Intelligent operation and maintenance method and system for photovoltaic inspection robot and electronic equipment

Abstract

The invention relates to the field of data processing, and provides an intelligent operation and maintenance method and system for a photovoltaic inspection robot and electronic equipment, which can avoid the relevance failure among data by carrying out differentiated extraction on data of different sensors, only retaining local key information, reducing the transmission of invalid information and carrying out protocol unification and timestamp alignment. Then, on the basis of analysis and recognition, a polling path is generated through a dynamic polling priority queue and a dual-mode path planning algorithm, and finally, polling and operation and maintenance are conducted in sequence according to the path and an operation and maintenance task queue. According to the invention, the transmission and association of invalid information can be reduced, the calculation pressure and communication occupation of system hardware can be reduced, the response speed of the system can be improved, and on the basis, the system can utilize the dynamic inspection priority queue and dual-mode path planning to realize instant response aiming at different conditions, and the inspection efficiency, accuracy and effectiveness can be improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to an intelligent operation and maintenance method, system and electronic equipment for a photovoltaic inspection robot. Background Art

[0002] In the traditional photovoltaic inspection field, the common track-mounted inspection robot, for example, is equipped with multiple sensors to enable real-time monitoring and perception of equipment and the environment within the distribution room. It can detect indoor temperature and humidity, and effectively monitor toxic and hazardous gases. The onboard infrared thermal imager can detect the temperature of the distribution equipment, automatically issuing alarms for over-temperature and high-temperature conditions. The robot can also detect partial discharges in the distribution room's distribution equipment, with the results uploaded to the AIoT robot inspection platform in real time. Images of photovoltaic panels are captured using image acquisition equipment, and the data is then transmitted back to the control center for simple analysis to determine if any panels are significantly damaged. In terms of mobility, the robot relies on preset tracks or simple path planning algorithms to navigate within the photovoltaic power station.

[0003] In the existing technology, in order to ensure the accuracy and effectiveness of inspections, real-time inspection path planning may be adopted, which is usually achieved by improving the path-finding rule algorithm to fully consider the inspection paths actually required in different situations. However, this method is not effective in real-time path-finding in emergency situations. When faced with complex environments and large amounts of data, due to the poor pertinence of data processing in existing technologies, a large amount of invalid information is processed and transmitted multiple times, affecting the system response speed. For example, directly transmitting and analyzing all sensor data between modules, directly performing comprehensive processing on different types of data, directly determining the inspection path based on the path planning algorithm, etc., will waste a lot of hardware resources to varying degrees, resulting in a slower final system response speed. When the response speed is too slow, even if the information about the external situation is very accurate, the system processing is too slow, and it is ultimately difficult to make real-time adjustments based on the actual situation, which to a certain extent affects the recognition accuracy and effectiveness.

[0004] Therefore, in order to ensure the accuracy and effectiveness of inspections, it is necessary to solve the problem from multiple dimensions at the same time. Summary of the Invention

[0005] In response to the problem in the prior art that a large amount of invalid information is processed and transmitted multiple times, resulting in low inspection accuracy and effectiveness, the present invention provides a photovoltaic inspection robot intelligent operation and maintenance method, system and electronic equipment. By performing differential extraction of data from different sensors, only local key information is retained, the transmission of invalid information is reduced, and protocol unification and timestamp alignment are performed to avoid the failure of correlation between data. Then, based on analysis and identification, an inspection path is generated through a dynamic inspection priority queue and a dual-mode path planning algorithm. Finally, inspections and operations are carried out in sequence according to the path and operation and maintenance task queue. The present invention can reduce the transmission and association of invalid information, reduce the computing pressure and communication occupancy of system hardware, and improve the system response speed. On this basis, the system can use dynamic inspection priority queues and dual-mode path planning to achieve immediate response to different situations, improving inspection efficiency, accuracy and effectiveness.

[0006] The following are the technical solutions of the present invention.

[0007] The present invention provides an intelligent operation and maintenance method for a photovoltaic inspection robot, comprising: Collect image data, temperature data, and environmental data of photovoltaic modules in photovoltaic power plants, and perform differential extraction to retain only local key information; Through standardized data interfaces, the protocols of multiple types of key information in the data acquisition module are unified and timestamps are aligned to obtain standardized data. Conduct defect identification and operation status analysis on standardized data to obtain analysis results; Obtain PV power plant layout data, combine analysis results, environmental data, and time synchronization information, and generate inspection routes using dynamic inspection priority queues and a dual-mode path planning algorithm; Generate an operation and maintenance task queue based on the analysis results and store the entire process data; Perform movement and maintenance operations on the robot body according to the inspection path and operation and maintenance task queue.

[0008] In the present invention, data dimensions are reduced through differentiated extraction, transmission volume is compressed synchronously, and communication load is reduced; data temporal and spatial consistency and parsing efficiency are ensured through protocol unification and timestamp alignment, avoiding association failure; on this basis, path reconstruction in special circumstances is achieved through dynamic inspection priority queues and dual-mode path planning algorithms, thereby improving inspection efficiency and pertinence.

[0009] As a possible implementation method, the image data, temperature data of photovoltaic modules in the photovoltaic power station and environmental data of the photovoltaic power station are collected and differentially extracted to retain only local key information, including: By using the feature significance scoring function, the region with higher weight is selected for ROI extraction, and only the local image containing potential defects is retained to focus on the potential defect area in the image data. At the same time, the sliding window filtering algorithm is used to extract the temperature and environmental data.

[0010] In the present invention, different methods of extraction are performed on data from different sensor sources, only local key information is retained, and the processing and transmission of invalid data is reduced.

