A method and system for intelligent operation and maintenance scheduling of distributed photovoltaic power plants

Abstract

This invention relates to the field of photovoltaic power plant operation and maintenance, specifically to an intelligent operation and maintenance scheduling method and system for distributed photovoltaic power plants. An intelligent operation and maintenance scheduling system for distributed photovoltaic power plants includes: a task generation module, a task aggregation module, and a dynamic scheduling and execution module. This invention overcomes the technical challenges of inaccurate fault location and difficulty in monitoring latent degradation in distributed photovoltaic power plants by establishing a collaborative diagnostic mechanism that integrates string-level IV curve analysis and UAV image recognition. This method accurately correlates and cross-validates electrical characteristics and visual features in the spatiotemporal dimension, enabling precise identification of the smallest fault unit and clear distinction of fault types. Simultaneously, it effectively isolates environmental noise, achieving early warning of slow component performance degradation, thereby upgrading the operation and maintenance mode from overall extensive monitoring to component-level refined management.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power plant operation and maintenance, and specifically to an intelligent operation and maintenance scheduling method and system for distributed photovoltaic power plants. Background Technology

[0002] Distributed photovoltaic power stations face significant challenges in operation and maintenance due to their dispersed layout, heterogeneous equipment, and substantial environmental impact. Currently, the commonly used operation and maintenance technologies in the industry have several key shortcomings, which directly restrict the improvement of operation and maintenance efficiency and the economic benefits of the power station. Specifically:

[0003] First, at the fault diagnosis level, existing technologies lack the ability to perform refined analysis by fusing multi-source data, resulting in vague fault location and difficulty in detecting hidden performance degradation. Currently, most mainstream monitoring systems rely on summary alarms and total power generation data reported by inverters, and cannot acquire and deeply analyze core electrical characteristic data that can reflect the health status of components. At the same time, for appearance faults such as dust accumulation, hot spots, and physical damage, they rely heavily on low-frequency manual inspections or single image recognition, failing to align electrical performance data with visual inspection data in time and space and to make comprehensive judgments. This makes it impossible for the system to accurately distinguish fault types, locate the smallest fault unit, and effectively identify and warn of the slow and hidden power generation performance degradation caused by component aging, PID effect, etc., resulting in a continuous "boiling frog" loss of power generation revenue.

[0004] Second, at the operation and maintenance scheduling level, the existing model is highly passive, fragmented, and experience-based, failing to achieve proactive optimization and resource coordination based on economic assessment. Current methods typically aim to handle individual emergency alarms, adopting a passive "order-dispatch" response model. Scheduling decisions are often based on human experience, which cannot scientifically assess and prioritize a large number of scattered, non-urgent operation and maintenance needs, nor does it have an effective mechanism to intelligently aggregate these scattered tasks in the spatiotemporal dimension to form composite work orders with economies of scale. At the same time, after task dispatch, there is a lack of dynamic monitoring and triggering mechanisms for real-time resource status, external environment, and economic thresholds, resulting in suboptimal operation and maintenance path planning, low resource utilization, and high overall costs.

[0005] To address the two major shortcomings mentioned above, this invention proposes an intelligent operation and maintenance scheduling method and system for distributed photovoltaic power plants. Summary of the Invention

[0006] This invention overcomes the technical challenges of inaccurate fault location and difficulty in monitoring latent degradation in distributed photovoltaic power stations by establishing a collaborative diagnostic mechanism that integrates string-level IV curve analysis and UAV image recognition. The method accurately correlates and cross-validates electrical characteristics and visual features in the spatiotemporal dimension, enabling precise identification of the smallest fault unit and clear distinction of fault types. At the same time, it effectively isolates environmental noise and achieves early warning of slow degradation of component performance, thereby upgrading the operation and maintenance mode from overall extensive monitoring to component-level refined management.

[0007] A smart operation and maintenance scheduling method for distributed photovoltaic power plants includes:

[0008] Step S100: Collect operational and environmental data from multiple distributed photovoltaic power stations; wherein, the operational data includes string-level current, voltage, and power data, as well as inverter status and alarm information, and the environmental data includes irradiance, ambient temperature, and weather forecast information of the power station location;

[0009] Based on the aforementioned operational and environmental data, the operation and maintenance needs of each power station are diagnosed and identified.

[0010] For each of the aforementioned maintenance requirements, a preliminary maintenance task item is generated. The preliminary maintenance task item includes at least the fault type, geographical location information, estimated processing time, recommended processing time window, and an economic priority evaluation value. The economic priority evaluation value is calculated based on the ratio of the expected power generation revenue generated from processing the maintenance requirement to the estimated maintenance cost.

[0011] Step S200: Aggregate multiple preliminary operation and maintenance task items that meet the preset geographical proximity conditions to form a regional task cluster;

[0012] Based on the weather forecast information and power grid dispatch plan, a suitable maintenance time window for outdoor operations is determined;

[0013] Multiple preliminary operation and maintenance tasks belonging to the same regional task cluster and whose recommended processing time window overlaps with the maintenance time window are integrated and their work paths are planned to generate a composite operation and maintenance work order. The composite operation and maintenance work order has an overall economic evaluation value, which is higher than the economic priority evaluation value of any single preliminary operation and maintenance task item it contains.

[0014] Step S300: Place the composite operation and maintenance work order into the task queue to be executed, and configure it with trigger execution conditions including at least the readiness status of associated operation and maintenance resources, no negative external event interference, and overall economic performance.

[0015] The triggering conditions are continuously monitored. When all conditions are met, the composite maintenance work order is activated from the queue of tasks to be executed and sent to the corresponding maintenance terminal for scheduling and execution.

[0016] Dynamically optimize and adjust composite operation and maintenance work orders that are in the execution state.

[0017] Preferably, the operational data also includes photovoltaic module image data collected by image sensors deployed on maintenance drones or fixed monitoring points;

[0018] The process of diagnosing and identifying the operation and maintenance needs of each power station based on operational and environmental data includes the following sub-steps:

[0019] Step S101: Based on the single-diode equivalent circuit model of the photovoltaic module, the core physical parameters of the model include photocurrent, diode saturation current, series resistance, and parallel resistance; using the real-time irradiance and real-time ambient temperature in the environmental data, the nominal values ​​of the core physical parameters are corrected; the corrected parameters are substituted into the current-voltage equation of the single-diode equivalent circuit model, and the theoretical IV curve of the photovoltaic string under standard healthy conditions is generated through numerical calculation;

[0020] Step S102: Based on the collected string current and voltage data, plot the actual IV curve and extract the set of characteristic parameters from the actual IV curve. The set of characteristic parameters includes short-circuit current, open-circuit voltage, maximum power point current and maximum power point voltage.

