Photovoltaic cleaning robot operation scheduling method and system based on environment prediction

Abstract

The invention discloses a photovoltaic cleaning robot operation scheduling method and system based on environment prediction, and belongs to the technical field of photovoltaic power station intelligent operation and maintenance and robots, and the method comprises the steps: obtaining weather forecast data, photovoltaic panel real-time dust accumulation data and historical power generation efficiency attenuation data, and generating environment prediction parameters; generating a dynamic weight value based on the environment prediction parameters and a preset power generation loss threshold value, and judging the sweeping emergency degree of each photovoltaic panel area; distributing an operation priority according to the cleaning emergency degree, and generating an initial cleaning task sequence; and a multi-robot cooperative path planning result is generated by combining the photovoltaic power station topological structure and the real-time position data of the robots. According to the method, the environment prediction technology and the dynamic scheduling strategy are combined, the cleaning task can be adaptively adjusted according to weather changes and dust accumulation conditions, the robot operation path is optimized, and the power generation efficiency of a photovoltaic power station and the utilization rate of robot resources are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the fields of intelligent operation and maintenance of photovoltaic power stations and robotics technology, and in particular to an operation scheduling method and system for photovoltaic cleaning robots based on environmental prediction. Background Art

[0002] As an important form of clean energy, photovoltaic power generation's efficiency is directly affected by the surface cleanliness of photovoltaic panels. The accumulation of pollutants such as dust and bird droppings on the panels significantly reduces light absorption, thereby affecting overall power generation. Therefore, regular and effective cleaning of photovoltaic panels is crucial for ensuring the efficient operation of photovoltaic power plants. Automated cleaning robots, due to their efficiency and convenience, are increasingly being used in large-scale photovoltaic power plants.

[0003] Currently, scheduling operations for photovoltaic cleaning robots mostly relies on fixed periodic plans or simple threshold-triggered mechanisms. These scheduling methods often struggle to adapt to complex and changing lighting and weather conditions, as well as the dynamic variations in dust accumulation across photovoltaic panels. This can lead to untimely cleaning of some areas, resulting in continued power generation losses, or cleaning operations performed at unnecessary times, wasting energy and water resources. Overall scheduling efficiency and cost-effectiveness need to be improved. Summary of the Invention

[0004] To solve the above problems, the present invention provides a photovoltaic cleaning robot operation scheduling method and system based on environmental prediction. It combines environmental prediction technology with dynamic scheduling strategies, can adaptively adjust cleaning tasks according to weather changes and dust accumulation, optimize the robot's operation path, and significantly improve the power generation efficiency of the photovoltaic power station and the utilization rate of robot resources.

[0005] The above objectives can be achieved through the following solutions:

[0006] A photovoltaic cleaning robot operation scheduling method based on environmental prediction includes obtaining weather forecast data, real-time dust accumulation data of photovoltaic panels, and historical power generation efficiency attenuation data, and generating environmental prediction parameters based on the weather forecast data, the real-time dust accumulation data of photovoltaic panels, and the historical power generation efficiency attenuation data; generating dynamic weight values ​​based on the environmental prediction parameters and in combination with preset power generation loss thresholds, and judging the cleaning urgency of each photovoltaic panel area by means of the dynamic weight values; assigning an operation priority to each photovoltaic panel area according to the judged cleaning urgency, and generating an initial cleaning task sequence based on preset robot operation capability parameters; generating a multi-robot collaborative path planning result based on the initial cleaning task sequence and in combination with preset photovoltaic power station topology data and the obtained real-time robot position data.

[0007] Optionally, generating environmental prediction parameters includes: performing time series processing on the weather forecast data to extract environmental risk characteristic values ​​within a preset time period in the future, wherein the environmental risk characteristic values ​​include wind level characteristics, rainfall and snowfall probability characteristics, and air humidity fluctuation range; analyzing the correspondence between the amount of dust accumulation and the power generation decline rate in the historical power generation efficiency attenuation data to generate a power generation loss rate curve; correlating and matching the environmental risk characteristic values ​​with the power generation loss rate curve to generate environmental prediction parameters.

[0008] Optionally, the generation of dynamic weight values ​​includes: judging whether natural cleaning conditions are triggered based on the rainfall and snowfall probability characteristics in the environmental risk characteristic values, and if the probability indicated by the judged rainfall and snowfall probability characteristics exceeds a preset judgment threshold, reducing the cleaning urgency of the corresponding photovoltaic panel area; calculating the estimated power generation loss value corresponding to the current dust accumulation amount based on the power generation loss rate curve, and generating a power generation benefit value in combination with the wind level characteristics in the environmental prediction parameters; generating and dynamically adjusting the dynamic weight value based on the power generation benefit value and a preset energy benefit model.

[0009] Optionally, the assigning of operation priorities to each photovoltaic panel area includes: dividing each photovoltaic panel area into high-efficiency grids, medium-efficiency grids and low-efficiency grids according to the cleaning urgency, and matching and allocating the number of robots to each divided grid based on the endurance parameter in the preset robot operation capability parameter; establishing a priority task queue for the divided high-efficiency grid, and inserting the established priority task queue into the head of the initial cleaning task sequence to adjust the initial cleaning task sequence; dynamically adjusting the coverage method of the divided low-efficiency grid according to the acquired robot power data, and generating an updated cleaning task sequence based on the adjusted initial cleaning task sequence.

[0010] Optionally, generating a multi-robot collaborative path planning result includes: generating a corresponding initial path for each robot according to the updated cleaning task sequence; obtaining coordinate data and task execution status data uploaded by the robot in real time, and detecting whether there is a path conflict area based on the obtained coordinate data and task execution status data; if it is determined that there is a path conflict area, determining the coordinates of the path conflict area, and reallocating the robot passage order based on the dynamic weight value associated with the priority task queue to generate a multi-robot collaborative path planning result.

[0011] Optionally, the reallocation of the robot passage order includes: generating a priority sorting list according to the power generation efficiency value corresponding to the current task being performed by each robot; dividing the first priority and the second priority based on the priority sorting list, wherein the first priority is higher than the second priority, locking the path usage right of the path conflict area for the identified first-priority robot, and triggering the identified second-priority robot to perform a detour path calculation; if the estimated time for performing the detour path calculation exceeds a preset tolerance threshold, migrating the current task of the second-priority robot to the identified idle robot.

