse task distribution across different geographic clusters.

[0159] For maintenance personnel Calculate the task proportion of each geographic cluster in its work order:

[0160]

[0161] in, Indicates maintenance personnel The work order belongs to a geographical cluster The proportion of tasks, For the task Clustering labels, The total number of tasks assigned to this person.

[0162] Calculate Shannon entropy based on the task ratio of each cluster:

[0163]

[0164] in, This represents the total number of geographical clusters. A small constant to prevent errors in logarithmic operations. This is especially important as task allocation becomes more diverse (i.e., distributed across more different geographical clusters). The higher the value, the better the diversity.

[0165] Combining the three local objective components and the Shannon entropy regularization term, the corrected local fitness is obtained:

[0166]

[0167] in, This is the entropy weighting coefficient, used to balance optimization performance and allocation diversity.

[0168] Using the modified local fitness and dynamic temperature parameters Calculate the probability of each operations and maintenance personnel being selected:

[0169]

[0170] In the formula, For maintenance personnel The probability of being selected The dynamic temperature parameter for the current iteration. This indexes people. The negative sign indicates that the lower the fitness (i.e., the better the performance), the higher the probability of selection.

[0171] Temperature parameters Controlling the degree of randomness in the selection process: when When the value is large, the selection probability tends to be uniformly distributed, increasing the randomness of the search; when... When the population is smaller, select individuals with better fitness to increase the targeting of the search.

[0172] Parent selection process: 1. Calculate the corrected local fitness of all individuals in the current population. 2. Based on the current temperature parameters Calculate the probability of selection 3. Use the roulette wheel selection method to select parent individuals based on probability distribution; 4. Repeat the selection process until the required number of parent individuals are obtained.

[0173] Compared to traditional random selection or tournament selection, this probability-weighted selection mechanism can better utilize individual fitness information, while maintaining appropriate selection pressure through dynamic temperature regulation, effectively improving the algorithm's search efficiency and solution quality.

[0174] S26. Perform genetic operations to generate a progeny population. Genetic operations include:

[0175] With crossover probability A dedicated crossover operator for performing task block swapping;

[0176] With the probability of mutation Perform a 2-opt local mutation search and apply a rollback mechanism.

[0177] Specifically, in this embodiment, the task block swapping dedicated crossover operator is as follows:

[0178] Randomly select an operations and maintenance personnel from parent A. The work order is then used to randomly select a continuous task segment from that work order. Then, another maintenance worker who satisfies the clustering constraints is randomly selected from parent B. (i.e., for each task in the segment) All are satisfied ), insert the task fragment into The random location of the work order, and at the same time These tasks that already existed in the work order are removed to ensure that tasks are not assigned repeatedly.

[0179] The 2-opt local search mutation operator is as follows:

[0180] Randomly select a work order from a specific chromosome for an operations and maintenance personnel, and apply the 2-opt algorithm to this task list for local path optimization. During the iteration of 2-opt, if the mutated local fitness... Then, based on the rollback probability Roll back to the old state to avoid degradation. For the local fitness after mutation, The local fitness before mutation, This represents the current iteration number. This represents the maximum number of iterations.

[0181] S27, By merging parent populations and offspring population Forming a joint population Crowding distance is calculated for the joint population, and non-dominated ranking is performed to select superior individuals to form the next generation of the population. .

[0182] Specifically, in this embodiment, the population merging process maintains complete information about all individuals, including attributes such as chromosome encoding, objective function value, and local fitness.

[0183] For any two individuals in the joint population and ,individual Dominant Individual (recorded as) If and only if:

[0184]

[0185]

[0186] in, These are the objective functions for total maintenance cost, total task completion time, and response time, respectively.

[0187] The joint population is stratified using a fast non-dominated sorting algorithm:

[0188] First floor ( ): Identify all non-dominated individuals, i.e., individuals not dominated by any other individuals; Subsequent layer construction: Continue searching for non-dominated individuals from the remaining individuals until all individuals are assigned to the appropriate layer; Layer labeling: Assign a non-dominated level to each individual. , Individuals in , Individuals in And so on.