[0011] As a possible implementation method, the method of performing protocol unification and timestamp alignment on multiple types of key information of the data acquisition module through a standardized data interface to obtain standardized data includes: A unified data structure with a pre-defined format is defined for image, temperature, and environmental data. The structure includes a data type label, acquisition device ID, raw value, and quality factor, where the quality factor reflects the integrity and credibility of the current data. It is connected to a precision clock synchronization system and uses the difference between the master clock broadcast signal and the local timing of the slave node to dynamically calibrate the sampling time to ensure that the time error of all sampling nodes in the system is controlled within the preset error. Each frame of data is bound to a unique timestamp when it is generated. By minimizing the time alignment cost function between different data sequences, the distance matrix is ​​traversed through dynamic programming to backtrack the optimal alignment path. After the alignment is completed, the result is constrained with the target error window to ensure the consistency of multi-source data in the time dimension.

[0012] This invention utilizes a standardized data interface and timestamp alignment mechanism to uniformly define a data format encompassing "data type tags, device IDs, raw values, and quality factors," addressing heterogeneous multi-source data protocols and improving data parsing efficiency across modules. Furthermore, by quantifying data integrity through quality factors, the system automatically filters low-reliability data, preventing invalid data from entering the processing flow. This improves data reliability from the source and reduces the risk of misjudgment in subsequent analysis. Furthermore, through a dynamic master-slave node calibration mechanism and utilizing the Dynamic Time Warping (DTW) algorithm to minimize the time alignment cost, the sampling time error across the entire system is controlled, ensuring strict temporal alignment of multi-source data (such as images, temperature, and environmental parameters). This ultimately avoids data correlation failures caused by frequency discrepancies and improves the accuracy of subsequent analysis of multi-physics coupling faults.

[0013] As a possible implementation, performing defect identification and operating status analysis on the standardized data to obtain analysis results includes: A lightweight convolutional neural network model based on knowledge distillation optimization is used to identify defects in photovoltaic module image data. It uses a fusion structure of gated recurrent units and Transformer encoders to extract long-term dependency information and assigns dynamic weights to different feature dimensions through a multi-head attention mechanism to perform operational status analysis. Among them, defect identification and operation status analysis are based on a distributed computing architecture. Image recognition tasks and status analysis tasks are processed by independent computing nodes respectively. Task allocation between nodes is dynamically adjusted in real time through a load balancing algorithm. The load balancing algorithm improves processing efficiency by minimizing the overall task time.

[0014] As a possible implementation method, the method of acquiring photovoltaic power station layout data, combining analysis results, environmental data, and time synchronization information, and generating an inspection path through a dynamic inspection priority queue and a dual-mode path planning algorithm includes: Determine the dynamic inspection priority queue based on the defect level and environmental risk level. The priority of each PV module is generated based on the defect level, environmental risk level and corresponding weights, with the defect level weighting being greater than the environmental risk level weighting. Set up a dual-mode path planning engine, where: In normal mode, an improved ant colony algorithm is used to combine the distribution of defect hotspots with future weather forecast data to construct a global path. In this algorithm, each ant makes decisions based on a heuristic function and pheromone intensity during the path selection process, and periodically and automatically re-evaluates the global path to ensure that all potential high-risk areas are covered. When at least one of the defect level or environmental risk level is at the highest level, the system switches to emergency mode. In emergency mode, a cost function is constructed based on real-time environmental data and component location information. The comprehensive cost of the current path node is calculated based on the actual cost from the starting node to the current node and the heuristic cost from the current node to the target node. Node expansion is performed and path planning is completed with the goal of minimizing the comprehensive cost.

[0015] In the present invention, the normal mode and emergency mode dynamically adjust the path in response to different situations, and rely on low-latency data with high-precision timestamps to ensure the spatiotemporal consistency of path planning, prioritize inspections of high-risk components, significantly enhance preventive maintenance capabilities, and reduce the robot's ineffective movement distance.

[0016] As a possible implementation manner, the step of obtaining the defect level includes: Obtain the defect type and confidence level obtained by defect identification, and determine the defect level based on the confidence level.

[0017] As a possible implementation method, if the generation of the inspection path is blocked, an alternative inspection path is generated with reference to prior knowledge, including: pre-selection through a minimum path cost function based on a defect distribution density function, in which the higher the defect density, the greater the path cost, thereby guiding path planning to avoid or prioritize coverage of high-risk areas.

[0018] As a possible implementation method, generating an operation and maintenance task queue based on the analysis results and storing full-process data at the same time include: An operation and maintenance task queue is generated based on the defect level and processing time. The priority of each operation and maintenance task is generated based on the defect level, processing time and corresponding weights, and the weight of the defect level is greater than the weight of the processing time.

[0019] The present invention also provides a photovoltaic inspection robot intelligent operation and maintenance system for executing any one of the above photovoltaic inspection robot intelligent operation and maintenance methods, including: A data acquisition module is used to collect image data, temperature data of photovoltaic modules in a photovoltaic power station, and environmental data of the photovoltaic power station; The data fusion processing module is used to unify the protocols and align the timestamps of multiple types of key information from the data acquisition module through a standardized data interface to obtain standardized data; Deep learning processing module, used to perform defect identification and operation status analysis on standardized data to obtain analysis results; The path optimization module is used to obtain PV power plant layout data, combine analysis results, environmental data and time synchronization information, and generate inspection paths through dynamic inspection priority queues and a dual-mode path planning algorithm; The operation and maintenance management module is used to generate an operation and maintenance task queue based on the analysis results and store the entire process data; The robot body is used to perform movement and operation and maintenance operations according to the inspection path and operation and maintenance task queue.