[0012] Optionally, generating the multi-robot collaborative path planning results also includes: obtaining the robot's historical movement trajectory data, extracting the path repetition rate characteristics and the energy consumption distribution characteristics; based on the path repetition rate characteristics and the energy consumption distribution characteristics, optimizing the distribution position of the turning points of each path in the multi-robot collaborative path planning results, and generating an energy-balanced path; updating the multi-robot collaborative path planning results according to the energy-balanced path.

[0013] Optionally, the method also includes: when an emergency environmental event is detected, interrupting low-priority tasks according to the cleaning urgency, and reallocating the robot to perform emergency cleaning tasks; based on the emergency cleaning tasks, generating a dynamically compressed spiral progressive path so that the robot completes a preset coverage rate within the impact period of the emergency environmental event.

[0014] Optionally, the method also includes: obtaining actual power generation efficiency data of the photovoltaic panels after the task is completed, and generating cleaning effect verification parameters based on the actual power generation efficiency data of the photovoltaic panels; comparing the cleaning effect verification parameters with a preset verification threshold, and if it is determined that the cleaning effect verification parameters deviate from the preset verification threshold, adjusting the association matching rules of the environmental prediction parameters.

[0015] Based on the same inventive concept, the present invention also provides a photovoltaic cleaning robot operation scheduling system based on environmental prediction, the system including: a data acquisition module, used to obtain weather forecast data, real-time dust accumulation data of photovoltaic panels and historical power generation efficiency attenuation data; a prediction model module, connected to the data acquisition module, used to generate environmental prediction parameters and calculate power generation benefit values; a scheduling generation module, connected to the prediction model module, used to assign job priorities according to the power generation benefit values ​​and generate multi-robot collaborative path planning results; a dynamic adjustment module, connected to the scheduling generation module, used to respond to sudden environmental events and update path planning; a verification feedback module, used to optimize the matching rules of the prediction model module according to the actual power generation efficiency data after the task is completed.

[0016] Compared with the prior art, the present invention has the following advantages:

[0017] 1. Improved the targeting and intelligence of cleaning operations: Through environmental prediction technology, weather forecasts, dust accumulation, and historical power generation efficiency data are integrated to dynamically adjust cleaning task priorities and robot scheduling. This overcomes the resource waste and inefficiency problems of the traditional fixed scheduling model, significantly improving the targeting and intelligence of operations.

[0018] 2. Optimized robot resource utilization and operating efficiency: The system adopts a multi-robot collaborative path planning and conflict resolution mechanism, combined with the photovoltaic power station topology and real-time location data, to optimize the operating path and resource allocation, improve cleaning efficiency, extend equipment life, and have high practical value and economic benefits;

[0019] 3. Enhanced system response to sudden environmental events: Through real-time monitoring and dynamic adjustment mechanisms, low-priority tasks are quickly interrupted when an emergency occurs, and robots are reallocated to perform emergency cleaning, ensuring that power generation efficiency is not affected, thereby improving the robustness and reliability of the system.

[0020] 4. Achieved closed-loop verification and continuous optimization of the cleaning effect: Verification parameters are generated through actual power generation efficiency data after the task, compared with preset thresholds, and the matching rules of environmental prediction parameters are dynamically adjusted to achieve closed-loop verification of the cleaning effect and continuous optimization of the system, thereby improving the stability and effectiveness of long-term operation.

[0021] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0023] Figure 1 The present invention is a flowchart of a method and system for scheduling photovoltaic cleaning robot operations based on environmental prediction.

[0024] Figure 2 1 is a graph showing a power generation loss rate curve and a dust accumulation level curve according to an embodiment of the present invention.

[0025] Figure 3This is a graph showing changes in urgency of cleaning of a photovoltaic panel area and responses to environmental events according to an embodiment of the present invention.

[0026] Figure 4 3. This is a comparison diagram of the emergency cleaning coverage path and the conventional path according to an embodiment of the present invention.

[0027] Figure 5 The present invention is a schematic diagram of a photovoltaic cleaning robot operation scheduling method and system based on environmental prediction according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0029] Reference Figure 1 One embodiment of the present invention proposes a photovoltaic cleaning robot operation scheduling method and system based on environmental prediction. By combining environmental prediction technology with a dynamic scheduling strategy, it can adaptively adjust cleaning tasks according to weather changes and dust accumulation, optimize the robot's operation path, and significantly improve the power generation efficiency of the photovoltaic power station and the utilization rate of robot resources.

[0030] The method of this embodiment specifically includes:

[0031] Acquiring weather forecast data, real-time dust accumulation data of photovoltaic panels, and historical power generation efficiency attenuation data, and generating environmental prediction parameters based on the weather forecast data, the real-time dust accumulation data of photovoltaic panels, and the historical power generation efficiency attenuation data;

[0032] Based on the environmental prediction parameters and in combination with a preset power generation loss threshold, a dynamic weight value is generated, and the cleaning urgency of each photovoltaic panel area is determined by the dynamic weight value;

[0033] Assigning a work priority to each photovoltaic panel area according to the determined cleaning urgency, and generating an initial cleaning task sequence based on preset robot operation capability parameters;

[0034] According to the initial cleaning task sequence, combined with the preset photovoltaic power station topology data and the acquired real-time position data of the robots, a multi-robot collaborative path planning result is generated.

[0035] By combining environmental prediction technology with dynamic scheduling strategies, cleaning tasks can be adaptively adjusted according to weather changes and dust accumulation, the robot's operating path can be optimized, and the power generation efficiency of the photovoltaic power station and the utilization rate of robot resources can be significantly improved.