[0189] Crowded distance Used to measure individuals The density in the target space is calculated using the following formula:

[0190]

[0191] in, and For individuals In the Neighboring individuals on a target and These are the maximum and minimum values ​​of the target in the current layer, respectively.

[0192] Detailed calculation steps: For each non-dominated layer Calculate the crowding distance for each individual; for each objective function Individuals in this layer are sorted by Sort by values ​​in ascending order; set the crowding distance between the first and last individuals after sorting to infinity; for intermediate individuals, sum their local crowding distances on each target.

[0193] From the joint population, according to the following priority order Select Individuals form the next generation population. Prioritize smaller ones Individuals with the same value; in the same Among the individuals, prioritize the larger ones. Individuals with value.

[0194] S28. Determine if the iteration stopping condition has been met. If yes, output the current Pareto optimal solution set. If no, set the iteration count t = t + 1 and update the dynamic temperature parameter. Then return to step S25 to continue the iteration.

[0195] Specifically, in this embodiment, the iteration stopping condition is reaching the maximum number of iterations. Or it may reach the convergence condition, where the convergence condition is: when the rate of change of the non-dominated solution set within the continuous observation window is less than the convergence threshold, it is determined to be converged.

[0196] The temperature parameter adopts a dynamic adjustment strategy based on Logistic chaotic mapping:

[0197]

[0198] in, The parameter is used to define the chaos, causing the system to exhibit chaotic characteristics. For the first Temperature parameters of the generation, initial values ; To determine the amplitude of the disturbance, a periodic disturbance is introduced; This represents the current iteration number. This represents the maximum number of iterations.

[0199] To prevent temperature parameters from deviating excessively from the reasonable range, boundary constraints are set:

[0200]

[0201] It should be noted that the upper and lower bounds in the boundary constraints are empirical values. A lower bound of 0.1 ensures that the selection mechanism maintains a minimum degree of randomness, while an upper bound of 50.0 ensures that the algorithm can still utilize fitness information for biased selection during the initial and perturbation phases. Furthermore, in practice, the upper bound value should match the numerical range of the local fitness, allowing... The calculation is reasonable.

[0202] When the algorithm meets the termination condition, it will start from the current population. Extract the first non-dominated layer As the Pareto optimal solution set:

[0203]

[0204] If the termination condition is not met, perform the following update operation: Increment the iteration counter: Temperature parameter update: Apply the temperature update formula described above. Return to step S25 to begin a new round of probability-weighted selection, forming a complete evolutionary cycle.

[0205] Specifically, in one embodiment of the present invention, step S3 selects the final execution scheme from the Pareto optimal solution set and converts it into an executable format to output a complete operation and maintenance scheduling scheme.

[0206] The final selection of the solution includes both automatic selection mode and manual selection mode:

[0207] Automatic Selection Mode: Automatically selects the optimal solution from the solution set based on preset preferences. A multi-objective decision function calculates a comprehensive score for each solution, selecting the solution with the highest comprehensive score as the final execution plan. The decision function considers a weighted combination of three objectives: total maintenance cost F1, total task completion time F2, and response time F3. The weights can be configured according to actual maintenance strategy requirements.

[0208] Manual Selection Mode: The Pareto frontier is presented to the operations manager in a visual format for manual selection. The visualization interface displays a 3D scatter plot, with each point representing a Pareto optimal solution, and the three axes corresponding to the three optimization objectives. The operations manager can view detailed information for each solution and make decisions based on the current operational strategy and resource status.

[0209] Once the final scheme is determined, the chromosome encoding format will be converted into a standardized execution scheme:

[0210] Personnel Dispatch List: for each operations and maintenance personnel Generate independent work orders, including basic personnel information and a list of assigned tasks. Detailed task information (location coordinates) Urgency weight Estimated working hours Resource requirements (and clustering assignment information).

