Multi-machine cooperative autonomous inspection system suitable for mountainous area photovoltaic power station

Abstract

The invention relates to the technical field of new energy electric power, in particular to a multi-machine cooperative autonomous inspection system suitable for a mountainous area photovoltaic power station, which comprises a task and decision-making layer, a cooperative sensing layer and a field execution layer which are communicated in sequence. The field execution layer executes an inspection action based on the dynamic task instruction and collects original inspection data; the collaborative sensing layer receives the original data, performs fusion and positioning calculation in combination with preset artificial anchor spatial distribution information and photovoltaic array geometric prior information, and generates sensing data containing positioning information and a target state; and the task and decision-making layer generates a dynamic task instruction comprising an inspection path and / or a detection task according to the sensing data in combination with an environment map comprising digital elevation model information, a road network, a photovoltaic array vector and communication sight distance information. According to the system, closed-loop control is realized through three-layer cooperation, the adaptive capability and the operation efficiency of photovoltaic inspection are improved, and the inspection precision and the response speed are optimized.

Description

Technical Field

[0001] This invention relates to the field of new energy power technology, specifically to a multi-machine collaborative autonomous inspection system suitable for photovoltaic power stations in mountainous areas. Background Technology

[0002] With the introduction of the "dual carbon" target, photovoltaic power generation, as an important component of clean energy, is rapidly expanding its application in non-flat areas such as mountains and hills. Compared with plains, photovoltaic power stations in mountainous areas are often more widely distributed, with complex terrain conditions. The arrangement of modules is more constrained by the terrain, and the inspection tasks not only include the detection of surface defects and hot spots on the modules, but also involve various objects such as combiner boxes, brackets, cables, and ground access roads. Therefore, the efficiency and accuracy of inspection results directly affect the long-term stable operation of the power station and the control of operation and maintenance costs.

[0003] Currently, photovoltaic power plant inspections primarily rely on manual labor and single-unit drones. Manual inspections require maintenance personnel to walk for extended periods within the power plant area, facing high temperatures, intense radiation, and poor road conditions, resulting in high workload and safety hazards. While single-unit drones improve inspection efficiency, limitations in flight time, payload, and data processing capabilities often restrict coverage to a limited area, failing to meet the needs of large-scale inspections in complex terrain. Furthermore, existing drone inspections mostly employ a single-unit operation mode, lacking inter-platform collaboration, leading to data redundancy, blind spots, and unstable communication links.

[0004] In mountainous environments, inspection tasks face challenges such as blocked satellite positioning signals, communication interruptions caused by terrain obstruction, and sudden changes in weather conditions. Existing inspection methods struggle to guarantee continuous coverage and robust positioning, and lack the ability to combine real-time screening of collected data with remote in-depth diagnostics, resulting in untimely anomaly detection and high false alarm and false negative rates. Therefore, there is an urgent need for an inspection system that integrates multiple platforms (air and ground), enables multimodal perception, supports intelligent scheduling and edge-cloud collaborative diagnostics, and addresses the technical problems of low inspection efficiency, poor communication stability, and insufficient diagnostic accuracy in mountainous photovoltaic power stations operating in complex environments. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a multi-machine collaborative autonomous inspection system suitable for photovoltaic power stations in mountainous areas, which addresses the shortcomings of the prior art. This system solves the technical problems that existing inspection methods cannot guarantee the continuity of large-area coverage and the robustness of positioning, and also lack the ability to combine real-time screening and remote in-depth diagnosis of collected data.

[0006] The objective of this invention is achieved through the following technical solutions: In a first aspect, the present invention provides a multi-machine collaborative autonomous inspection system suitable for photovoltaic power stations in mountainous areas, comprising a task and decision layer, a collaborative perception layer, and a field execution layer that communicate sequentially. The field execution layer is used to execute corresponding inspection actions based on dynamic task instructions and collect raw inspection data. The collaborative perception layer is used to receive the raw inspection data collected by the field execution layer, and based on the preset spatial distribution information of artificial anchors and the geometric prior information of the photovoltaic array, to fuse and locate the raw inspection data to obtain perception data containing location information and target status. The task and decision layer is used to generate dynamic task instructions containing inspection paths and / or detection tasks based on the perception data output by the collaborative perception layer and in combination with a pre-constructed environmental map; the environmental map includes digital elevation model information, road network, photovoltaic array vector and communication line-of-sight information.

[0007] As a further improvement of the present invention, the field execution layer includes fixed-wing UAVs, multi-rotor UAVs, and ground unmanned vehicles; The fixed-wing UAV is used to respond to the coverage inspection command in the dynamic mission command, perform a cruise within a set mountain range and collect the first raw data; The multi-rotor UAV is used to respond to the fine inspection command or communication relay command in the dynamic mission command, perform low-altitude hovering or fine inspection flight and collect second raw data. The unmanned ground vehicle is used to respond to the near-ground inspection command in the dynamic task command, travel along the inspection road and collect third raw data.