[0036] Optionally, the generation environment prediction parameters include:

[0037] Performing time series processing on the weather forecast data to extract environmental risk characteristic values ​​within a preset time period in the future, wherein the environmental risk characteristic values ​​include wind force level characteristics, rainfall and snowfall probability characteristics, and air humidity fluctuation range;

[0038] Specifically, weather forecast data is obtained through the meteorological service interface. These data usually cover information such as wind speed, rainfall probability, snowfall probability and air humidity within a preset time period in the future. Time series processing is performed on these data to extract environmental risk characteristic values ​​that can quantify the impact of the environment on dust accumulation on photovoltaic panels. These characteristic values ​​include wind force level characteristics, rainfall and snowfall probability characteristics and air humidity fluctuation range. The wind force level characteristics divide the wind speed into levels 0-12 according to the Beaufort wind scale standard, which is used to evaluate the impact of wind on dust accumulation or dispersion. The rainfall and snowfall probability characteristics indicate the possibility of rainfall or snowfall within a preset time period in the future, with a value range of 0 to 1, which is used to measure the cleaning effect of natural precipitation on dust accumulation. The air humidity fluctuation range is obtained by calculating the difference between the maximum and minimum humidity values ​​within the preset time period, in percentage, and is used to analyze the impact of humidity on dust adhesion. In order to comprehensively evaluate environmental risks, a comprehensive environmental risk characteristic value is calculated. The formula is as follows:

[0039] R=w f *F+w p *P+w h *H,

[0040] Where R is the comprehensive environmental risk characteristic value, which represents the comprehensive quantitative result of environmental risk; F is the wind force level, ranging from 0 to 12; P is the probability of rain or snow, ranging from 0 to 1; H is the humidity fluctuation amplitude, unit %; w f 、w p 、w h are the weight coefficients of each feature.

[0041] Analyzing the corresponding relationship between the amount of dust accumulation and the power generation decline rate in the historical power generation efficiency decay data to generate a power generation loss rate curve;

[0042] Specifically, historical power generation efficiency decay data was used to analyze the relationship between dust accumulation and power generation decline rate. This historical data records the changes in power generation under different levels of dust accumulation. Dust accumulation is measured by sensors in grams per square meter, and the power generation decline rate represents the percentage of power generation efficiency loss caused by dust accumulation. By analyzing this data, a power generation loss rate curve was fitted to describe the impact of dust accumulation on power generation efficiency. A quadratic function model was used to fit this relationship, with the following formula:

[0043] L=a*D 2 +b*D+c,

[0044] Where L is the power generation efficiency loss caused by dust accumulation, in %; D is the amount of dust accumulated on the surface of the photovoltaic panel, in g / m 2 ; a, b, c are coefficients obtained by fitting historical data. Through this formula, the corresponding power generation decline rate can be calculated according to any dust accumulation value to form a power generation loss rate curve, such as Figure 2 shown.

[0045] The environmental risk characteristic value is correlated and matched with the power generation loss rate curve to generate environmental prediction parameters.

[0046] Specifically, the obtained comprehensive environmental risk characteristic value is combined with the power generation decline rate to generate an environmental prediction parameter through correlation matching. This parameter comprehensively reflects the potential impact of environmental conditions and dust accumulation on photovoltaic power generation efficiency, providing a basis for subsequent cleaning operation scheduling. The calculation formula is:

[0047] E=k r *R+k l *L,

[0048] Where, E is the environmental prediction parameter; R is the comprehensive environmental risk characteristic value; L is the power generation decline rate; k r 、k l are the weight coefficients of the two parts respectively.

[0049] For example, suppose the weather forecast data shows that the wind force level in the next 24 hours is level 4, the probability of rainfall is 20%, the air humidity fluctuation range is 10%, and the historical data shows that the current dust accumulation is 8g / m 2 First, calculate the comprehensive environmental risk characteristic value, which is 0.5×4+0.3×0.2+0.2×10, and the result is 2.06. Secondly, calculate the power generation decline rate, which is 0.02×8 2+0.03×8 + 0.01, the result is 1.53%. Finally, calculate the environmental prediction parameter, which is 0.6×2.06 + 0.4×1.53, and the result is 1.848. This environmental prediction parameter value can be used to determine whether to schedule the cleaning robot for operation in the subsequent stage.

[0050] Optionally, the generating of the dynamic weight value includes:

[0051] According to the rainfall and snowfall probability characteristics in the environmental risk characteristic values, determine whether to trigger the natural cleaning condition, and if the probability indicated by the determined rainfall and snowfall probability characteristics exceeds the preset judgment threshold, then reduce the cleaning urgency of the corresponding photovoltaic panel area;

[0052] Specifically, extract the rainfall and snowfall probability characteristic values from the environmental prediction parameter, denoted as P, which represents the probability of rainfall or snowfall occurring within a preset future duration, and the value range is from 0 to 1. Preset a judgment threshold, denoted as Tp, for example, Tp = 0.7, to evaluate whether natural precipitation is sufficient to clean the photovoltaic panel. When P≥Tp, the probability of rainfall or snowfall is relatively high, which may trigger the natural cleaning condition. At this time, reduce the cleaning urgency to reduce unnecessary cleaning tasks. When P<Tp, natural precipitation is not sufficient to clean the photovoltaic panel, and the cleaning urgency remains unchanged.

[0053] Based on the power generation loss rate curve, calculate the estimated value of the power generation loss corresponding to the current dust accumulation amount, and combine the wind force level characteristic in the environmental prediction parameter to generate a power generation benefit value;

[0054] Specifically, use the generated power generation loss rate curve and combine the current dust accumulation amount per unit area on the surface of the photovoltaic panel obtained in real time by the sensor to calculate the estimated value of the power generation loss caused by dust accumulation at present. At the same time, extract the wind force level characteristic from the environmental prediction parameter. Considering the estimated value of the power generation loss and the wind force level characteristic comprehensively to generate a power generation benefit value. The calculation of this power generation benefit value can refer to the following formula:

[0055] B = L loss *(1 + k w *F wind )

[0056] In the formula, B is the power generation benefit value, which reflects the quantitative index of the potential benefit that may be brought by performing the cleaning operation; L loss is the estimated value of the power generation loss calculated according to the current dust accumulation amount and the power generation loss rate curve; F wind is the quantified wind force level characteristic; k w is the wind force influence correction coefficient, which is used to characterize the influence degree of wind force conditions on the immediate power generation benefit or the suitability of the cleaning operation.

[0057] A dynamic weight value is generated and dynamically adjusted according to the power generation benefit value and a preset energy benefit model.

[0058] Specifically, based on the generated power generation benefit value and combined with a preset energy benefit model, the dynamic weight value of each photovoltaic panel area is finally generated and dynamically adjusted. The energy benefit model may integrate factors such as current market electricity prices, operation and maintenance cost targets, and the overall power generation strategy of the power station. The calculated dynamic weight value will be the direct basis for subsequent judgment of cleaning urgency and allocation of work priorities. The dynamic weight value can be calculated by referring to the following formula:

[0059] W=coeff B *B+M adj ,

[0060] Where W is the dynamic weight value; B is the calculated power generation benefit value; coeff B is the benefit conversion coefficient, which is used to adjust the power generation benefit value to the appropriate weight range; M adj It is a comprehensive adjustment item output by the energy benefit model, which can change dynamically according to real-time economic or strategic factors, such as Figure 3 shown.