[0211] Path planning instruction: Calculates the optimal navigation path based on each operations and maintenance personnel's task sequence. Starting from the operations and maintenance personnel's starting position... Once started, standardized navigation instructions containing GPS coordinates, estimated travel distance, and time are generated based on the location coordinates in the mission sequence.

[0212] Material and Tool List: Based on the resource requirements of all assigned tasks The system compiles and generates a bill of materials. It summarizes the resource requirements of all tasks, removes duplicates, and generates a personalized bill of materials for each operations and maintenance personnel.

[0213] Overall Timeline: Create a Gantt chart-style timeline containing the schedules for all tasks. For each task... Calculate start time and duration Completion Time It generates an independent timeline for each operations and maintenance personnel, showing the task execution order and schedule.

[0214] In addition, such as Figure 3 As shown, the present invention also provides an operation and maintenance management platform for a photovoltaic power generation system, the platform being used to implement the method described in any one of the above methods, the platform comprising:

[0215] The data acquisition module is used to acquire real-time operating status data of photovoltaic equipment, fault alarm information, location information of maintenance personnel, and skill resource data. This module connects to the photovoltaic monitoring system, GPS positioning system, and human resource management system to collect real-time operating status data of photovoltaic equipment (such as power generation, equipment temperature, current, and voltage), fault alarm information (including fault type, location coordinates, and urgency level), and location information of maintenance personnel (starting and ending coordinates). ) and skill resource data (skill sets) hourly labor cost The module uses a standardized data interface protocol to ensure data accuracy and real-time performance.

[0216] The task management module receives operation and maintenance task requests, classifies tasks, assesses priorities, and analyzes resource requirements. This module also receives fault alarms and periodic maintenance requests from the data acquisition module and automatically generates a set of operation and maintenance tasks. Each task Includes geographic coordinates Urgency weight Estimated working hours and resource requirements The module categorizes and prioritizes tasks based on fault type, impact scope, and historical experience, providing standardized task data for subsequent intelligent scheduling.

[0217] The intelligent scheduling module performs multi-objective optimization computation based on the improved NSGA-II algorithm to generate a Pareto optimal solution set. This module executes the complete algorithm flow of step S2, including parameter initialization, k-means geographic clustering, clustering enhancement encoding, mixed population initialization, probability-weighted selection, genetic operations, and environment selection. The module employs a parallel computing architecture to improve algorithm execution efficiency, handling large-scale task scheduling problems through multi-threading. The final output is a Pareto optimal solution set containing multiple non-dominated solutions, providing decision-makers with diverse scheduling scheme options.

[0218] The decision support module provides two modes: automatic selection and manual selection. It determines the final execution plan from the Pareto optimal solution set. In the automatic selection mode, the comprehensive score is automatically calculated based on preset preference weights to select the optimal solution. In the manual selection mode, the Pareto frontier is displayed through a 3D visualization interface. Each solution is displayed in scatter form with three target values: total maintenance cost, total task completion time, and response time. The maintenance manager can interactively view the detailed information of the solution and make manual selections.

[0219] The execution monitoring module is used to output a complete operation and maintenance execution plan, including a personnel scheduling list, route planning instructions, material and tool list, and time schedule, and to monitor the execution status in real time.

[0220] The data storage module stores historical operation and maintenance data, optimized parameter configurations, and scheduling plan records. This module stores historical operation and maintenance data for experience accumulation and data mining, optimized parameter configuration information for algorithm tuning, and scheduling plan records for performance evaluation and continuous improvement.