[0020] The present invention also provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the steps of the above-mentioned photovoltaic inspection robot intelligent operation and maintenance method are implemented.

[0021] The present invention also provides a storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are loaded and executed by a processor, the steps of the above-mentioned photovoltaic inspection robot intelligent operation and maintenance method are implemented.

[0022] The substantial effects of the present invention include: The present invention reduces data dimensions through differentiated extraction, synchronously compresses transmission volume, and reduces communication load; ensures data temporal and spatial consistency and parsing efficiency through protocol unification and timestamp alignment, avoiding association failure; on this basis, realizes path reconstruction in special circumstances through dynamic inspection priority queues and dual-mode path planning algorithms, thereby improving inspection efficiency and pertinence. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flow chart of an embodiment of the present invention; Figure 2 It is a system block diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0025] It should be understood that in various embodiments of the present invention, the size of the sequence number of each process does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0026] It should be understood that in the present invention, "include" and "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or apparatuses.

[0027] It should be understood that in the present invention, "multiple" refers to two or more. "And / or" is only a description of the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "Contains A, B and C", "Contains A, B, C" means that A, B, and C are all included, "Contains A, B or C" means that one of A, B, and C is included, and "Contains A, B and / or C" means that any one, any two, or any three of A, B, and C are included.

[0028] The technical solution of the present invention is described in detail below with reference to specific embodiments. The embodiments may be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0029] Example: Intelligent operation and maintenance method of photovoltaic inspection robot, such as Figure 1 As shown, including: Collect image data, temperature data, and environmental data of photovoltaic modules in photovoltaic power plants, and perform differential extraction to retain only local key information; Through standardized data interfaces, the protocols of multiple types of key information in the data acquisition module are unified and timestamps are aligned to obtain standardized data. Conduct defect identification and operation status analysis on standardized data to obtain analysis results; Obtain PV power plant layout data, combine analysis results, environmental data, and time synchronization information, and generate inspection routes using dynamic inspection priority queues and a dual-mode path planning algorithm; Generate an operation and maintenance task queue based on the analysis results and store the entire process data; Perform movement and maintenance operations on the robot body according to the inspection path and operation and maintenance task queue.

[0030] This embodiment also provides a photovoltaic inspection robot intelligent operation and maintenance system for executing the above photovoltaic inspection robot intelligent operation and maintenance method, such as Figure 2 As shown, including: A data acquisition module is used to collect image data, temperature data of photovoltaic modules in a photovoltaic power station, and environmental data of the photovoltaic power station; The data fusion processing module is used to unify the protocols and align the timestamps of multiple types of key information from the data acquisition module through a standardized data interface to obtain standardized data; Deep learning processing module, used to perform defect identification and operation status analysis on standardized data to obtain analysis results; The path optimization module is used to obtain PV power plant layout data, combine analysis results, environmental data and time synchronization information, and generate inspection paths through dynamic inspection priority queues and a dual-mode path planning algorithm; The operation and maintenance management module is used to generate an operation and maintenance task queue based on the analysis results and store the entire process data; The robot body is used to perform movement and operation and maintenance operations according to the inspection path and operation and maintenance task queue.

[0031] There is also a communication module to realize data transmission between modules, integrating data buffer area and hardware acceleration interface.

[0032] As a possible implementation method, the image data, temperature data of photovoltaic modules in the photovoltaic power station and environmental data of the photovoltaic power station are collected and differentially extracted to retain only local key information, including: Through the collaborative work of multiple sensor types, efficient collection and preliminary processing of key operational data from photovoltaic power plants are achieved. The image acquisition unit consists of multiple high-definition cameras, and the edge computing unit integrates an attention mechanism to improve the intelligence of image data processing and transmission efficiency. This attention mechanism uses a feature saliency scoring function to focus on potential defect areas in the image. Its core is the weighted feature extraction formula: ; in, represents the attention weight of the i-th region, is the characteristic response value of the region, and n is the number of regions in which the image is divided. The system selects regions with higher weights for ROI extraction, retaining only local images containing potential defects, significantly reducing the redundancy and transmission overhead of image data. The temperature acquisition unit uses infrared temperature sensors to capture the surface thermal information of photovoltaic modules, while the environmental acquisition unit includes light, wind speed, and humidity sensors to collect key factors affecting the environment's photovoltaic operation. The edge computing unit uses a sliding window filtering algorithm for temperature and environmental data to improve data stability. Its mathematical expression is: ; in, is the smoothed output value at the tth moment, is the original sampling value, and w is the window width. The filtered data is further dimensional compressed through principal component analysis to retain the main information of the data and improve the efficiency of subsequent processing. The core conversion formula of principal component analysis is: ; Where X is the standardized data matrix, W is the principal component direction matrix obtained by eigenvalue decomposition, and Z is the feature matrix after dimensionality reduction. To further constrain the compression ratio, the edge computing unit determines the number of principal components to be selected by accumulating the variance contribution rate and selects the minimum number of principal components k that meets the following conditions: ; in, represents the i-th eigenvalue, m is the total number of original dimensions, The variance contribution rate threshold for retention is set between 0.2 and 0.3 to ensure that the compressed data retains its core features. All pre-processed data is sent to the communication module via a standardized data bus, which then forwards it to the data fusion processing module for protocol matching and time alignment, ensuring the real-time and consistency of system-level data flow.