[0061] For example, assume that the probability of rain or snowfall in a photovoltaic panel area is 0.2, which is less than the preset judgment threshold of 0.7. Therefore, the urgency of cleaning is not reduced due to the expected natural precipitation. If the estimated power loss value calculated by the power loss rate curve and the current dust accumulation in the area is 0.9%, the corresponding value of the wind level characteristic obtained from the environmental prediction parameters is 3, and the set wind impact correction coefficient is 0.05, then the power generation benefit value of the area can be calculated as 0.9% × (1 + 0.05 × 3) = 0.9% × 1.15 ≈ 1.035. Furthermore, if the set benefit conversion coefficient is 80 and the comprehensive adjustment item given by the preset energy benefit model based on the current operating strategy is 10, then the dynamic weight value of the area can be calculated as 80 × 1.035 + 10 = 82.8 + 10 = 92.8. This calculated dynamic weight value will be used in subsequent scheduling decisions.

[0062] Optionally, allocating operation priorities to each photovoltaic panel area includes:

[0063] Divide each photovoltaic panel area into a high-efficiency grid, a medium-efficiency grid, and a low-efficiency grid according to the cleaning urgency, and allocate the number of robots to each divided grid based on the endurance parameter in the preset robot operation capability parameters;

[0064] Specifically, the physical area of ​​the photovoltaic power station is divided into grids with different efficiency levels based on the cleaning urgency and the expected power generation efficiency of each photovoltaic panel area. Then, based on the preset robot operation capability parameters, especially the robot endurance parameter after a single charge, the number of robots to perform cleaning tasks is preliminarily matched and allocated to each divided grid. For example, the number of robots assigned to a certain grid is N. robot It can be estimated as follows:

[0065]

[0066] Where N robot is the number of robots assigned; A grid is the area of ​​the grid; T unit is the time required for the robot to clean a unit area; T battery is the effective endurance of the robot after it is fully charged; U factor is the comprehensive utilization factor of the robot, taking into account indirect operation time such as path movement and obstacle avoidance; ceil() represents the rounding-up function.

[0067] Establishing a priority task queue for the divided high-efficiency grid, and inserting the established priority task queue into the head of the initial cleaning task sequence to adjust the initial cleaning task sequence;

[0068] Specifically, the goal is to ensure that high-value cleaning tasks are prioritized. For each of the identified high-efficiency grids, the corresponding cleaning tasks are integrated to create a priority task queue. Tasks within this priority task queue can be further sorted based on the precise cleaning urgency or expected power generation benefit of each high-efficiency grid. This prioritized task queue is then inserted as a whole at the front of the initial cleaning task sequence, completing the initial adjustment of the initial cleaning task sequence and forming an intermediate task sequence that reflects the priority strategy.

[0069] The coverage mode of the divided low-efficiency grid is dynamically adjusted according to the acquired robot power data, and an updated cleaning task sequence is generated based on the adjusted initial cleaning task sequence.

[0070] Specifically, we focus on flexibly handling the cleaning tasks of low-efficiency grids based on the real-time power of the robot while ensuring high-priority tasks. We obtain the current power data of each robot in real time. Based on this power data, we dynamically adjust the cleaning coverage mode of the low-efficiency grids divided above. For example, the coverage mode M cover The choice can depend on the robot's current remaining battery percentage B level The comparison results with several preset power thresholds can be expressed in the following logic:

[0071] If B level ≥B MThresh , then M cover Select complete cleaning;

[0072] If B LThresh ≤B level <B MThresh , then M cover Select Partial Cleaning or Quick Cleaning mode;

[0073] If B level <B LThresh , then M cover Select Do not clean yet or Clean only critical points.

[0074] After completing the dynamic decision on the low-efficiency grid coverage method, these decisions are applied to the intermediate task sequence formed in the previous step, and finally a fully updated cleaning task sequence is generated.

[0075] For example, assume that a photovoltaic power station area is divided into a high-efficiency grid A and several other grids based on cleaning urgency and efficiency assessment. The robot's cleaning time per unit area is 0.0015 hours per square meter, the effective endurance on a single charge is 2.5 hours, and the comprehensive utilization factor is 0.8. The number of robots assigned to grid A is calculated as ceil((5000 × 0.0015) / (2.5 × 0.8)) = 4. The tasks of these robots are organized into a priority task queue and inserted at the head of the initial task sequence. Subsequently, if a robot preparing to perform a task for a low-efficiency grid has a remaining battery percentage of 40%, while the low battery threshold is set at 30% and the medium battery threshold is 60%, then because the battery level falls between the two thresholds, a partial cleaning strategy may be implemented for this low-efficiency grid. Task details are adjusted accordingly, ultimately forming an updated cleaning task sequence.

[0076] Optionally, generating a multi-robot collaborative path planning result includes:

[0077] Generating a corresponding initial path for each robot according to the updated cleaning task sequence;

[0078] Specifically, this step plans a preliminary route for the robot that performs each task in the updated cleaning task sequence. In the task sequence, each robot participating in the operation is assigned one or more photovoltaic panel areas that it is responsible for cleaning. Subsequently, a path planning algorithm is used to generate an initial path for each robot and its assigned task. Commonly used path planning algorithms include the A* algorithm or the Dijkstra algorithm. In addition, for cleaning tasks that cover a large area, specific coverage strategies such as zigzag traversal or bow traversal can also be selected. The optimization goal of path generation is usually to minimize a comprehensive cost function. The calculation of the total cost of the path can refer to the following formula:

[0079] C p =k d *∑L s +k e *∑E s ,

[0080] Where C p is the total path cost for the robot to complete the assigned task; ∑L s is the total travel distance of the path; ∑E s is the estimated total energy consumption during path execution; k d and k e are the preset weight coefficients of distance cost and energy consumption cost respectively. Based on the topological structure data of the photovoltaic power station, an initial path that minimizes the total cost of this path will be planned.