[0221] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for operation and maintenance management of a photovoltaic power generation system, characterized in that, include: S1. Obtain the set of pending operation and maintenance tasks. and a collection of idle maintenance personnel Each of the operation and maintenance tasks Includes geographic location coordinates, urgency weight, estimated operation time, and resource requirements for each operations and maintenance personnel. Includes skill set, starting location coordinates, and hourly labor cost; This represents the total number of maintenance tasks. Total number of maintenance personnel; S2. The operation and maintenance tasks are scheduled and optimized by the improved NSGA-II algorithm. A multi-objective optimization problem is established to minimize the total operation and maintenance cost, the total task completion time, and the response time. The multi-objective optimization is solved by using a clustering-enhanced coding structure, a hybrid population initialization strategy, and a probability-weighted genetic operator, and the Pareto optimal solution set is output. S3. By selecting a final solution from the Pareto optimal solution set and converting it into an executable format, the output is a complete operation and maintenance scheduling solution including a personnel scheduling list, path planning instructions, a material and tool list, and an overall time plan. Step S2 includes: S21. Set population size Maximum number of iterations Crossover probability Probability of mutation Initial temperature parameters Cluster number K and clustering preference threshold ; S22. Use the k-means algorithm to cluster the geographic coordinates of all operation and maintenance tasks to obtain K geographic clusters, and then cluster each operation and maintenance task. Assign cluster labels The clustering preference weight vector is calculated based on the Euclidean distance between the starting position of the maintenance personnel and each cluster center. A clustering-enhanced coding scheme is used to define chromosome X as... ,in To be allocated to maintenance personnel Task sequence, For this person The clustering preference weight vector is used, and clustering constraints are set to limit task allocation; S23. By establishing a multi-objective optimization function that minimizes total maintenance cost, total task completion time, and response time, the optimization objectives are determined. S24. Generate the initial population using a mixed population initialization strategy. 20% of the individuals were generated using a cluster-constrained greedy seed, and 80% of the individuals were generated using a cluster-constrained Latin hypercube sampling method. The number of iterations was set to t=0. S25. By calculating the corrected local fitness of each work order in each chromosome. Based on the modified local fitness and dynamic temperature parameters A probability-weighted selection mechanism is used to select parent individuals for reproduction; S26. Perform genetic operations to generate a progeny population. Genetic operations include: With crossover probability A dedicated crossover operator for performing task block swapping; With the probability of mutation Perform a 2-opt local mutation search and apply a rollback mechanism; S27, By merging parent populations and offspring population Forming a joint population Crowding distance is calculated for the joint population, and non-dominated ranking is performed to select superior individuals to form the next generation of the population. ; S28. Determine if the iteration stopping condition has been met. If yes, output the current Pareto optimal solution set. If no, set the iteration count t = t + 1 and update the dynamic temperature parameter. Then return to step S25 to continue the iteration.

2. The operation and maintenance management method for a photovoltaic power generation system according to claim 1, characterized in that, Step S22 includes: The k-means algorithm was used to locate the task position. Clustering is performed to obtain K geographical clusters, and each task... Assign a cluster label Among them, the cluster center calculation is as follows: ； In the formula, Let k be the cluster center of the k-th geographical cluster. Let k be the number of tasks in the k-th geographic cluster. For the task Geographical coordinates; For each operations and maintenance personnel Calculate the preference weight vector for each geographic cluster. : ； In the formula, For maintenance personnel The preference weight vector for geographical cluster k, Let be the Euclidean distance function. For maintenance personnel The starting position coordinates, To prevent division by zero of small constants; The final weight vector is obtained through normalization: ； In the formula, K is the total number of clusters. For clustering index; Set clustering constraints: Task Only when the clustering preference condition is met Only then can they be assigned to personnel. ,in This is the clustering preference threshold. For the task Clustering labels, Represents the weight vector The Each component, namely .

3. The operation and maintenance management method for a photovoltaic power generation system according to claim 1, characterized in that, Multi-objective optimization functions include: Minimize total operating costs: ； In the formula, Chromosomes The corresponding total operation and maintenance cost, Total number of maintenance personnel For maintenance personnel hourly labor costs To be allocated to maintenance personnel The task set, For the task The estimated working hours For unit distance transportation cost, For maintenance personnel Total path distance; Minimize total task completion time: ； In the formula, Chromosomes The corresponding total task completion time, This represents the total number of maintenance tasks. For the task Completion time; Minimize response time: ； In the formula, Chromosomes The corresponding total response time, For the task The urgency level weight.