[0021] Step S103: Compare the values ​​of each parameter in the set of feature parameters extracted from the actual IV curve with the corresponding theoretical feature parameters in the theoretical IV curve; when the relative deviation between the measured value and the theoretical value of any feature parameter exceeds the first threshold preset for that parameter, it is determined that the photovoltaic string has a fault or performance abnormality, and a first diagnostic result is generated, which includes the abnormal string identifier and the abnormality type.

[0022] Step S104: Input the photovoltaic module image data into a pre-trained image recognition model. The model outputs a state classification result for each photovoltaic module in the image, including normal, dust accumulation, shadow occlusion, or physical damage. A second diagnostic result is generated, which includes abnormal component identification and state classification.

[0023] Step S105: For the same spatial location, perform spatiotemporal alignment and logical association between the first diagnostic result and the second diagnostic result;

[0024] When only the first diagnostic result exists, the generated maintenance requirement point is an electrical performance anomaly based on the IV curve;

[0025] When only a second diagnostic result exists, the generated maintenance requirement point is an abnormal appearance state based on the image;

[0026] When the first diagnostic result and the second diagnostic result both point to the same component or string, the generated maintenance requirement point is a composite anomaly, and the status classification result is used as a refined description of the fault type.

[0027] Preferably, in step S200, multiple preliminary operation and maintenance task items that meet preset geographical proximity conditions are aggregated to form a regional task cluster. Specific operations include:

[0028] Step S201: Convert the geographic location information contained in each preliminary operation and maintenance task item into unified latitude and longitude coordinates;

[0029] Step S202: Based on the latitude and longitude coordinates of all preliminary maintenance tasks to be processed, a density-based spatial clustering algorithm is used to group them. The core parameters of the density-based spatial clustering algorithm include the neighborhood search radius and the minimum task number threshold. When the straight-line distance between any two tasks in a group of preliminary maintenance tasks is not greater than the neighborhood search radius, and the total number of tasks in the group is not less than the minimum task number threshold, the group of tasks is clustered into a regional task cluster.

[0030] Step S203: Calculate a location coordinate representing the geographical center of each formed regional task cluster for subsequent job path planning.

[0031] Preferably, in step S200, generating a composite maintenance work order specifically includes the following sub-steps:

[0032] Step S204: For any regional task cluster, select all preliminary operation and maintenance tasks whose recommended processing time window overlaps with the determined maintenance time window; integrate these tasks and their included fault types and estimated processing time information to form a set of tasks to be planned.

[0033] Step S205: Using the representative coordinates of the regional task cluster as the starting and ending points of the path, the specific geographical location of each task item in the set of tasks to be planned as the necessary path points, and minimizing the sum of the total estimated processing time and the total path travel time of all task items as the optimization objective, the Traveling Salesman Problem algorithm is used to calculate and generate the optimal job sequence and the corresponding estimated total working time.

[0034] Step S206: Based on the set of tasks to be planned, the optimal work sequence, and the estimated total working hours, generate the composite operation and maintenance work order; wherein, the overall economic evaluation value of the composite operation and maintenance work order is calculated by dividing the sum of the expected power generation revenue of all task items in the set by the total operation and maintenance cost calculated based on the estimated total working hours and the unit working hour cost.

[0035] Preferably, configuring trigger execution conditions for the composite maintenance work order and monitoring its execution specifically includes the following sub-steps:

[0036] Step S301: Configure a set of quantifiable trigger execution conditions for each composite maintenance work order placed in the task queue. These conditions include:

[0037] Resource readiness conditions: The distance between the real-time geographical location of at least one operations and maintenance team and the representative coordinates of the task cluster in the region to which the composite operations and maintenance work order belongs is less than a preset distance threshold, and the team's current status is "idle" or "about to complete the task".

[0038] No negative external event conditions: Based on weather forecast information, confirm that there are no forecasts of precipitation or wind levels below the safe operation threshold within the recommended processing time window of the composite operation and maintenance work order; and confirm that the power grid dispatch system has not issued any power rationing instructions that would affect the shutdown of power plants in this area;

[0039] Overall economic efficiency criteria: The overall economic efficiency evaluation value of the composite maintenance work order is not lower than a preset economic threshold.

[0040] Step S302: Continuously monitor the triggering execution conditions of each composite maintenance work order in the queue of tasks to be executed; if and only if all the triggering execution conditions of a composite maintenance work order are met at the same time, the system automatically transitions its status from "pending triggering" to "executable";

[0041] Step S303: Assign the composite operation and maintenance work order that transitions to "executable" to the operation and maintenance team that meets the resource ready state condition. The assignment is based on the calculation of the comprehensive dispatch score of each candidate team. The formula for calculating the comprehensive dispatch score Z is: Z=α*(1 / D)+β*S.

[0042] Where D is the real-time path distance between the team's real-time geographical location and the representative coordinates of the regional task cluster, S is the percentage of skill items in the team's skill library that match the fault types included in the composite maintenance work order, and α and β are preset weighting coefficients.

[0043] The composite maintenance work order is assigned to the maintenance team with the highest comprehensive dispatch score Z, and the detailed operation instructions of the work order are sent to the maintenance terminal of the corresponding team.

[0044] Preferably, the calculation of the economic priority evaluation value of the preliminary maintenance task item and the overall economic evaluation value of the composite maintenance work order are both achieved through the following unified benefit and cost calculation steps:

[0045] Calculation of expected power generation revenue:

[0046] For any point of operation and maintenance requirement or a set of tasks consisting of it, the expected power generation revenue is determined through the following steps:

[0047] Step 1: Combine the fault type corresponding to the maintenance requirement point, the historical power generation data sequence of the fault point, and the irradiance prediction sequence within the future preset time period to form a multi-dimensional feature vector;

[0048] Step 2: Input the multi-dimensional feature vector into a pre-trained power generation recovery prediction model; the power generation recovery prediction model is a neural network model based on a long short-term memory network architecture, which is trained through historical fault samples and is used to map the nonlinear relationship between fault features and recoverable power generation.

[0049] Step 3: The power generation recovery prediction model outputs a value, which is the expected power generation that can be recovered in the future within a preset period after the demand is processed; multiply the expected recoverable power generation by the preset electricity price, and the result is the expected power generation revenue;

[0050] Calculation of estimated operation and maintenance costs:

[0051] For a single initial maintenance task, the estimated maintenance cost consists of labor cost and material cost. Specifically, the estimated processing time is multiplied by the preset unit labor cost, and the material cost matched from the standard material library according to the fault type is added.

[0052] For a composite maintenance work order, its estimated total maintenance cost consists of total working hours cost, total material cost and mileage cost. Specifically, it is as follows: multiply the estimated total working hours generated by the work path planning in step S200 by the comprehensive cost per unit working hour, add the sum of the material costs matched by all task items in the work order, and add the product of the total mileage determined based on the work path planning and the preset cost per unit mileage.