[0081] Acquire coordinate data and task execution status data uploaded by the robot in real time, and detect whether there is a path conflict area based on the acquired coordinate data and task execution status data;

[0082] Specifically, in order to achieve dynamic conflict detection, the current precise coordinate data uploaded by each robot in real time, as well as the task execution status data, are continuously obtained through the robot's onboard positioning module and communication interface. The positioning module is, for example, a global positioning system module or a Beidou positioning module. The task execution status data may include information such as the task number currently being executed by the robot, the progress percentage of the completed task, and the current driving speed vector. Based on these acquired real-time data, the positions of any two robots, such as robot and robot, after a short period of predicted time in the future are predicted. The predicted position vector of the robot after the future time can be estimated by the following formula:

[0083] P k,pred =P k,curr +V k,curr *Δt,

[0084] Where, P k,pred is the predicted position vector of robot k after the predicted time Δt in the future; P k,curris the current position vector of robot k; V k,curr is the current velocity vector of robot k. The Euclidean distance between the predicted positions of any two robots is calculated and compared with a preset safety distance threshold. If the calculated predicted distance is less than the safety distance threshold, a path conflict is determined between the two robots, and the area near their predicted positions is marked as a path conflict area.

[0085] If it is determined that a path conflict area exists, the coordinates of the path conflict area are determined, and based on the dynamic weight value associated with the priority task queue, the robot passage order is reallocated to generate a multi-robot collaborative path planning result.

[0086] Specifically, when a path conflict area is determined, the precise geographic coordinates of the path conflict area are first determined. Subsequently, to resolve the conflict, the order of passage between the conflicting robots is redistributed based on the established priority task queue and the dynamic weight values ​​associated with the tasks or corresponding robots in the queue. The allocation of the passage order can be based on the calculation of a passage priority value. The higher the passage priority value, the greater the possibility that the robot will be given priority passage. The calculation of this passage priority value can refer to the following formula:

[0087] P val,k =w task *W k +w queue *I k ,

[0088] Where, P val,k is the priority value of robot k; W k is the dynamic weight value corresponding to the current task of robot k; I k is an indicator. If the task of robot k comes from the priority task queue, the indicator is 1, otherwise it is 0; w task and w queue The two weights are the dynamic task weight and the preset weight coefficients for the priority queue. The robot with the higher priority value is given priority to pass through the conflicting path area, requiring other robots to adjust their paths or travel timing. Through this conflict resolution and coordinated path adjustment, a conflict-free multi-robot collaborative path planning result is ultimately generated.

[0089] For example, assume that, based on the updated cleaning task sequence, initial paths have been generated for Robot A and Robot B, respectively. After obtaining their real-time coordinates and velocity data, the position prediction formula calculates that the predicted distance between the two robots will be less than the preset safety distance of 2 meters in the next 10 seconds. Therefore, a path conflict zone is determined and the coordinates of this zone are determined. Furthermore, assume that the dynamic weight of Robot A's current task is 85, and its task originates from the priority task queue, while the dynamic weight of Robot B's current task is 60, and it does not originate from the priority task queue. If the weight coefficients in the priority calculation formula are set to 1 and 15, the priority of Robot A is calculated as 1×85+15×1=100, and the priority of Robot B is calculated as 1×60+15×0=60. Since Robot A has a higher priority than Robot B, Robot A is granted right of way. Robot B will perform an avoidance maneuver, such as pausing or planning a new path. The adjusted set of conflict-free paths forms part of the generated multi-robot collaborative path planning results.

[0090] Optionally, the reallocation of the robot passage order includes:

[0091] Generate a priority ranking list based on the power generation benefit value corresponding to the current task being performed by each robot;

[0092] Specifically, the goal is to prioritize tasks based on their economic value when paths conflict between robots or when coordination of travel order is required. All robots involved in potential path interactions or conflicts are identified and their current tasks are retrieved. For each current task, the calculated power generation benefit value corresponding to that task is retrieved. The robots are then sorted in descending order based on these power generation benefit values ​​to generate a priority list. This list clearly identifies which robots' tasks should be prioritized in resource competition situations.

[0093] Based on the priority ranking list, dividing the robot into a first priority and a second priority, wherein the first priority is higher than the second priority, locking the path use right of the path conflict area for the identified first priority robot, and triggering the identified second priority robot to perform detour path calculation;

[0094] Specifically, this step is to execute specific right-of-way allocation and avoidance strategies based on the generated priority list. From the priority list, the first-priority robot identified, for example, the robot with the highest ranking or the power generation benefit value higher than a certain threshold, will be granted the right to use the priority path of the conflict area, that is, it will be allowed to pass through the area according to the original plan or fine-tuned path. This process can be called locking the path. At the same time, for other second-priority robots identified in the list, they will be triggered to perform detour path calculation. The goal of the detour path calculation is to plan a new alternative path for these second-priority robots that can avoid the locked conflict area, while satisfying other constraints as much as possible, such as minimizing the additional driving distance or time. The cost of the detour path can be one of its evaluation indicators, for example:

[0095] C detour =L extra +w t *T extra ,

[0096] Where C detour is the comprehensive cost of the detour path; L extra is the additional driving distance of the detour route compared to the original route; T extra is the extra time consumed by the detour; t is the weight coefficient of time cost. Select the detour path with lower overall cost.

[0097] If the estimated time required to perform the detour path calculation exceeds a preset tolerance threshold, the current task of the second-priority robot is migrated to the identified idle robot.

[0098] Specifically, this step provides a remedial measure for situations where the detour cost is too high, in order to ensure overall work efficiency. After performing the detour path calculation for the second-priority robot, the estimated time of the detour path will be obtained. This estimated time is compared with a preset tolerance threshold. The tolerance threshold can be set according to factors such as the urgency of the task and the time window of the overall work plan. If the calculated estimated time of the detour path is greater than the preset tolerance threshold, it means that the cost of performing the detour is too high for the second-priority robot, which may seriously affect its subsequent tasks or overall efficiency. In this case, the task migration strategy will be executed to migrate the current task of the second-priority robot to another robot whose current status is an identified idle robot or a robot with a lower load and can complete the task faster. Task migration involves updating the task list and path planning of the relevant robots.