[0008] As a further improvement of the present invention, the field execution layer also includes a fixed base, which is installed in the inverter room and key combiner box in the mountain ridge, and includes a broadband positioning base station, an RTK differential signal relay unit, a millimeter wave communication unit and a fast charging interface. The broadband positioning base station is used to provide relative positioning signals for fixed-wing drones, multi-rotor drones or ground unmanned vehicles that enter the positioning coverage area. The RTK differential signal relay unit is used to receive and forward GNSS differential correction signals to correct the absolute positioning accuracy of the fixed-wing UAV, multi-rotor UAV or ground unmanned vehicle. The millimeter-wave communication unit is used to establish high-speed data transmission links between fixed-wing UAVs, multi-rotor UAVs, or ground unmanned vehicles, or between different fixed bases. The fast charging interface is used to connect the fixed-wing UAV, multi-rotor UAV, or ground unmanned vehicle.

[0009] As a further improvement of the present invention, the task and decision layer includes a graph construction module and a dynamic scheduling module; The map construction module is used to construct and update a dynamic environmental map reflecting terrain, equipment topology and communication links based on the environmental map and the positioning information in the sensing data. The dynamic scheduling module is used to periodically calculate and output the dynamic task instructions, taking the dynamic environment map as input and taking task completion efficiency, equipment energy consumption and operation safety as joint optimization objectives.

[0010] As a further improvement of the present invention, the constraints of the joint optimization objective include: the remaining energy of the fixed-wing UAV and the multi-rotor UAV is not lower than the energy threshold, the signal-to-noise ratio of the communication link is not lower than the communication threshold, the driving slope of the ground unmanned vehicle does not exceed the slope threshold, and the minimum safe distance is maintained between fixed-wing UAVs or between multi-rotor UAVs.

[0011] As a further improvement of the present invention, the dynamic scheduling module adopts a rolling time-domain optimization method, performing optimization calculations once at a set time interval.

[0012] As a further improvement of the present invention, it also includes an edge-cloud collaboration layer, which includes an edge computing module and a cloud analytics module; The edge computing module is used to receive the sensing data, analyze the sensing data to obtain edge screening results including abnormal suspected areas and preliminary categories, and feed the edge screening results back to the task and decision layer. The cloud analysis module is used to receive the edge screening results uploaded by the edge computing module, and perform parameter inversion and deep diagnosis based on the physical model to generate cloud analysis results containing anomaly scores and diagnostic reports; the cloud analysis results are used to correct the optimization strategies of the task and decision layers and the analysis model of the edge computing module.

[0013] As a further improvement of the present invention, the edge computing module includes: hot spot screening based on infrared thermal imaging, crack detection based on visible light images, and vegetation occlusion identification based on near-infrared images.

[0014] As a further improvement of the present invention, the anomaly score generated in the cloud analysis module is a value between 0 and 1; when the anomaly score is greater than or equal to a first threshold, a detailed inspection work order is triggered based on the cloud analysis result; when the anomaly score is between a second threshold and the first threshold, the cloud analysis result triggers a review mark; the first threshold is greater than the second threshold.

[0015] As a further improvement of the present invention, the artificial anchors are deployed at intervals of 80 to 150 meters, and their absolute coordinates are obtained by RTK measurement; when the collaborative sensing layer calculates the pose of the multi-mode sensor array, it matches the image coordinates of the identified artificial anchors with their absolute coordinates to correct the positioning error caused by the obstruction of satellite signals.

[0016] Secondly, this invention provides a multi-machine collaborative autonomous inspection method suitable for photovoltaic power stations in mountainous areas, applied to a multi-machine collaborative autonomous inspection system suitable for photovoltaic power stations in mountainous areas, the method comprising: Based on the initial environment map, initial task allocation and global path planning are performed for multi-machine collaboration, and initial scheduling instructions are generated and sent to the field execution layer; Each device in the field execution layer moves and collects raw inspection data according to the scheduling instructions. Based on the collected raw inspection data, combined with the preset spatial distribution information of artificial anchors and the geometric prior information of the photovoltaic array, the raw inspection data is fused and located to obtain perception data containing positioning information and target status. Based on the perceived data and combined with the pre-constructed environmental map, dynamic task instructions containing inspection paths and / or detection tasks are generated; the environmental map includes digital elevation model information, road network, photovoltaic array vectors, and communication line-of-sight information.