[0053] Calculation of economic priority evaluation value:

[0054] The economic priority evaluation value of a single initial operation and maintenance task item is calculated by dividing the expected power generation revenue of the task item by its estimated operation and maintenance cost;

[0055] The overall economic evaluation value of a composite operation and maintenance work order is calculated by dividing the sum of the expected power generation revenue of all tasks within the work order by the estimated total operation and maintenance cost of the work order.

[0056] Preferably, the dynamic optimization and adjustment of composite maintenance work orders in the execution state includes the following steps:

[0057] When the maintenance team performs on-site operations based on the issued composite maintenance work orders, they collect information on newly discovered fault points that were not included in the original work orders during the operation process. The fault point information includes at least: fault type, identifier of fault component or string, and geographical location information obtained through the terminal.

[0058] Upon receiving the fault point information, the newly added fault point is treated as an independent maintenance requirement point, and the economic priority evaluation value of the maintenance requirement point is calculated.

[0059] Based on the following quantifiable rules, the system automatically determines whether to dynamically insert the newly added fault point into the currently executing composite maintenance work order:

[0060] Economic threshold rule: The economic priority evaluation value of the newly added point must be greater than or equal to a preset dynamic insertion economic threshold.

[0061] Resource real-time matching rules: All material types required to process the new point must exist in the material list carried by the execution team according to the composite operation and maintenance work order;

[0062] Path and Time Tolerance Rules: The current location reported by the execution team in real time is taken as the new path starting point. The geographical location of the newly added point is taken as the new mandatory path point. The geographical locations of all unfinished sub-tasks in the composite maintenance work order are compared with those of the new path. The Traveling Salesman Problem algorithm is called again to plan the work path. The total time of the newly planned path is calculated and compared with the estimated time of the original remaining tasks. If the time increment does not exceed the preset maximum allowable time increment ratio, the path disturbance is deemed acceptable.

[0063] A dynamic optimization instruction is generated only if a newly added fault point simultaneously meets all of the above rules. The instruction includes: detailed information about the newly added point, its economic priority evaluation value, and the recommended job sequence after replanning.

[0064] An intelligent operation and maintenance scheduling system for distributed photovoltaic power plants includes:

[0065] The task generation module includes:

[0066] The data acquisition unit is used to collect operational and environmental data from multiple distributed photovoltaic power stations. The operational data includes string-level current, voltage, and power data, as well as inverter status and alarm information. The environmental data includes irradiance, ambient temperature, and weather forecast information of the power station location.

[0067] The diagnostic identification unit is used to diagnose and identify the operation and maintenance needs of each power station based on the operational data and environmental data.

[0068] The task item generation unit is used to generate a preliminary operation and maintenance task item for each of the operation and maintenance demand points. The preliminary operation and maintenance task item includes at least the fault type, geographical location information, estimated processing time, recommended processing time window, and an economic priority evaluation value. The economic priority evaluation value is calculated based on the ratio of the expected power generation revenue generated from processing the operation and maintenance demand point to the estimated operation and maintenance cost.

[0069] The task aggregation module includes:

[0070] The cluster aggregation unit is used to aggregate multiple preliminary operation and maintenance task items that meet the preset geographical proximity conditions to form a regional task cluster.

[0071] The time window determination unit is used to determine a suitable maintenance time window for outdoor operations based on the weather forecast information and the power grid dispatch plan.

[0072] The work order generation unit is used to integrate and plan the work paths of multiple preliminary operation and maintenance tasks that belong to the same regional task cluster and whose recommended processing time window overlaps with the maintenance time window, and generate a composite operation and maintenance work order. The composite operation and maintenance work order has an overall economic evaluation value, which is higher than the economic priority evaluation value of any single preliminary operation and maintenance task item it contains.

[0073] The dynamic scheduling and execution module includes:

[0074] The condition configuration unit is used to place the composite operation and maintenance work order into the task queue to be executed, and configure trigger execution conditions for it, including at least the readiness status of associated operation and maintenance resources, no negative external event interference, and overall economic performance, and continuously monitor the trigger execution conditions.

[0075] The work order activation unit is used to activate the composite operation and maintenance work order from the queue of tasks to be executed when all triggering execution conditions are met, and send it to the corresponding operation and maintenance terminal for scheduling and execution.

[0076] The dynamic adjustment unit is used to dynamically optimize and adjust composite operation and maintenance work orders that are in the execution state.

[0077] The present invention has the following advantages:

[0078] 1. This invention overcomes the technical challenges of inaccurate fault location and difficulty in monitoring hidden degradation in distributed photovoltaic power stations by establishing a collaborative diagnostic mechanism that integrates string-level IV curve analysis and UAV image recognition. This method accurately correlates and cross-validates electrical characteristics and visual features in the spatiotemporal dimension, enabling precise identification of the smallest fault unit and clear distinction of fault types. At the same time, it effectively removes environmental noise and achieves early warning of slow degradation of component performance, thereby upgrading the operation and maintenance mode from overall extensive monitoring to component-level refined management.

[0079] 2. This invention fundamentally changes the traditional passive and fragmented operation and maintenance scheduling mode by designing a task aggregation algorithm with economic evaluation as the core and a dynamic condition-triggered execution strategy. The system intelligently aggregates scattered operation and maintenance needs into composite work orders with economies of scale, and automatically activates execution only at the optimal moment when resource, environmental and economic conditions are simultaneously met. This achieves global optimization of operation and maintenance paths and maximizes resource utilization, significantly improving the comprehensive economic benefits of a single operation and maintenance task. Attached Figure Description

[0080] Figure 1 This is a schematic diagram of the intelligent operation and maintenance scheduling system for distributed photovoltaic power plants used in an embodiment of the present invention. Detailed Implementation

[0081] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this invention.

[0082] Example 1: An intelligent operation and maintenance scheduling method for distributed photovoltaic power plants, comprising:

[0083] Step S100: Collect operational and environmental data from multiple distributed photovoltaic power stations; wherein, the operational data includes string-level current, voltage, and power data, as well as inverter status and alarm information, and the environmental data includes irradiance, ambient temperature, and weather forecast information of the power station location;

[0084] In practice, operational data is uploaded to the cloud platform in real time through smart sensors and communication modules installed in each power station; string-level data enables monitoring granularity down to each power generation unit, which is the basis for accurate diagnosis; environmental data can be obtained from the meteorological service interface for subsequent analysis.

[0085] Based on the aforementioned operational and environmental data, the operation and maintenance needs of each power station are diagnosed and identified.

[0086] For example, the system analysis revealed that the real-time power curve of string 3 in power plant A was significantly lower than that of other strings in the same power plant under the same irradiation conditions, and its IV curve shape was abnormal. Combined with the historical performance data of this string, the system can automatically diagnose and identify a maintenance requirement: string 3 in power plant A has a suspected fault, which leads to a decrease in power generation performance.