4. The operation and maintenance management method for a photovoltaic power generation system according to claim 1, characterized in that, Mixed population initialization strategies include: Cluster-constrained greedy seed generation accounts for 20% of the population and includes two types: cost-priority seed and urgency-priority seed. The cost-priority seed assigns each task to the nearest maintenance personnel who meets the cluster constraints, while the urgency-priority seed assigns tasks to idle maintenance personnel who meet the cluster constraints in descending order of urgency. Clustering constraint Latin hypercube sampling accounts for 80% of the population. The work order length of each maintenance personnel is used as a sampling dimension, and the total dimension is the number of maintenance personnel m. Latin hypercube sampling is performed within the range of task allocation to generate uniformly distributed work order length combinations, ensuring that the population has orthogonal and uniform coverage in the multidimensional search space. The greedy task filling mechanism fills the work order with a corresponding number of specific tasks for each maintenance personnel's work order based on the work order length determined by Latin hypercube sampling. The filling process uses a comprehensive greedy score to select tasks, which comprehensively considers skill matching degree, clustering preference weight and geographical distance factors, and prioritizes the allocation of tasks with the highest comprehensive score that meet the clustering constraints.

5. The operation and maintenance management method for a photovoltaic power generation system according to claim 3, characterized in that, In step S25, the probability-weighted selection mechanism is as follows: Calculate the corrected local fitness for each work order: ； In the formula, For maintenance personnel Correct local fitness, For maintenance personnel Sub-cost, For maintenance personnel Subtask completion time, For maintenance personnel Sub-response time, For entropy weighting coefficients, This is the Shannon entropy regularization term; Calculate the selection probability using the softmax function: ； In the formula, For maintenance personnel The probability of being selected The dynamic temperature parameter for the current iteration. Index for people.

6. The operation and maintenance management method for a photovoltaic power generation system according to claim 5, characterized in that, In step S26: The dedicated cross operator uses a task block exchange method to select continuous task segments of operation and maintenance personnel work orders from parent A and insert them into operation and maintenance personnel work orders in parent B that satisfy clustering constraints. The 2-opt local search mutation operator optimizes the path of the task sequence and employs a rollback mechanism: when the local fitness is mutated... At that time, based on the rollback probability Rollback to the old state, where For the local fitness after mutation, The local fitness before mutation, This represents the current iteration number. This represents the maximum number of iterations.

7. The operation and maintenance management method for a photovoltaic power generation system according to claim 1, characterized in that, The iteration stopping condition is reaching the maximum number of iterations. Or it may reach the convergence condition, where the convergence condition is: when the rate of change of the non-dominated solution set within the continuous observation window is less than the convergence threshold, it is determined to be converged.

8. The operation and maintenance management method for a photovoltaic power generation system according to claim 1, characterized in that, In step S3, selecting the final execution plan from the Pareto optimal solution set includes two modes: The automatic selection mode automatically selects the optimal solution from the solution set based on preset preferences. The manual selection mode presents the Pareto frontier in a visual way for operations managers to make manual selections.

9. An operation and maintenance management platform for a photovoltaic power generation system, characterized in that, The platform is used to implement the method as described in any one of claims 1-8, the platform comprising: The data acquisition module is used to acquire real-time data on the operating status of photovoltaic equipment, fault alarm information, location information of maintenance personnel, and skill resource data. The task management module is used to receive operation and maintenance task requests, classify tasks, evaluate priorities, and analyze resource requirements. The intelligent scheduling module is used to perform multi-objective optimization calculations based on the improved NSGA-II algorithm and generate Pareto optimal solution sets. The decision support module provides both automatic and manual selection modes to determine the final execution plan from the Pareto optimal solution set. The execution monitoring module is used to output a complete operation and maintenance execution plan, including a personnel scheduling list, route planning instructions, material and tool list, and time schedule, and to monitor the execution status in real time. The data storage module is used to store historical operation and maintenance data, optimization parameter configurations, and scheduling scheme records.

Images