[0033] In this embodiment, for image data, the ROI region extraction technology based on the attention mechanism has significant advantages. Traditional image data transmission often contains a large amount of redundant information, while this embodiment only transmits local images that may have defects, greatly reducing invalid data transmission. This not only reduces the pressure on the data transmission bandwidth, but also enables subsequent processing modules to focus on key areas and improve processing efficiency. Taking a medium-sized photovoltaic power station as an example, the amount of image data collected every day can reach several GB. After adopting this technology, the amount of data transmitted can be reduced by about 80%, effectively alleviating the burden of network transmission. For temperature and environmental data, the sliding window filter combined with the principal component analysis method can reduce the data dimension while retaining key information. The sliding window filter algorithm removes noise interference by smoothing the time series data, making the data more stable and reliable. Principal component analysis further extracts the main features of the data, reduces the data dimension by more than 50% while ensuring data quality, and improves the speed and efficiency of subsequent data processing.

[0034] As a possible implementation method, the method of performing protocol unification and timestamp alignment on multiple types of key information of the data acquisition module through a standardized data interface to obtain standardized data includes: The data fusion processing module undertakes the core tasks of standardization and time alignment of multi-source heterogeneous data. Through the collaborative work of the protocol conversion unit and the time synchronization unit, it ensures that the data received by the subsequent deep learning module is consistent in structure and timing. The protocol conversion unit uniformly defines the data structure in JSON format for three types of data: image, temperature, and environment. The structure contains the data type label, acquisition device ID, original value, and quality factor. The quality factor reflects the integrity and credibility of the current data, and its value range is [0,1]. The unified data structure improves the data parsing efficiency between the processing units in the system, and optimizes the transmission structure through field-level indexing. The time synchronization unit is connected to the IEEE 1588 precision clock synchronization system, and uses the difference between the master clock broadcast signal and the local timing of the slave node to dynamically calibrate the sampling time to ensure that the time error of all sampling nodes in the system is controlled within 1 Within.

[0035] This embodiment adds high-precision timestamps to image, temperature, and environmental data, and with the help of a synchronized clock system, can accurately ensure the consistency of multi-source data in the time dimension. In the actual operation of a photovoltaic power station, the time when different types of sensors collect data may differ. If the time is not synchronized, it will cause serious errors in the analysis of the status of photovoltaic modules. For example, when analyzing the relationship between the temperature change and power generation efficiency of photovoltaic modules, if the timestamps of the temperature data and the power generation data are inconsistent, incorrect conclusions may be drawn, affecting the operation and maintenance decisions of the power station. This embodiment uses a time synchronization mechanism to control the time error within a very small range, avoiding the status analysis error caused by time misalignment, and providing a solid guarantee for the stable operation of the system.

[0036] In this embodiment, each frame of data is bound to a unique timestamp when it is generated. The timestamp is expressed as: ; in, Indicates the timestamp of the i-th frame data, is the initial time of the master clock, The synchronization offset between the frame data and the master clock is continuously updated through the bidirectional messaging mechanism in the IEEE 1588 protocol to maintain synchronization accuracy. To eliminate timing drift caused by different sampling frequencies of different sensor types, the system introduces the Dynamic Time Warping (DTW) algorithm, the core of which is to minimize the time alignment cost function between different data sequences. The recursive formula for the DTW cumulative distance is: ; in, and Represent the sampling values ​​of two different types of sensors at the i-th and j-th moments, respectively. To minimize the cumulative cost of aligning to this point, the formula traverses the distance matrix through dynamic programming and backtracks to find the optimal alignment path. To meet the system's strict control of time alignment error, after DTW alignment is completed, the result is constrained with the target error window, which satisfies: ; in, and are the timestamps of the two types of data after alignment, The system sets a time alignment error tolerance of 5ms to ensure high temporal consistency of multi-source data. The standardized data is pushed to the deep learning processing module via a message queue mechanism using the MQTT protocol. This ensures low-latency transmission while providing high concurrent processing capabilities to support the real-time computing needs of the image recognition and state analysis units.

[0037] In this embodiment, a standardized data interface and timestamp alignment mechanism are used to uniformly define a data format encompassing "data type tags, device IDs, raw values, and quality factors." This addresses the heterogeneity of multi-source data protocols and improves data parsing efficiency across modules. Furthermore, by quantifying data integrity through quality factors, the system automatically filters low-reliability data, preventing invalid data from entering the processing flow. This improves data reliability from the source and reduces the risk of misjudgment in subsequent analysis. A dynamic master-slave node calibration mechanism utilizes the dynamic time warping (DTW) algorithm to minimize the time alignment cost, controlling sampling time errors across the entire system and ensuring strict temporal alignment of multi-source data (such as images, temperature, and environmental parameters). This ultimately avoids data correlation failures caused by frequency discrepancies and improves the accuracy of subsequent analysis of multi-physics coupling faults.

[0038] For example, when a photovoltaic module hot spot fails, the time difference between infrared temperature data and image data can be precisely matched, avoiding fault location errors caused by time misalignment. Alternatively, when a wind speed sensor (high-frequency sampling) and a camera (low-frequency sampling) simultaneously capture module vibrations caused by strong winds, the system can precisely align their timing, avoiding data correlation failures caused by frequency differences and improving the accuracy of deep learning models for analyzing multi-physics coupled faults.