[0087] For each of the aforementioned maintenance needs, a preliminary maintenance task item is generated. This preliminary maintenance task item includes at least the fault type, geographical location information, estimated processing time, recommended processing time window, and an economic priority evaluation value. The economic priority evaluation value is calculated based on the ratio of the expected power generation revenue generated from processing the maintenance need to the estimated maintenance cost. Continuing the previous example, the system generates a task item for this need: the fault type is labeled "string performance anomaly"; the geographical location is the specific coordinates of power station A; the estimated processing time is set to 2 hours based on an experience database; the recommended processing time window is set to 9:00-11:00 the next day based on a future clear weather forecast; and the economic priority evaluation value is calculated by comparing the predicted additional power generation value of the string after repair with the estimated costs of travel, labor, etc., thereby quantifying the urgency and economic efficiency of the task.

[0088] Step S200: Aggregate multiple preliminary operation and maintenance tasks that meet the preset geographical proximity conditions to form a regional task cluster. In practical applications, the system will periodically scan all pending preliminary operation and maintenance tasks. For example, it may find that five power stations from different factory rooftops in the same industrial park have generated a total of eight tasks, all within 3 kilometers of each other. The system will then automatically aggregate these eight tasks to form a regional task cluster named "XX Industrial Park", laying the foundation for centralized processing.

[0089] Based on the weather forecast information and power grid dispatch plan, a suitable maintenance time window for outdoor operations is determined. For example, the system queries the weather forecast and confirms that the weather in the industrial park will be sunny and the wind force will be less than level 3 the next day. At the same time, the power grid plan is checked and it is confirmed that there is no temporary maintenance outage arrangement. Accordingly, the system determines that "8:00-16:00 the next day" is a suitable maintenance time window for outdoor operations in this area.

[0090] Multiple preliminary operation and maintenance tasks belonging to the same regional task cluster and whose recommended processing time window overlaps with the maintenance time window are integrated and their work paths are planned to generate a composite operation and maintenance work order. The composite operation and maintenance work order has an overall economic evaluation value, which is higher than the economic priority evaluation value of any single preliminary operation and maintenance task item it contains.

[0091] The system selects six tasks recommended for processing the next day from the "XX Industrial Park" cluster. Then, starting from the location of the maintenance station and using the latitude and longitude of the six task points as waypoints, the system calls a path planning algorithm to calculate an access sequence with the shortest total travel time. Finally, the system generates a composite maintenance work order that includes the six specific tasks and the optimized route.

[0092] Step S300: Place the composite operation and maintenance work order into the task queue to be executed, and configure it with trigger execution conditions including at least the readiness status of associated operation and maintenance resources, no negative external event interference, and overall economic performance.

[0093] The triggering conditions are continuously monitored. When all conditions are met, the composite maintenance work order is activated from the queue of tasks to be executed and sent to the corresponding maintenance terminal for scheduling and execution.

[0094] For example, the next day, the system's real-time monitoring found that after completing the morning tasks, the GPS location of maintenance team B showed that it was 5 kilometers away from the park and its status changed to "idle"; the actual weather conditions in the park were consistent with the forecast; all the triggering conditions were met at the same time, and the system immediately activated the composite maintenance work order and pushed it to team B's smart terminal; team B could efficiently complete this clustered maintenance operation by following the work order navigation and guidance.

[0095] The operational data also includes photovoltaic module image data collected by image sensors deployed on maintenance drones or fixed monitoring points; this provides visualized appearance status information for diagnosis; in actual deployment, drones can be automatically scheduled to fly over the power station periodically, or high-definition cameras installed on inspection robots or fixed towers can be used to collect visible light and infrared thermal images of photovoltaic modules.

[0096] Dynamically optimize and adjust composite operation and maintenance work orders that are in the execution state.

[0097] The process of diagnosing and identifying the operation and maintenance needs of each power station based on operational and environmental data includes the following sub-steps:

[0098] Step S101: Based on the single-diode equivalent circuit model of the photovoltaic module, the core physical parameters of the model include photocurrent, diode saturation current, series resistance, and parallel resistance; using the real-time irradiance and real-time ambient temperature in the environmental data, the nominal values ​​of the core physical parameters are corrected; the corrected parameters are substituted into the current-voltage equation of the single-diode equivalent circuit model, and the theoretical IV curve of the photovoltaic string under standard healthy conditions is generated through numerical calculation;

[0099] This step establishes the baseline for electrical diagnostics. The single-diode equivalent circuit is a classic physical model in the photovoltaic field, and its current-voltage equation can accurately describe the characteristics of the PN junction. In the model, the photocurrent is directly proportional to the irradiance, and parameters such as the diode saturation current are exponentially related to the temperature. During implementation, the system pre-stores the nominal parameters of each type of component. At the diagnostic moment, the system acquires the current real-time irradiance and ambient temperature, and performs mathematical corrections on the above nominal parameters based on recognized physical relationships, thereby obtaining a set of "ideal" parameters under the current specific environmental conditions. Subsequently, this set of parameters is substituted into the model equation, and a series of discrete current-voltage points from 0V to the open-circuit voltage are solved using numerical calculation methods such as the Newton-Raphson method, thus forming a "theoretical IV curve" under the current environment. This curve represents the expected performance of the string under perfect performance without any faults.

[0100] Step S102: Based on the collected string current and voltage data, plot the actual IV curve and extract the set of characteristic parameters from the actual IV curve. The set of characteristic parameters includes short-circuit current, open-circuit voltage, maximum power point current and maximum power point voltage.

[0101] Step S103: Compare the values ​​of each parameter in the set of feature parameters extracted from the actual IV curve with the corresponding theoretical feature parameters in the theoretical IV curve; when the relative deviation between the measured value and the theoretical value of any feature parameter exceeds the first threshold preset for that parameter, it is determined that the photovoltaic string has a fault or performance abnormality, and a first diagnostic result is generated, which includes the abnormal string identifier and the abnormality type.

[0102] Step S104: Input the photovoltaic module image data into a pre-trained image recognition model. The model outputs a state classification result for each photovoltaic module in the image, including normal, dust accumulation, shadow occlusion, or physical damage. A second diagnostic result is generated, which includes abnormal component identification and state classification.

[0103] This step is the visual diagnostic process. The pre-trained image recognition model is preferably a convolutional neural network model. This model typically includes convolutional layers, pooling layers, and fully connected layers. Its input is a pre-processed component image, and its output is the probability of each image patch belonging to each state category. The training process of the model is as follows: collect and manually label a large number of historical component image samples to construct a training dataset; use this dataset to perform supervised training on the CNN model, and optimize the network weights through the backpropagation algorithm until the model can accurately distinguish between typical appearance states such as "normal", "dust accumulation", "shadow occlusion", and "physical damage"; when applied, the system calls the model to automatically analyze the images taken by the drone, output the state label and confidence level of each component, and thus generate a second diagnostic result identified by the specific component location.