[0099] For example, assume that Robot A and Robot B meet in a conflict zone. Based on the power generation efficiency values ​​corresponding to their respective tasks, Robot A is ranked higher in the generated priority list and is identified as the first-priority robot, while Robot B is the second-priority robot. The right to use the path through the conflict zone is locked for Robot A. Subsequently, a detour path calculation is performed for Robot B, and a detour path is obtained, which is expected to take 15 minutes. If the preset tolerance threshold is 10 minutes, since 15 minutes exceeds 10 minutes, the detour cost is judged to be too high. At this time, a search is performed to see if there is currently an identified idle Robot C. If Robot C exists and is suitable for performing Robot B's current task, the task will be removed from Robot B's task queue and assigned to Robot C. At the same time, a new path for Robot C to perform this task will be planned. Robot B may be reassigned to other tasks or returned to standby.

[0100] Optionally, generating a multi-robot collaborative path planning result further includes:

[0101] Obtain the robot's historical movement trajectory data and extract path repetition rate characteristics and energy consumption distribution characteristics;

[0102] Specifically, this step aims to mine information that can be used for path optimization from the robot's historical operation data. The robot's historical movement trajectory data is obtained from the storage system. This data usually contains a series of coordinate points with timestamps recorded by each robot when performing tasks in the past. Based on this trajectory data, feature extraction is performed. The extraction of path repetition rate features can be achieved, for example, by counting the frequency of a specific road section or area traversed by the robot, or calculating the overlap between different paths. A simplified path segment repetition rate can be expressed as:

[0103]

[0104] Where R seg is the repetition rate of a specific road segment; N pass is the total number of times the road section is passed by the robot during the statistical period; N totalOps The total number of tasks or runs performed by the robot during the statistical period. Extracting energy consumption distribution features may involve analyzing energy consumption records corresponding to different sections or actions in the historical trajectory, thereby identifying high-energy consumption areas or high-energy consumption path patterns.

[0105] Based on the path repetition rate characteristics and the energy consumption distribution characteristics, optimizing the distribution positions of the turning points of each path in the multi-robot collaborative path planning result to generate an energy-balanced path;

[0106] Specifically, the extracted features are used to improve the path quality, with particular attention paid to the balance of energy consumption. Based on the path repetition rate characteristics and energy consumption distribution characteristics obtained through analysis, each robot path in the generated multi-robot collaborative path planning results is optimized, with the focus on adjusting the distribution positions of the turning points in the path. For example, if the repetition rate of a path segment is too high and unnecessary, or if some turning points have significantly high historical energy consumption due to factors such as small angles and complex terrain, the optimization algorithm will try to smooth these turning points, or fine-tune the path within the allowable range to avoid high-energy consumption points. The goal of the optimization is to generate an energy-balanced path, that is, to make the energy consumption of each robot path as low as possible, or to make the energy consumption distribution between robots more reasonable, while meeting the task requirements. The optimization of path turning points can introduce a cost evaluation function to evaluate the energy consumption impact of turning points:

[0107] C turn =k a *Δθ+k e *E histTurn ,

[0108] Where C turn is the optimization evaluation cost of the turning point; Δθ is the turning angle; E histTurn is the historical average energy consumption characteristic value of this type of turning point; k a and k e are the corresponding weight coefficients. The optimization process attempts to adjust the turning points to minimize the sum or maximum value along the path.

[0109] The multi-robot collaborative path planning result is updated according to the energy consumption balanced path.

[0110] Specifically, after generating energy-balanced paths, this step integrates these optimized paths into the final job plan. The original or pre-optimized paths for the corresponding robots in the multi-robot collaborative path planning results are replaced with the newly generated energy-balanced paths. This update ensures that the paths ultimately delivered to each robot are optimized for energy consumption and repeatability, thereby improving overall job economics and long-term operational efficiency. The updated multi-robot collaborative path planning results serve as the final basis for the robots' actual task execution.

[0111] For example, it is assumed that it is analyzed from the historical movement trajectory data of the robot that there is a sharp turning point T1 in the planned path P1 of a certain robot. The path repetition rate characteristics of this point show that it is frequently used for unnecessary U-turns, and the energy consumption distribution characteristics indicate that the historical average energy consumption here is relatively high. In the optimization step, based on these characteristics, the distribution position of this turning point T1 is adjusted, for example, by introducing two gentle turning points T2 and T3 to replace T1, thereby forming a new energy-balanced path P1 passing through T2 and T3. If the optimization evaluation cost of the turning angle and the optimization evaluation cost of the historical energy consumption characteristics are both reduced, the path P1' is confirmed to be better. Finally, in the original multi-robot collaborative path planning result, the energy-balanced path P1' is used to replace the original path P1, thereby completing the update of the multi-robot collaborative path planning result.

[0112] Optionally, the method further includes:

[0113] When an emergency environmental event is detected, low-priority tasks are interrupted according to the cleaning urgency, and robots are reallocated to perform emergency cleaning tasks;

[0114] Specifically, this step describes the method's response mechanism for unexpected, urgent environmental changes. Sudden environmental events are detected through specific monitoring channels, such as accessing an external meteorological disaster warning system, analyzing abnormal readings from sensors within the PV plant, or receiving manually issued alarm signals. Such sudden environmental events can include, for example, a short-lived severe sandstorm, hail, or a sudden drop in energy efficiency in a specific area due to unexpected pollution. Once such an event is detected, its potential impact on various areas of the PV plant is immediately assessed. Based on the generated cleaning urgency for each PV panel area, currently executing or pending low-priority tasks are interrupted. This interruption may involve pausing the robots performing these tasks and returning them to a standby state. Based on the nature, scope, and estimated duration of the sudden environmental event, one or more emergency cleaning tasks are defined. These emergency cleaning tasks typically aim to quickly restore power generation in critical areas or mitigate the negative impact of the event. Available robots are reallocated, prioritizing them to perform these newly generated emergency cleaning tasks.

[0115] Based on the emergency cleaning task, a dynamically compressed spiral progressive path is generated so that the robot can complete a preset coverage rate within the impact period of the sudden environmental event.