[0039] As a possible implementation, performing defect identification and operating status analysis on the standardized data to obtain analysis results includes: The deep learning processing module, centered around parallel computing, accurately analyzes image and time series data through the collaborative work of the image recognition unit and the state analysis unit. The image recognition unit employs a lightweight convolutional neural network model based on knowledge distillation optimization. This model uses ResNet-50 as the teacher model and constructs a loss function by minimizing the difference between the output distributions of the student and teacher models. This reduces the number of parameters by over 60%, making it suitable for deployment on embedded GPU platforms. The distillation loss function is defined as: ; in, represents the student model prediction label, is the true label, and are the logits outputs of the student model and the teacher model respectively, T is the temperature coefficient, is the loss weight coefficient, is the cross entropy loss, is the Kullback-Leibler divergence. This model achieves real-time performance with a single-image inference latency of less than 20ms on embedded platforms such as the NVIDIA Jetson. The state analysis unit utilizes a fusion structure of a gated recurrent unit (GRU) and a Transformer encoder to enhance the global perception and multidimensional data representation capabilities of time series modeling. The GRU state update process is as follows: ; ; in, is the current hidden state, is the current input, To update the gate, To reset the gate, and is the weight matrix. The Transformer encoder is used to extract long-term dependency information and assign dynamic weights to different feature dimensions through a multi-head attention mechanism. The calculation formula is: ; Among them, Q, K, and V are query, key, and value matrices respectively. This mechanism significantly enhances the model's ability to selectively focus on different data dimensions, such as temperature, current, and humidity, achieving a state analysis delay of no more than 10ms. The entire deep learning processing module operates on a distributed computing architecture, with image recognition and state analysis tasks handled by independent computing nodes. Task allocation is dynamically adjusted in real time between nodes using a load balancing algorithm. The load balancing objective function is: ; in, is the computational complexity of the task currently being processed by the i-th node, The load balancing mechanism minimizes the overall task time and improves processing efficiency by allocating computing power to the node. This parallel architecture increases overall system throughput by more than three times, ensuring recognition accuracy while meeting the requirements of real-time data processing and large-scale deployment in photovoltaic inspections.

[0040] This embodiment utilizes knowledge distillation technology in the CNN model, a key approach to achieving model lightweighting while ensuring recognition accuracy. Taking the simplification of ResNet-50 to MobileNet-V3 as an example, knowledge distillation transfers knowledge from the teacher model (ResNet-50) to the student model (MobileNet-V3). This reduces model parameters by over 60% while still maintaining high recognition accuracy. This lightweight model is suitable for embedded GPU deployment, significantly reducing hardware costs and energy consumption. In practical applications, embedded devices have limited computing resources, making traditional large-scale CNN models inefficient. However, this optimized lightweight model can achieve real-time performance with a single-image inference latency of no more than 20ms on embedded GPUs, meeting the stringent real-time requirements of photovoltaic inspections. For RNN time series analysis tasks, the gating mechanism optimization of the Long Short-Term Memory (LSTM) network effectively mitigates the vanishing gradient problem. Furthermore, the parallel processing of multi-dimensional environmental data using a time series prediction model significantly improves the speed of state analysis. When analyzing the operating status of photovoltaic modules, it can quickly and accurately capture the changing trends of temperature, current, environment and other data, discover potential faults in advance, and provide a basis for timely maintenance.

[0041] Furthermore, dynamic allocation of computing resources through a load balancing algorithm prevents overloading of a single node and improves the system's overall processing capacity. A large-scale photovoltaic power plant generates enormous amounts of data daily. Relying solely on a single computing node would result in slow processing and fail to meet real-time requirements. By adopting a distributed computing architecture, the system's throughput has increased by more than three times, enabling the processing and analysis of large amounts of data in a short period of time, ensuring efficient photovoltaic inspections.

[0042] As a possible implementation method, the method of acquiring photovoltaic power station layout data, combining analysis results, environmental data, and time synchronization information, and generating an inspection path through a dynamic inspection priority queue and a dual-mode path planning algorithm includes: Determine the dynamic inspection priority queue based on the defect level and environmental risk level. The priority of each PV module is generated based on the defect level, environmental risk level and corresponding weights, with the defect level weighting being greater than the environmental risk level weighting. Set up a dual-mode path planning engine, where: In normal mode, an improved ant colony algorithm is used to combine the distribution of defect hotspots with future weather forecast data to construct a global path. In this algorithm, each ant makes decisions based on a heuristic function and pheromone intensity during the path selection process, and periodically and automatically re-evaluates the global path to ensure that all potential high-risk areas are covered. When at least one of the defect level or environmental risk level is at the highest level, the system switches to emergency mode. In emergency mode, a cost function is constructed based on real-time environmental data and component location information. The comprehensive cost of the current path node is calculated based on the actual cost from the starting node to the current node and the heuristic cost from the current node to the target node. Node expansion is performed and path planning is completed with the goal of minimizing the comprehensive cost.