[0104] Step S105: For the same spatial location, perform spatiotemporal alignment and logical association between the first diagnostic result and the second diagnostic result;

[0105] When only the first diagnostic result exists, the generated maintenance requirement point is an electrical performance anomaly based on the IV curve;

[0106] When only a second diagnostic result exists, the generated maintenance requirement point is an abnormal appearance state based on the image;

[0107] When the first diagnostic result and the second diagnostic result both point to the same component or string, the generated maintenance requirement point is a composite anomaly, and the status classification result is used as a refined description of the fault type.

[0108] In step S200, multiple preliminary operation and maintenance task items that meet the preset geographical proximity conditions are aggregated to form a regional task cluster. Specific operations include:

[0109] Step S201: Convert the geographic location information contained in each preliminary operation and maintenance task item into unified latitude and longitude coordinates;

[0110] Step S202: Based on the latitude and longitude coordinates of all preliminary maintenance tasks to be processed, a density-based spatial clustering algorithm is used to group them. The core parameters of the density-based spatial clustering algorithm include the neighborhood search radius and the minimum task number threshold. When the straight-line distance between any two tasks in a group of preliminary maintenance tasks is not greater than the neighborhood search radius, and the total number of tasks in the group is not less than the minimum task number threshold, the group of tasks is clustered into a regional task cluster.

[0111] This step is the core of intelligent aggregation, and its key lies in the use of a "density-based spatial clustering algorithm", preferably the DBSCAN algorithm. This algorithm is particularly suitable for scenarios such as distributed photovoltaic power stations that are geographically irregularly scattered, because it can discover clusters of arbitrary shapes and effectively identify and filter out discrete, sparse outliers.

[0112] Step S203: Calculate a location coordinate representing the geographical center of each formed regional task cluster for subsequent job path planning.

[0113] In step S200, a composite maintenance work order is generated, which specifically includes the following sub-steps:

[0114] Step S204: For any regional task cluster, select all preliminary operation and maintenance tasks whose recommended processing time window overlaps with the determined maintenance time window; integrate these tasks and their included fault types and estimated processing time information to form a set of tasks to be planned.

[0115] This step involves time-based feasibility filtering and information integration for cluster tasks. For example, the system has formed a regional cluster containing 8 tasks and determined a maintenance time window of "tomorrow (Wednesday) 9:00-17:00". The system will verify the "recommended processing time window" of each task in the cluster one by one: if a task is recommended to be processed on Tuesday or Wednesday, it overlaps with the maintenance window and is selected; if another task is recommended to be processed only on Thursday, it is temporarily excluded. Finally, all selected tasks that can be executed collaboratively in terms of time are integrated into a set of tasks to be planned, providing a clear input target for the subsequent spatial path optimization.

[0116] Step S205: Using the representative coordinates of the regional task cluster as the starting and ending points of the path, the specific geographical location of each task item in the set of tasks to be planned as the necessary path points, and minimizing the sum of the total estimated processing time and the total path travel time of all task items as the optimization objective, the Traveling Salesman Problem algorithm is used to calculate and generate the optimal job sequence and the corresponding estimated total working time.

[0117] This step is the core of joint spatial and temporal optimization. Its key technology lies in modeling and solving the operation and maintenance path planning problem as a classic "traveling salesman problem".

[0118] Problem Modeling: The Traveling Salesman Problem (TSP) is a combinatorial optimization problem, classically described as follows: given a series of cities and the distance between each pair of cities, find the shortest loop that visits each city once and returns to the starting city. In this invention, the problem is specifically adapted as follows: the "departure / return base of maintenance vehicles" and "each power station in the set of tasks to be planned" are mapped to the "starting point / end point" and "must-pass cities" in the TSP, respectively; the "distance" is expanded to include a comprehensive "time cost" that includes travel time and on-site processing time.

[0119] Optimization objective: The objective function is to minimize the "total working time", which consists of two parts: 1) Total path travel time: estimated from the total path length and average vehicle speed; 2) Total estimated processing time: the sum of the estimated processing times of all task items in the set; This is a fixed value that does not affect the path sorting, but it does affect the calculation of total working time.

[0120] Algorithm Solution: In this embodiment, a genetic algorithm is preferably used. The system uses the genetic algorithm to intelligently search and iteratively optimize the access order of task points, and finally outputs an "optimal job sequence" that minimizes the total working time (e.g., base -> task point C -> task point A -> task point B -> return to base), and calculates the "estimated total working time" corresponding to the sequence.

[0121] Step S206: Based on the set of tasks to be planned, the optimal work sequence, and the estimated total working hours, generate the composite operation and maintenance work order; wherein, the overall economic evaluation value of the composite operation and maintenance work order is calculated by dividing the sum of the expected power generation revenue of all task items in the set by the total operation and maintenance cost calculated based on the estimated total working hours and the unit working hour cost.

[0122] Configuring trigger conditions for the composite maintenance work order and monitoring its execution includes the following sub-steps:

[0123] Step S301: Configure a set of quantifiable trigger execution conditions for each composite maintenance work order placed in the task queue. These conditions include:

[0124] Resource readiness conditions: The distance between the real-time geographical location of at least one operations and maintenance team and the representative coordinates of the task cluster in the region to which the composite operations and maintenance work order belongs is less than a preset distance threshold, and the team's current status is "idle" or "about to complete the task".

[0125] No negative external event conditions: Based on weather forecast information, confirm that there are no forecasts of precipitation or wind levels below the safe operation threshold within the recommended processing time window of the composite operation and maintenance work order; and confirm that the power grid dispatch system has not issued any power rationing instructions that would affect the shutdown of power plants in this area;

[0126] Overall economic efficiency criteria: The overall economic efficiency evaluation value of the composite maintenance work order is not lower than a preset economic threshold.

[0127] Step S302: Continuously monitor the triggering execution conditions of each composite maintenance work order in the queue of tasks to be executed; if and only if all the triggering execution conditions of a composite maintenance work order are met at the same time, the system automatically transitions its status from "pending triggering" to "executable";

[0128] Step S303: Assign composite maintenance work orders whose status transitions to "executable" to maintenance teams that meet the resource readiness conditions. The assignment is based on the comprehensive dispatch score of each candidate team, and the formula for calculating the comprehensive dispatch score Z is: Z = α*(1 / D) + β*S

[0129] Where D is the real-time path distance between the team's real-time geographical location and the representative coordinates of the regional task cluster, S is the percentage of skill items in the team's skill library that match the fault types included in the composite maintenance work order, and α and β are preset weighting coefficients.

[0130] The composite maintenance work order is assigned to the maintenance team with the highest comprehensive dispatch score Z, and the detailed operation instructions of the work order are sent to the maintenance terminal of the corresponding team.