[0116] Specifically, this step plans a special and efficient path for the robot performing the emergency cleaning task. This path is called a dynamically compressed spiral progressive path, and its design goal is to complete the operation of achieving a preset coverage rate on the target area as quickly as possible within a limited impact period of the sudden environmental event, for example, within a critical period of 2 hours when a sandstorm is expected to last, or pollutants need to be removed within 1 hour. A spiral progressive path usually refers to a robot moving in a spiral manner from the edge of the area to the center, or covering a rectangular area with a spiral-like trajectory that continuously shrinks or expands. Its dynamic compression characteristics are reflected in the fact that the path parameters can be dynamically adjusted according to the urgency of the emergency task, the remaining available time, and the real-time development of the sudden environmental event. For example, if the remaining time is tight, the pitch or scanning spacing can be appropriately increased to sacrifice some fineness in exchange for faster coverage speed. The dynamic adjustment of the parameters of the spiral path can be referred to as follows:

[0117]

[0118] Where S p is the dynamic pitch currently calculated; S p,base is the standard or basic pitch; T remain is the remaining available time for the emergency cleaning task; T totalPeriod is the estimated total impact period of the emergency environmental event or the total time limit of the emergency task; λ is an adjustment factor used to control the sensitivity of the pitch to the remaining time. By adopting this path strategy, it is possible to ensure that the robot completes the cleaning of key areas with the highest efficiency in emergency situations, striving to achieve the preset cleaning coverage target within the impact period, such as Figure 4 shown.

[0119] For example, suppose a photovoltaic power plant is suddenly hit by a regional dust storm, with the impact expected to last three hours. Upon detection of this event, all routine cleaning tasks assessed as low priority are immediately interrupted. Based on the severity of the dust coverage and the importance of the affected area, an emergency cleaning task is generated, with the goal of completing 70% cleaning coverage of Area A within three hours. Robots A and B, both located nearby, are selected to perform this emergency task. A dynamic, compressed, spiral progressive path strategy is employed when generating paths for these robots. Initially, for example, in the first hour, due to the long remaining available time, the dynamic pitch may be close to the base pitch to ensure cleaning quality. Over time, if progress falls behind schedule or decreases, the pitch is dynamically increased according to the aforementioned formula to accelerate coverage, ensuring that the preset coverage target of 70% is achieved as closely as possible before the end of the three-hour impact period.

[0120] Optionally, the method further includes:

[0121] Acquiring actual power generation efficiency data of the photovoltaic panels after the task is completed, and generating cleaning effect verification parameters based on the actual power generation efficiency data of the photovoltaic panels;

[0122] Specifically, the purpose of this step is to quantitatively evaluate the actual cleaning effect after the cleaning task is completed. During a predetermined observation period after the robot completes the cleaning operation of the designated photovoltaic panel area, the actual power generation efficiency data of the photovoltaic panels in the cleaned area is obtained through the data acquisition and monitoring system or inverter and other equipment of the photovoltaic power station. This data reflects the power generation performance of the photovoltaic panels under real environmental conditions after cleaning. Based on the actual power generation efficiency data of the photovoltaic panels obtained, a cleaning effect verification parameter is further generated. This parameter is intended to objectively evaluate whether the cleaning operation has achieved the expected effect. For example, the calculation method of a cleaning effect verification parameter can be as follows:

[0123]

[0124] Where V clean is the calculated cleaning effect verification parameter, and its value is usually between 0 and 1. The closer it is to 1, the better the cleaning effect. after is the actual power generation efficiency of the photovoltaic panel after the task is completed; η before is the power generation efficiency of the photovoltaic panel before cleaning; η ideal It is the theoretical or reference power generation efficiency of the photovoltaic panel under ideal clean conditions.

[0125] The cleaning effect verification parameter is compared with a preset verification threshold. If it is determined that the cleaning effect verification parameter deviates from the preset verification threshold, the association matching rule of the environmental prediction parameter is adjusted.

[0126] Specifically, after obtaining the cleaning effect verification parameters, this step determines whether the cleaning effect meets the standards by comparing them with the preset standards, and adaptively adjusts the internal rules of the prediction model accordingly to achieve continuous optimization. The generated cleaning effect verification parameters are compared with one or more preset verification thresholds. For example, a lower threshold and an expected target threshold may be set. If it is determined that the cleaning effect verification parameters deviate from the preset verification threshold, for example, when the calculated cleaning effect verification parameters are continuously lower than the lower threshold, it indicates that the actual cleaning effect has not met expectations, which may mean that there is a deviation in the previous assessment of environmental factors or the effectiveness of the cleaning strategy. At this point, it is necessary to adjust the association matching rules of the environmental prediction parameters. The environmental prediction parameters are generated through a series of association matching, and the generation process depends on several internal model parameters or rules. Adjusting these association matching rules can be achieved through a learning algorithm to update the relevant rules or coefficients according to the degree of deviation. For example, if a coefficient is part of the rule to be adjusted, it can be updated in the following way:

[0127] k adjust,new =k adjust,old +Δk×δ V ,

[0128] Where k adjust,new is the new value of the adjusted coefficient; k adjust,old is the coefficient value before adjustment; Δk is a learning rate or adjustment step size. Through this feedback adjustment mechanism, the internal models and rules used to generate environmental prediction parameters can continuously approach the actual situation, thereby improving the accuracy and effectiveness of subsequent scheduling decisions.

[0129] Based on the same inventive concept, the present invention also provides a photovoltaic cleaning robot operation scheduling system based on environmental prediction, such as Figure 5 As shown, the system includes:

[0130] Data acquisition module, used to obtain weather forecast data, real-time dust accumulation data of photovoltaic panels and historical power generation efficiency attenuation data;

[0131] A prediction model module, connected to the data acquisition module, for generating environmental prediction parameters and calculating power generation benefit values;

[0132] a scheduling generation module, connected to the prediction model module, for allocating job priorities according to the power generation benefit value and generating multi-robot collaborative path planning results;

[0133] A dynamic adjustment module, connected to the scheduling generation module, for responding to sudden environmental events and updating the path planning;

[0134] The verification feedback module is used to optimize the matching rules of the prediction model module according to the actual power generation efficiency data after the task is completed.

[0135] It should be noted that the functional division and information interaction between the aforementioned modules are logical. Physically, they can be integrated into the same software platform or deployed in a distributed manner. The connections between them represent data and control flows, designed to collaboratively achieve the dynamic optimization of building energy consumption of the present invention. The foregoing description is merely an exemplary embodiment of the present invention and is not intended to limit its scope.