[0043] In this embodiment, the path optimization module is one of the keys to achieving efficient inspections and risk avoidance. It integrates a dynamic inspection priority queue and a dual-mode path planning engine, and relies on the output data of the deep learning processing module and the data fusion processing module to achieve dynamic scheduling. The dynamic inspection priority queue (data priority queue) first evaluates the weight of the photovoltaic components. This evaluation constructs a comprehensive priority index based on the defect level and the environmental risk level. Emergency defect information is 5 priority levels higher than conventional defects. At the same time, if the environmental risk level is "dangerous", the inspection weight of the components in the corresponding area is further increased. Comprehensive priority weight Expressed as: ; in, represents the defect level code of the i-th component (5 for urgent defects and 1 for regular defects), Indicates the environmental risk level code (3 for danger, 2 for warning, and 1 for safety). and For the corresponding weight coefficient, the system defaults to , , ensuring that defects influence the dominant path sorting. During the path planning process, the normal mode uses an improved ant colony algorithm combined with the defect hotspot distribution and the weather forecast data for the next 72 hours to construct a global path. In this algorithm, each ant makes a decision based on the heuristic function and pheromone intensity during the path selection process. Its probability function is defined as: ; in, is the transfer probability of the kth ant from node i to node j, represents the path pheromone intensity, is the heuristic function value based on defect density and weather impact, and α and β control the weights of the two. This mode automatically re-evaluates the global path every 30 seconds to ensure that all potential high-risk areas are covered. In special circumstances, such as when the real-time wind speed exceeds level 10 or an emergency defect is detected, the system immediately switches to emergency mode and uses A The algorithm generates an obstacle avoidance path. This path is constructed based on the cost function of real-time environment data and component position information: ; in, is the comprehensive cost of the current path node n, is the actual cost from the starting node to the current node, Estimation of the heuristic cost from the current node to the target node, path planning to minimize Node expansion is performed for the target. The path optimization module connects with the data fusion processing module to obtain environmental parameters such as wind speed and humidity with high-precision time stamps. This ensures that the environmental data used in path planning and the component status data are strictly consistent in time, ultimately achieving a simultaneous improvement in inspection efficiency and system responsiveness.

[0044] This embodiment assigns the highest weight to photovoltaic module data marked as "urgent defects", such as cracks and overheating, which can trigger instant path planning. This means that when an emergency defect is detected in a photovoltaic module, the robot can quickly adjust the inspection path and reach the location of the problem module for inspection and maintenance as soon as possible, effectively reducing the loss of power generation due to the failure to handle the fault in a timely manner. For example, in the actual operation of a photovoltaic power station, when an emergency defect of overheating is detected in a photovoltaic module, the system plans a path to the module within 1 minute through a priority queue mechanism, allowing the robot to handle it in time and avoid further damage to the module. The batch processing mode is used for routine inspection data, which balances real-time performance and computing efficiency, ensuring that while ensuring the timely handling of emergency defects, routine inspection tasks can also be completed efficiently.

[0045] In this embodiment, the normal mode and emergency mode dynamically adjust the path in response to different situations, and rely on low-latency data with high-precision timestamps to ensure the spatiotemporal consistency of path planning, prioritize inspections of high-risk components, significantly enhance preventive maintenance capabilities, and reduce the robot's ineffective movement distance.

[0046] For example, in normal mode, a globally optimal path is generated based on an improved ant colony algorithm, combined with historical data and predicted environmental parameters. This approach fully leverages historical data and environmental prediction information, enabling the robot to conduct inspections along the optimal path under normal circumstances, improving inspection efficiency and coverage. For example, based on historical defect distribution and weather forecast data for the next 72 hours, the robot can plan a path in advance that avoids potential high-risk areas while ensuring a comprehensive inspection of all components. When the deep learning module detects a sudden defect or real-time environmental parameters exceed a safety threshold, the system immediately switches to emergency mode and uses the Dijkstra algorithm to generate a local obstacle avoidance path to ensure the robot's safety. In the event of sudden severe weather, the robot can quickly adjust its path to avoid being affected by harsh environmental conditions such as strong winds while continuing to complete its inspection mission, ensuring the stability and adaptability of the system.

[0047] As a possible implementation manner, the step of obtaining the defect level includes: Obtain the defect type and confidence level obtained by defect identification, and determine the defect level based on the confidence level.

[0048] For example, the image recognition unit uses a lightweight CNN model (such as MobileNet-V3 based on knowledge distillation) to infer PV module images and output defect types (such as cracks, dust accumulation, breakage, and overheating) and corresponding confidence values ​​C (range: 0 ≤ C ≤ 1). When the confidence value C ≥ 0.95, the defect level is 5; when 0.8 ≤ C < 0.95, the defect level is 3; and in other cases, the defect level is 1. It should be noted that the threshold value can be other values, and more levels can be divided.

[0049] As a possible implementation method, if the generation of the inspection path is blocked, an alternative inspection path is generated with reference to prior knowledge, including: pre-selection through a minimum path cost function based on a defect distribution density function, in which the higher the defect density, the greater the path cost, thereby guiding path planning to avoid or prioritize coverage of high-risk areas.

[0050] For example, preselection is performed by using the minimum path cost criterion based on the defect distribution density function. The path cost function is defined as: ; in, is the total cost of path p, is the Euclidean distance from the i-th node to the i+1-th node in the path, is the historical defect density of the area corresponding to the node, γ is the defect penalty coefficient. The higher the defect density, the greater the path cost, which guides path planning to avoid or prioritize covering high-risk areas.

[0051] In this embodiment, the communication module, a key connection unit of the intelligent operation and maintenance system, integrates an efficient caching mechanism and a low-latency transmission channel. This ensures the system's overall real-time and robustness, while enabling high-speed data forwarding and autonomous recovery in the event of a network outage. The module's built-in data cache, with a capacity of ≥2GB, is preloaded with PV plant layout grid data, a historical defect coordinate library, and an optimal path template generated through ant colony optimization. This allows the path optimization module to generate alternative inspection routes based on prior knowledge in the event of a network outage.