[0131] The calculation of the economic priority evaluation value of the preliminary maintenance task item and the overall economic evaluation value of the composite maintenance work order are both achieved through the following unified benefit and cost calculation steps:

[0132] Calculation of expected power generation revenue:

[0133] For any point of operation and maintenance requirement or a set of tasks consisting of it, the expected power generation revenue is determined through the following steps:

[0134] Step 1: Combine the fault type corresponding to the maintenance requirement point, the historical power generation data sequence of the fault point, and the irradiance prediction sequence within the future preset time period to form a multi-dimensional feature vector;

[0135] Step 2: Input the multi-dimensional feature vector into a pre-trained power generation recovery prediction model. The power generation recovery prediction model is a neural network model based on a long short-term memory network architecture, which is trained using historical fault samples and is used to map the nonlinear relationship between fault features and recoverable power generation. The core of the model is a multi-layer LSTM neural network, whose hidden layer structure is designed to capture complex long-term time dependencies and nonlinear mappings in the input features, such as the continuous impact of historical performance degradation trends on future power generation capacity, and the power generation loss features under the interaction of different fault types and irradiation conditions. The network is finally connected to a fully connected output layer, which outputs a single scalar value, namely the total power generation that can be recovered within the preset future time period after the fault is repaired, as predicted by the model.

[0136] Step 3: The power generation recovery prediction model outputs a value, which is the expected power generation that can be recovered in the future within a preset period after the demand is processed; multiply the expected recoverable power generation by the preset electricity price, and the result is the expected power generation revenue;

[0137] Calculation of estimated operation and maintenance costs:

[0138] For a single initial maintenance task, the estimated maintenance cost consists of labor cost and material cost. Specifically, the estimated processing time is multiplied by the preset unit labor cost, and the material cost matched from the standard material library according to the fault type is added.

[0139] For a composite maintenance work order, its estimated total maintenance cost consists of total working hours cost, total material cost and mileage cost. Specifically, it is as follows: multiply the estimated total working hours generated by the work path planning in step S200 by the comprehensive cost per unit working hour, add the sum of the material costs matched by all task items in the work order, and add the product of the total mileage determined based on the work path planning and the preset cost per unit mileage.

[0140] Calculation of economic priority evaluation value:

[0141] The economic priority evaluation value of a single initial operation and maintenance task item is calculated by dividing the expected power generation revenue of the task item by its estimated operation and maintenance cost;

[0142] The overall economic evaluation value of a composite operation and maintenance work order is calculated by dividing the sum of the expected power generation revenue of all tasks within the work order by the estimated total operation and maintenance cost of the work order.

[0143] Dynamically optimize and adjust composite maintenance work orders that are in the execution state, specifically including the following steps:

[0144] When the maintenance team performs on-site operations based on the issued composite maintenance work orders, they collect information on newly discovered fault points that were not included in the original work orders during the operation process. The fault point information includes at least: fault type, identifier of fault component or string, and geographical location information obtained through the terminal.

[0145] Upon receiving the fault point information, the newly added fault point is treated as an independent maintenance requirement point, and the economic priority evaluation value of the maintenance requirement point is calculated.

[0146] Based on the following quantifiable rules, the system automatically determines whether to dynamically insert the newly added fault point into the currently executing composite maintenance work order:

[0147] Economic threshold rule: The economic priority evaluation value of the newly added point must be greater than or equal to a preset dynamic insertion economic threshold.

[0148] Resource real-time matching rules: All material types required to process the new point must exist in the material list carried by the execution team according to the composite operation and maintenance work order;

[0149] Path and Time Tolerance Rules: The current location reported by the execution team in real time is taken as the new path starting point. The geographical location of the newly added point is taken as the new mandatory path point. The geographical locations of all unfinished sub-tasks in the composite maintenance work order are compared with those of the new path. The Traveling Salesman Problem algorithm is called again to plan the work path. The total time of the newly planned path is calculated and compared with the estimated time of the original remaining tasks. If the time increment does not exceed the preset maximum allowable time increment ratio, the path disturbance is deemed acceptable.

[0150] A dynamic optimization instruction is generated only if a newly added fault point simultaneously meets all of the above rules. The instruction includes: detailed information about the newly added point, its economic priority evaluation value, and the recommended job sequence after replanning.

[0151] Example 2: An intelligent operation and maintenance scheduling system for distributed photovoltaic power plants, such as... Figure 1 As shown, it includes:

[0152] The task generation module includes:

[0153] The data acquisition unit is used to collect operational and environmental data from multiple distributed photovoltaic power stations. The operational data includes string-level current, voltage, and power data, as well as inverter status and alarm information. The environmental data includes irradiance, ambient temperature, and weather forecast information of the power station location.

[0154] The diagnostic identification unit is used to diagnose and identify the operation and maintenance needs of each power station based on the operational data and environmental data.

[0155] The task item generation unit is used to generate a preliminary operation and maintenance task item for each of the operation and maintenance demand points. The preliminary operation and maintenance task item includes at least the fault type, geographical location information, estimated processing time, recommended processing time window, and an economic priority evaluation value. The economic priority evaluation value is calculated based on the ratio of the expected power generation revenue generated from processing the operation and maintenance demand point to the estimated operation and maintenance cost.

[0156] The task aggregation module includes:

[0157] The cluster aggregation unit is used to aggregate multiple preliminary operation and maintenance task items that meet the preset geographical proximity conditions to form a regional task cluster.

[0158] The time window determination unit is used to determine a suitable maintenance time window for outdoor operations based on the weather forecast information and the power grid dispatch plan.

[0159] The work order generation unit is used to integrate and plan the work paths of multiple preliminary operation and maintenance tasks that belong to the same regional task cluster and whose recommended processing time window overlaps with the maintenance time window, and generate a composite operation and maintenance work order. The composite operation and maintenance work order has an overall economic evaluation value, which is higher than the economic priority evaluation value of any single preliminary operation and maintenance task item it contains.

[0160] The dynamic scheduling and execution module includes:

[0161] The condition configuration unit is used to place the composite operation and maintenance work order into the task queue to be executed, and configure trigger execution conditions for it, including at least the readiness status of associated operation and maintenance resources, no negative external event interference, and overall economic performance, and continuously monitor the trigger execution conditions.

[0162] The work order activation unit is used to activate the composite operation and maintenance work order from the queue of tasks to be executed when all triggering execution conditions are met, and send it to the corresponding operation and maintenance terminal for scheduling and execution.

[0163] The dynamic adjustment unit is used to dynamically optimize and adjust composite operation and maintenance work orders that are in the execution state.