Claims

1. A photovoltaic cleaning robot operation scheduling method based on environmental prediction, characterized in that: The method comprises: Acquiring weather forecast data, real-time dust accumulation data of photovoltaic panels, and historical power generation efficiency attenuation data, and generating environmental prediction parameters based on the weather forecast data, the real-time dust accumulation data of photovoltaic panels, and the historical power generation efficiency attenuation data; Based on the environmental prediction parameters and in combination with a preset power generation loss threshold, a dynamic weight value is generated, and the cleaning urgency of each photovoltaic panel area is determined by the dynamic weight value; Assigning a work priority to each photovoltaic panel area according to the determined cleaning urgency, and generating an initial cleaning task sequence based on preset robot operation capability parameters; According to the initial cleaning task sequence, combined with the preset photovoltaic power station topology data and the acquired real-time position data of the robots, a multi-robot collaborative path planning result is generated.

2. The photovoltaic cleaning robot operation scheduling method based on environmental prediction according to claim 1 is characterized in that: The generation environment prediction parameters include: Performing time series processing on the weather forecast data to extract environmental risk characteristic values ​​within a preset time period in the future, wherein the environmental risk characteristic values ​​include wind force level characteristics, rainfall and snowfall probability characteristics, and air humidity fluctuation range; Analyzing the corresponding relationship between the amount of dust accumulation and the power generation decline rate in the historical power generation efficiency decay data to generate a power generation loss rate curve; The environmental risk characteristic value is correlated and matched with the power generation loss rate curve to generate environmental prediction parameters.

3. The photovoltaic cleaning robot operation scheduling method based on environmental prediction according to claim 2 is characterized in that: Generating a dynamic weight value includes: Determining whether a natural cleaning condition is triggered based on the rainfall or snowfall probability characteristics in the environmental risk characteristic value, and reducing the cleaning urgency of the corresponding photovoltaic panel area if the probability indicated by the rainfall or snowfall probability characteristics exceeds a preset judgment threshold; Based on the power generation loss rate curve, an estimated power generation loss value corresponding to the current dust accumulation amount is calculated, and combined with the wind force level characteristics in the environmental prediction parameters, a power generation benefit value is generated; A dynamic weight value is generated and dynamically adjusted according to the power generation benefit value and a preset energy benefit model.

4. The photovoltaic cleaning robot operation scheduling method based on environmental prediction according to claim 3 is characterized in that: The allocation of operation priorities to each photovoltaic panel area includes: Divide each photovoltaic panel area into a high-efficiency grid, a medium-efficiency grid, and a low-efficiency grid according to the cleaning urgency, and allocate the number of robots to each divided grid based on the endurance parameter in the preset robot operation capability parameters; Establishing a priority task queue for the divided high-efficiency grid, and inserting the established priority task queue into the head of the initial cleaning task sequence to adjust the initial cleaning task sequence; The coverage mode of the divided low-efficiency grid is dynamically adjusted according to the acquired robot power data, and an updated cleaning task sequence is generated based on the adjusted initial cleaning task sequence.

5. The photovoltaic cleaning robot operation scheduling method based on environmental prediction according to claim 4 is characterized in that: Generating the multi-robot collaborative path planning results includes: Generating a corresponding initial path for each robot according to the updated cleaning task sequence; Acquire coordinate data and task execution status data uploaded by the robot in real time, and detect whether there is a path conflict area based on the acquired coordinate data and task execution status data; If it is determined that a path conflict area exists, the coordinates of the path conflict area are determined, and based on the dynamic weight value associated with the priority task queue, the robot passage order is reallocated to generate a multi-robot collaborative path planning result.

6. The photovoltaic cleaning robot operation scheduling method based on environmental prediction according to claim 5 is characterized in that: The reallocation of the robot's passing order includes: Generate a priority ranking list based on the power generation benefit value corresponding to the current task being performed by each robot; Based on the priority ranking list, dividing the robot into a first priority and a second priority, wherein the first priority is higher than the second priority, locking the path use right of the path conflict area for the identified first priority robot, and triggering the identified second priority robot to perform detour path calculation; If the estimated time required to perform the detour path calculation exceeds a preset tolerance threshold, the current task of the second-priority robot is migrated to the identified idle robot.

7. The photovoltaic cleaning robot operation scheduling method based on environmental prediction according to claim 5 is characterized in that: Generating the multi-robot collaborative path planning result further includes: Obtain the robot's historical movement trajectory data and extract path repetition rate characteristics and energy consumption distribution characteristics; Based on the path repetition rate characteristics and the energy consumption distribution characteristics, optimizing the distribution positions of the turning points of each path in the multi-robot collaborative path planning result to generate an energy-balanced path; The multi-robot collaborative path planning result is updated according to the energy consumption balanced path.

8. The photovoltaic cleaning robot operation scheduling method based on environmental prediction according to claim 1 is characterized in that: The method further comprises: When an emergency environmental event is detected, low-priority tasks are interrupted according to the cleaning urgency, and robots are reallocated to perform emergency cleaning tasks; Based on the emergency cleaning task, a dynamically compressed spiral progressive path is generated so that the robot can complete a preset coverage rate within the impact period of the sudden environmental event.

9. The photovoltaic cleaning robot operation scheduling method based on environmental prediction according to claim 1, characterized in that: The method further comprises: Acquiring actual power generation efficiency data of the photovoltaic panels after the task is completed, and generating cleaning effect verification parameters based on the actual power generation efficiency data of the photovoltaic panels; The cleaning effect verification parameter is compared with a preset verification threshold. If it is determined that the cleaning effect verification parameter deviates from the preset verification threshold, the association matching rule of the environmental prediction parameter is adjusted.

10. A photovoltaic cleaning robot operation scheduling system based on environmental prediction, applied to a photovoltaic cleaning robot operation scheduling method based on environmental prediction according to any one of claims 1 to 9, characterized in that: The system comprises: Data acquisition module, used to obtain weather forecast data, real-time dust accumulation data of photovoltaic panels and historical power generation efficiency attenuation data; A prediction model module, connected to the data acquisition module, for generating environmental prediction parameters and calculating power generation benefit values; a scheduling generation module, connected to the prediction model module, for allocating job priorities according to the power generation benefit value and generating multi-robot collaborative path planning results; A dynamic adjustment module, connected to the scheduling generation module, for responding to sudden environmental events and updating the path planning; The verification feedback module is used to optimize the matching rules of the prediction model module according to the actual power generation efficiency data after the task is completed.

Images