[0052] The hardware acceleration interface connects to the dedicated AI chip of the robot body module and uses a dedicated bus channel to achieve low-latency communication in image recognition output and path instruction transmission. The end-to-end latency is strictly controlled within 1ms. The data throughput rate T is expressed as: ; Where S is the amount of data transmitted in a single transmission, in bytes, and Δt is the total transmission time, in seconds. This ensures that the real-time delivery of single-frame image results and path data instructions meets the accuracy requirements of dynamic inspections. Wireless communication uses a dual-link collaboration mechanism of Wi-Fi 6 and 4G LTE. In-station communication uses a Wi-Fi module to achieve low-latency instruction synchronization of ≤10ms. Remote communication uses a 4G module to complete data backhaul and status reporting with the cloud platform. The AES-256 high-strength encryption algorithm is used in transmission, and its encryption function is expressed as: ; Among them, C is the ciphertext data, P is the original plaintext data, Represents an AES encryption operation using a 256-bit key, k. This encryption strategy ensures data privacy and integrity during public network transmission, preventing the risk of leaking critical computational results during operations and maintenance, and building a secure, low-latency, and breakpoint-tolerant communication assurance system.

[0053] As a possible implementation method, generating an operation and maintenance task queue based on the analysis results and storing full-process data at the same time include: An operation and maintenance task queue is generated based on the defect level and processing time. The priority of each operation and maintenance task is generated based on the defect level, processing time and corresponding weights, and the weight of the defect level is greater than the weight of the processing time.

[0054] For example, the task scheduling unit of the operation and maintenance management module generates an operation and maintenance task queue based on the time synchronization data of the data fusion processing module and the priority path of the path optimization module. The priority of task scheduling is composed of "defect level + processing time", where the defect level is derived from the image recognition results of the deep learning processing module, and the processing time is dynamically adjusted according to the urgency of the task. The response delay of urgent tasks must be controlled within 1 minute to ensure a rapid response to system failures or equipment anomalies. The task queue generation mechanism evaluates the priority of tasks using the following cost function: ; in, is the priority of the i-th task, is the defect level corresponding to the task, is the processing time of the task, λ1 and λ2 are the weight coefficients of the defect level and processing time, and the system defaults to λ1=0.7 and λ2=0.3. This mechanism ensures that tasks with urgent defects are given priority. The task scheduling system continuously updates the task queue and adjusts the priority weights to ensure that high-priority tasks can be completed in the shortest time. The data storage unit uses a distributed time series database (such as InfluxDB) to store multi-dimensional sensor data and sets different storage periods according to the life cycle of the data. The raw sensor data is stored in time series with a storage period of 1 year. The storage period of feature data after preprocessing is 3 years, and the operation and maintenance logs are stored permanently. The storage and query of time series data are implemented through efficient indexing mechanisms and query algorithms, supporting multi-dimensional data retrieval with second-level accuracy, ensuring data access efficiency and accuracy. The query cost function of data retrieval can be expressed as: ; in, The cost of the query, is the weight of the i-th query field, The system dynamically evaluates the retrieval cost based on the query field and data volume. The human-computer interaction unit provides a real-time visualization interface, displaying the data flow status of each module, including key metrics such as data throughput, processing latency, and synchronization accuracy. Users can manually adjust the edge computing unit's ROI extraction threshold and priority queue weight parameters through this interface, optimizing system performance and responsiveness based on on-site operation and maintenance requirements. This interface supports flexible scheduling of tasks and parameters, adapting to different working environments and operating conditions.

[0055] As an implementation method, this example also implements hardware-software collaborative optimization, including heterogeneous computing acceleration: integrating a dedicated AI chip (such as a TPU / NPU) within the robot's main module to achieve hardware acceleration for image recognition and path optimization algorithms, reducing data transmission delays between modules. A cache and prefetching strategy: setting up a data cache in the communication module to pre-fetch PV plant layout data and historical defect distribution patterns. This enables the path optimization module to generate candidate paths based on prior knowledge even without real-time data input, improving response speed.

[0056] This embodiment also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor calls the computer program in the memory, the steps of the above-mentioned photovoltaic inspection robot intelligent operation and maintenance method are implemented.

[0057] This embodiment also provides a storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are loaded and executed by a processor, the steps of the above-mentioned photovoltaic inspection robot intelligent operation and maintenance method are implemented.

[0058] Through the description of the above implementation methods, technical personnel in the relevant field can understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional modules as needed, that is, the internal structure of the specific device can be divided into different functional modules to complete all or part of the functions described above.

[0059] In the embodiments provided in this application, it should be understood that the disclosed structures and methods can be implemented in other ways. For example, the embodiments of the structure described above are merely illustrative. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another structure, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, structure or unit, which can be electrical, mechanical or other forms.

[0060] Units described as separate components may or may not be physically separate, and components shown as units may be one physical unit or multiple physical units, that is, they may be located in one place or distributed in multiple places. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0061] In addition, the functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware or software functional units.

[0062] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a device (which can be a single-chip microcomputer, chip, etc.) or a processor (processor) to execute all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0063] The above content is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. Photovoltaic inspection robot intelligent operation and maintenance method, characterized in that: include: Collect image data, temperature data, and environmental data of photovoltaic modules in photovoltaic power plants, and perform differential extraction to retain only local key information; Through standardized data interfaces, the protocols of multiple types of key information in the data acquisition module are unified and timestamps are aligned to obtain standardized data. Conduct defect identification and operation status analysis on standardized data to obtain analysis results; Obtain PV power plant layout data, combine analysis results, environmental data, and time synchronization information, and generate inspection routes using dynamic inspection priority queues and a dual-mode path planning algorithm; Generate an operation and maintenance task queue based on the analysis results and store the entire process data; Perform movement and maintenance operations on the robot body according to the inspection path and operation and maintenance task queue.