[0164] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A smart operation and maintenance scheduling method for distributed photovoltaic power plants, characterized in that, include: Step S100: Collect operational and environmental data from multiple distributed photovoltaic power stations; wherein, the operational data includes string-level current, voltage, and power data, as well as inverter status and alarm information, and the environmental data includes irradiance, ambient temperature, and weather forecast information of the power station location; Based on the aforementioned operational and environmental data, the operation and maintenance needs of each power station are diagnosed and identified. For each of the aforementioned maintenance requirements, a preliminary maintenance task item is generated. The preliminary maintenance task item includes at least the fault type, geographical location information, estimated processing time, recommended processing time window, and an economic priority evaluation value. The economic priority evaluation value is calculated based on the ratio of the expected power generation revenue generated from processing the maintenance requirement to the estimated maintenance cost. Step S200: Aggregate multiple preliminary operation and maintenance task items that meet the preset geographical proximity conditions to form a regional task cluster; Based on the weather forecast information and power grid dispatch plan, a suitable maintenance time window for outdoor operations is determined; Multiple preliminary operation and maintenance tasks belonging to the same regional task cluster and whose recommended processing time window overlaps with the maintenance time window are integrated and their work paths are planned to generate a composite operation and maintenance work order. The composite operation and maintenance work order has an overall economic evaluation value, which is higher than the economic priority evaluation value of any single preliminary operation and maintenance task item it contains. The overall economic evaluation value of a composite operation and maintenance work order is calculated by dividing the sum of the expected power generation revenue of all tasks in the work order by the estimated total operation and maintenance cost of the work order. Step S300: Place the composite operation and maintenance work order into the task queue to be executed, and configure it with trigger execution conditions including at least the readiness status of associated operation and maintenance resources, no negative external event interference, and overall economic performance. The triggering conditions are continuously monitored. When all conditions are met, the composite maintenance work order is activated from the queue of tasks to be executed and sent to the corresponding maintenance terminal for scheduling and execution. Dynamically optimize and adjust composite operation and maintenance work orders that are in the execution state.

2. The intelligent operation and maintenance scheduling method for distributed photovoltaic power stations according to claim 1, characterized in that, The operational data also includes photovoltaic module image data collected by image sensors deployed on maintenance drones or fixed monitoring points; Based on the aforementioned operational and environmental data, the operation and maintenance needs of each power station are diagnosed and identified, specifically including the following sub-steps: Step S101: Based on the single diode equivalent circuit model of the photovoltaic module, the core physical parameters of the model include photocurrent, diode saturation current, series resistance and parallel resistance; using the real-time irradiance and real-time ambient temperature in the environmental data, the nominal values ​​of the core physical parameters are corrected; the corrected parameters are substituted into the current-voltage equation of the single diode equivalent circuit model, and the theoretical IV curve of the photovoltaic string under standard healthy conditions is generated through numerical calculation. Step S102: Based on the collected string current and voltage data, plot the actual IV curve and extract the set of characteristic parameters from the actual IV curve. The set of characteristic parameters includes short-circuit current, open-circuit voltage, maximum power point current and maximum power point voltage. Step S103: Compare the values ​​of each parameter in the set of feature parameters extracted from the actual IV curve with the corresponding theoretical feature parameters in the theoretical IV curve; when the relative deviation between the measured value and the theoretical value of any feature parameter exceeds the first threshold preset for that parameter, it is determined that the photovoltaic string has a fault or performance abnormality, and a first diagnostic result is generated, which includes the abnormal string identifier and the abnormality type. Step S104: Input the photovoltaic module image data into a pre-trained image recognition model. The model outputs a state classification result for each photovoltaic module in the image, including normal, dust accumulation, shadow occlusion, or physical damage. A second diagnostic result is generated, which includes abnormal component identification and state classification. Step S105: For the same spatial location, perform spatiotemporal alignment and logical association between the first diagnostic result and the second diagnostic result; When only the first diagnostic result exists, the generated maintenance requirement point is an electrical performance anomaly based on the IV curve; When only a second diagnostic result exists, the generated maintenance requirement point is an abnormal appearance state based on the image; When the first diagnostic result and the second diagnostic result both point to the same component or string, the generated maintenance requirement point is a composite anomaly, and the status classification result is used as a refined description of the fault type.

3. The intelligent operation and maintenance scheduling method for distributed photovoltaic power stations according to claim 2, characterized in that, In step S200, multiple preliminary operation and maintenance task items that meet the preset geographical proximity conditions are aggregated to form a regional task cluster. Specific operations include: Step S201: Convert the geographic location information contained in each preliminary operation and maintenance task item into unified latitude and longitude coordinates; Step S202: Based on the latitude and longitude coordinates of all preliminary maintenance tasks to be processed, a density-based spatial clustering algorithm is used to group them. The core parameters of the density-based spatial clustering algorithm include the neighborhood search radius and the minimum task number threshold. When the straight-line distance between any two tasks in a group of preliminary maintenance tasks is not greater than the neighborhood search radius, and the total number of tasks in the group is not less than the minimum task number threshold, the group of tasks is clustered into a regional task cluster. Step S203: Calculate a location coordinate representing the geographical center of each formed regional task cluster for subsequent job path planning.

4. The intelligent operation and maintenance scheduling method for distributed photovoltaic power stations according to claim 3, characterized in that, In step S200, a composite maintenance work order is generated, which specifically includes the following sub-steps: Step S204: For any regional task cluster, select all preliminary operation and maintenance tasks whose recommended processing time window overlaps with the determined maintenance time window; integrate these tasks and their included fault types and estimated processing time information to form a set of tasks to be planned. Step S205: Using the representative coordinates of the regional task cluster as the starting and ending points of the path, the specific geographical location of each task item in the set of tasks to be planned as the necessary path points, and minimizing the sum of the total estimated processing time and the total path travel time of all task items as the optimization objective, the Traveling Salesman Problem algorithm is used to calculate and generate the optimal job sequence and the corresponding estimated total working time. Step S206: Based on the set of tasks to be planned, the optimal work sequence, and the estimated total working hours, generate the composite operation and maintenance work order; wherein, the overall economic evaluation value of the composite operation and maintenance work order is calculated by dividing the sum of the expected power generation revenue of all task items in the set by the total operation and maintenance cost calculated based on the estimated total working hours and the unit working hour cost.

5. The intelligent operation and maintenance scheduling method for distributed photovoltaic power stations according to claim 4, characterized in that, Configuring trigger conditions for the composite maintenance work order and monitoring its execution includes the following sub-steps: Step S301: Configure a set of quantifiable trigger execution conditions for each composite maintenance work order placed in the task queue. These conditions include: Resource readiness conditions: The distance between the real-time geographical location of at least one operations and maintenance team and the representative coordinates of the task cluster in the region to which the composite operations and maintenance work order belongs is less than a preset distance threshold, and the team's current status is "idle" or "about to complete the task". No negative external event conditions: Based on weather forecast information, confirm that there are no forecasts of precipitation or wind levels below the safe operation threshold within the recommended processing time window of the composite operation and maintenance work order; and confirm that the power grid dispatch system has not issued any power rationing instructions that would affect the shutdown of power plants in this area; Overall economic efficiency criteria: The overall economic efficiency evaluation value of the composite maintenance work order is not lower than a preset economic threshold. Step S302: Continuously monitor the triggering execution conditions of each composite maintenance work order in the queue of tasks to be executed; if and only if all the triggering execution conditions of a composite maintenance work order are met at the same time, the system automatically transitions its status from "pending triggering" to "executable"; Step S303: Assign the composite operation and maintenance work order that transitions to "executable" to the operation and maintenance team that meets the resource ready state condition. The assignment is based on the calculation of the comprehensive dispatch score of each candidate team. The formula for calculating the comprehensive dispatch score Z is: Z=α*(1 / D)+β*S. Where D is the real-time path distance between the team's real-time geographical location and the representative coordinates of the regional task cluster, S is the percentage of skill items in the team's skill library that match the fault types included in the composite maintenance work order, and α and β are preset weighting coefficients. The composite maintenance work order is assigned to the maintenance team with the highest comprehensive dispatch score Z, and the detailed operation instructions of the work order are sent to the maintenance terminal of the corresponding team.

6. The intelligent operation and maintenance scheduling method for distributed photovoltaic power stations according to claim 5, characterized in that, The calculation of the economic priority evaluation value of the preliminary maintenance task item and the overall economic evaluation value of the composite maintenance work order are both achieved through the following unified benefit and cost calculation steps: Calculation of expected power generation revenue: For any point of operation and maintenance requirement or a set of tasks consisting of it, the expected power generation revenue is determined through the following steps: Step 1: Combine the fault type corresponding to the maintenance requirement point, the historical power generation data sequence of the maintenance requirement point, and the irradiance prediction sequence within the future preset time period to form a multi-dimensional feature vector; Step 2: Input the multi-dimensional feature vector into a pre-trained power generation recovery prediction model; the power generation recovery prediction model is a neural network model based on a long short-term memory network architecture, which is trained through historical fault samples and is used to map the nonlinear relationship between fault features and recoverable power generation. Step 3: The power generation recovery prediction model outputs a value, which is the expected power generation that can be recovered in the future within a preset period after the demand is processed; multiply the expected recoverable power generation by the preset electricity price, and the result is the expected power generation revenue; Calculation of estimated operation and maintenance costs: For a single initial maintenance task, the estimated maintenance cost consists of labor cost and material cost. Specifically, the estimated processing time is multiplied by the preset unit labor cost, and the material cost matched from the standard material library according to the fault type is added. For a composite maintenance work order, its estimated total maintenance cost consists of total working hours cost, total material cost and mileage cost. Specifically, it is as follows: multiply the estimated total working hours generated by the work path planning in step S200 by the comprehensive cost per unit working hour, add the sum of the material costs matched by all task items in the work order, and add the product of the total mileage determined based on the work path planning and the preset cost per unit mileage. Calculation of economic priority evaluation value: The economic priority evaluation value of a single initial operation and maintenance task item is calculated by dividing the expected power generation revenue of the task item by its estimated operation and maintenance cost.

7. The intelligent operation and maintenance scheduling method for distributed photovoltaic power stations according to claim 6, characterized in that, Dynamically optimize and adjust composite maintenance work orders that are in the execution state, specifically including the following steps: When the maintenance team performs on-site operations based on the issued composite maintenance work orders, they collect information on newly discovered fault points that were not included in the original work orders during the operation process. The fault point information includes at least: fault type, identifier of fault component or string, and geographical location information obtained through the terminal. Upon receiving the fault point information, the newly added fault point is treated as an independent maintenance requirement point, and the economic priority evaluation value of the maintenance requirement point is calculated. Based on the following quantifiable rules, the system automatically determines whether to dynamically insert the newly added fault point into the currently executing composite maintenance work order: Economic threshold rule: The economic priority evaluation value of the newly added fault point must be greater than or equal to a preset dynamic insertion economic threshold. Resource real-time matching rules: All material types required to handle the newly added fault point must exist in the material list carried by the execution team according to the composite maintenance work order; Path and Time Tolerance Rules: The current location reported by the execution team in real time is taken as the starting point of the new path. The geographical location of the newly added fault point is taken as the new mandatory path point. The geographical locations of all unfinished sub-tasks in the composite maintenance work order are compared with those of the newly added fault point. The Traveling Salesman Problem algorithm is called again to plan the work path. The total time of the newly planned path is calculated and compared with the estimated time of the original remaining tasks. If the time increment does not exceed the preset maximum allowable time increment ratio, the path disturbance is deemed acceptable. A dynamic optimization instruction is generated only if a newly added fault point simultaneously meets all of the above rules. The instruction includes: detailed information about the newly added fault point, its economic priority evaluation value, and the recommended job sequence after replanning.

8. An intelligent operation and maintenance scheduling system for distributed photovoltaic power plants, characterized in that, The system is applied to the intelligent operation and maintenance scheduling method for distributed photovoltaic power plants according to any one of claims 1-7, comprising: The task generation module includes: The data acquisition unit is used to collect operational and environmental data from multiple distributed photovoltaic power stations. The operational data includes string-level current, voltage, and power data, as well as inverter status and alarm information. The environmental data includes irradiance, ambient temperature, and weather forecast information of the power station location. The diagnostic identification unit is used to diagnose and identify the operation and maintenance needs of each power station based on the operational data and environmental data. The task item generation unit is used to generate a preliminary operation and maintenance task item for each of the operation and maintenance demand points. The preliminary operation and maintenance task item includes at least the fault type, geographical location information, estimated processing time, recommended processing time window, and an economic priority evaluation value. The economic priority evaluation value is calculated based on the ratio of the expected power generation revenue generated from processing the operation and maintenance demand point to the estimated operation and maintenance cost. The task aggregation module includes: The cluster aggregation unit is used to aggregate multiple preliminary operation and maintenance task items that meet the preset geographical proximity conditions to form a regional task cluster. The time window determination unit is used to determine a suitable maintenance time window for outdoor operations based on the weather forecast information and the power grid dispatch plan. The work order generation unit is used to integrate and plan the work paths of multiple preliminary operation and maintenance tasks that belong to the same regional task cluster and whose recommended processing time window overlaps with the maintenance time window, and generate a composite operation and maintenance work order. The composite operation and maintenance work order has an overall economic evaluation value, which is higher than the economic priority evaluation value of any single preliminary operation and maintenance task item it contains. The dynamic scheduling and execution module includes: The condition configuration unit is used to place the composite operation and maintenance work order into the task queue to be executed, and configure trigger execution conditions for it, including at least the readiness status of associated operation and maintenance resources, no negative external event interference, and overall economic performance, and continuously monitor the trigger execution conditions. The work order activation unit is used to activate the composite operation and maintenance work order from the queue of tasks to be executed when all triggering execution conditions are met, and send it to the corresponding operation and maintenance terminal for scheduling and execution. The dynamic adjustment unit is used to dynamically optimize and adjust composite operation and maintenance work orders that are in the execution state.

Images