2. The photovoltaic inspection robot intelligent operation and maintenance method according to claim 1 is characterized in that: The image data, temperature data of photovoltaic modules in the photovoltaic power station and environmental data of the photovoltaic power station are collected and differentially extracted to retain only local key information, including: By using the feature significance scoring function, the region with higher weight is selected for ROI extraction, and only the local image containing potential defects is retained to focus on the potential defect area in the image data. At the same time, the sliding window filtering algorithm is used to extract the temperature and environmental data.

3. The photovoltaic inspection robot intelligent operation and maintenance method according to claim 1 is characterized in that: The method of performing protocol unification and timestamp alignment on multiple types of key information of the data acquisition module through the standardized data interface to obtain standardized data includes: A unified data structure with a pre-defined format is defined for image, temperature, and environmental data. The structure includes a data type label, acquisition device ID, raw value, and quality factor, where the quality factor reflects the integrity and credibility of the current data. It is connected to a precision clock synchronization system and uses the difference between the master clock broadcast signal and the local timing of the slave node to dynamically calibrate the sampling time to ensure that the time error of all sampling nodes in the system is controlled within the preset error. Each frame of data is bound to a unique timestamp when it is generated. By minimizing the time alignment cost function between different data sequences, the distance matrix is ​​traversed through dynamic programming to backtrack the optimal alignment path. After the alignment is completed, the result is constrained with the target error window to ensure the consistency of multi-source data in the time dimension.

4. The photovoltaic inspection robot intelligent operation and maintenance method according to claim 1 is characterized in that: The defect identification and operation status analysis of the standardized data are performed to obtain analysis results, including: A lightweight convolutional neural network model based on knowledge distillation optimization is used to identify defects in photovoltaic module image data. It uses a fusion structure of gated recurrent units and Transformer encoders to extract long-term dependency information and assigns dynamic weights to different feature dimensions through a multi-head attention mechanism to perform operational status analysis. Among them, defect identification and operation status analysis are based on a distributed computing architecture. Image recognition tasks and status analysis tasks are processed by independent computing nodes respectively. Task allocation between nodes is dynamically adjusted in real time through a load balancing algorithm. The load balancing algorithm improves processing efficiency by minimizing the overall task time.

5. The photovoltaic inspection robot intelligent operation and maintenance method according to claim 1 is characterized in that: The method of obtaining photovoltaic power station layout data, combining analysis results, environmental data and time synchronization information, and generating an inspection path through a dynamic inspection priority queue and a dual-mode path planning algorithm includes: Determine the dynamic inspection priority queue based on the defect level and environmental risk level. The priority of each PV module is generated based on the defect level, environmental risk level and corresponding weights, with the defect level weighting being greater than the environmental risk level weighting. Set up a dual-mode path planning engine, where: In normal mode, an improved ant colony algorithm is used to combine the distribution of defect hotspots with future weather forecast data to construct a global path. In this algorithm, each ant makes decisions based on a heuristic function and pheromone intensity during the path selection process, and periodically and automatically re-evaluates the global path to ensure that all potential high-risk areas are covered. When at least one of the defect level or environmental risk level is the highest level, it switches to emergency mode. In emergency mode, a cost function is constructed based on real-time environmental data and component location information. The comprehensive cost of the current path node is calculated based on the actual cost from the starting node to the current node and the heuristic cost from the current node to the target node. Node expansion is performed and path planning is completed with the goal of minimizing the comprehensive cost.

6. The photovoltaic inspection robot intelligent operation and maintenance method according to claim 5 is characterized in that: The step of obtaining the defect level includes: Obtain the defect type and confidence level obtained by defect identification, and determine the defect level based on the confidence level.

7. The photovoltaic inspection robot intelligent operation and maintenance method according to claim 1 is characterized in that: If the generation of the inspection path is blocked, an alternative inspection path is generated with reference to prior knowledge, including: pre-selection through the minimum path cost function based on the defect distribution density function. In the path cost function, the higher the defect density, the greater the path cost, thereby guiding path planning to avoid or prioritize covering high-risk areas.

8. The photovoltaic inspection robot intelligent operation and maintenance method according to claim 1 is characterized in that: The operation and maintenance task queue is generated based on the analysis results, and the whole process data is stored at the same time, including: An operation and maintenance task queue is generated based on the defect level and processing time. The priority of each operation and maintenance task is generated based on the defect level, processing time and corresponding weights, and the weight of the defect level is greater than the weight of the processing time.

9. A photovoltaic inspection robot intelligent operation and maintenance system, configured to execute the photovoltaic inspection robot intelligent operation and maintenance method according to any one of claims 1 to 8, characterized in that: include: A data acquisition module is used to collect image data, temperature data of photovoltaic modules in a photovoltaic power station, and environmental data of the photovoltaic power station; The data fusion processing module is used to unify the protocols and align the timestamps of multiple types of key information from the data acquisition module through a standardized data interface to obtain standardized data; Deep learning processing module, used to perform defect identification and operation status analysis on standardized data to obtain analysis results; The path optimization module is used to obtain PV power plant layout data, combine analysis results, environmental data and time synchronization information, and generate inspection paths through dynamic inspection priority queues and a dual-mode path planning algorithm; The operation and maintenance management module is used to generate an operation and maintenance task queue based on the analysis results and store the entire process data; The robot body is used to perform movement and operation and maintenance operations according to the inspection path and operation and maintenance task queue.

10. An electronic device, characterized in that: It includes a memory and a processor, wherein the memory stores a computer program, and when the processor calls the computer program in the memory, it implements the steps of the photovoltaic inspection robot intelligent operation and maintenance method according to any one of claims 1 to 8.

Images