[0017] The beneficial effects of this invention are as follows: This invention provides a multi-machine collaborative autonomous inspection system suitable for photovoltaic power stations in mountainous areas. The system is robust. The field execution layer executes inspection actions and collects raw inspection data based on dynamic task instructions, achieving a direct correspondence between inspection actions and task instructions, ensuring the targeted and efficient data collection. After receiving the raw inspection data, the collaborative perception layer combines the preset spatial distribution information of artificial anchors with the prior geometric information of the photovoltaic array for fusion and positioning calculation, generating perception data containing positioning information and target status. Through multi-source information fusion, the accuracy and reliability of positioning calculation are improved, avoiding positioning errors caused by single-source data. Positional deviation; based on the sensing data output by the collaborative sensing layer, the task and decision layers combine the pre-constructed environmental map containing digital elevation model information, road network, photovoltaic array vector, and communication line-of-sight information to generate dynamic task instructions, realizing a deep integration of environmental information and sensing data, and optimizing the decision logic of inspection path planning and detection task allocation; the above three layers work together to form a closed-loop control structure, and through the generation, execution, feedback, and re-optimization mechanism of dynamic task instructions, significantly improve the adaptive capability and overall operating efficiency of the photovoltaic inspection system, achieving a dual improvement in inspection accuracy and response speed compared to the traditional fixed-path inspection method. Attached Figure Description

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

[0019] Figure 1 This is a schematic diagram of a multi-machine collaborative autonomous inspection system for photovoltaic power stations in mountainous areas, according to an embodiment of the present invention. Figure 2 This is a schematic diagram of a task and decision optimization process according to an embodiment of the present invention; Figure 3 This is a schematic diagram of a component anomaly diagnosis process according to an embodiment of the present invention; Figure 4 This is an internal structural diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives and technical solutions of this invention clearer and easier to understand, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0021] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and specific embodiments. The described embodiments are only some embodiments of the present invention, and not all embodiments.

[0022] Example 1 This embodiment provides a multi-machine collaborative autonomous inspection system suitable for photovoltaic power stations in mountainous areas, including a task and decision layer, a collaborative perception layer, and a field execution layer that communicate sequentially.

[0023] The field execution layer is used to execute corresponding inspection actions based on dynamic task instructions and collect raw inspection data. The collaborative perception layer is used to receive the raw inspection data collected by the field execution layer and, based on the preset spatial distribution information of artificial anchors and the geometric prior information of the photovoltaic array, fuse and perform positioning calculations on the raw inspection data to obtain perception data containing positioning information and target status. The task and decision layer is used to generate dynamic task instructions containing inspection paths and / or detection tasks based on the perception data output by the collaborative perception layer and in combination with the pre-constructed environmental map. The environmental map includes digital elevation model information, road network, photovoltaic array vector, and communication line-of-sight information.

[0024] The working principle of this embodiment is as follows: through sequential communication and coordination among the task and decision-making layer, the collaborative perception layer, and the field execution layer, the field execution layer executes dynamic task commands and collects raw inspection data. The collaborative perception layer fuses and calculates the location of the raw inspection data based on the preset spatial distribution information of artificial anchors and the geometric prior information of the photovoltaic array, obtaining perception data containing location information and target status. The task and decision-making layer combines a pre-constructed environmental map containing digital elevation model information, road network, photovoltaic array vector, and communication line-of-sight information, and generates dynamic task commands based on the perception data. The corresponding technical effect is that the field execution layer executes dynamic task commands and collects raw inspection data, achieving precise execution of inspection actions. The system effectively acquires raw data; the collaborative sensing layer performs data fusion and positioning calculation based on the spatial distribution information of preset artificial anchors and the geometric prior information of the photovoltaic array, improving the accuracy and positioning precision of raw inspection data processing and ensuring the reliability of positioning information and target status in the sensing data; the task and decision layer combines environmental maps containing multiple key information to generate dynamic task instructions, ensuring the scientific nature and relevance of dynamic task instructions, making inspection path planning and detection task allocation more in line with actual inspection needs; the three layers work together to achieve closed-loop linkage of inspection data collection, processing, and task instruction generation, significantly improving the overall efficiency and intelligence level of photovoltaic inspection, and ensuring the continuity and efficiency of the entire inspection process.

[0025] The on-site execution layer includes fixed-wing UAVs, multi-rotor UAVs, and ground unmanned vehicles. The fixed-wing UAVs are used to respond to the coverage inspection command in the dynamic mission instructions, perform a cruise within a set mountainous area, and collect first raw data. The multi-rotor UAVs are used to respond to the fine inspection command or communication relay command in the dynamic mission instructions, perform low-altitude hovering or fine inspection flight, and collect second raw data. The ground unmanned vehicles are used to respond to the near-ground inspection command in the dynamic mission instructions, travel along the inspection road, and collect third raw data.

[0026] The field execution layer also includes fixed bases, which are installed in inverter rooms and key combiner boxes located on mountain ridges. These bases include a broadband positioning base station, an RTK differential signal relay unit, a millimeter-wave communication unit, and a fast-charging interface. The broadband positioning base station provides relative positioning signals for fixed-wing UAVs, multi-rotor UAVs, or unmanned ground vehicles entering the positioning coverage area. The RTK differential signal relay unit receives and forwards GNSS differential correction signals to correct the absolute positioning accuracy of fixed-wing UAVs, multi-rotor UAVs, or unmanned ground vehicles. The millimeter-wave communication unit establishes high-speed data transmission links between fixed-wing UAVs, multi-rotor UAVs, or unmanned ground vehicles, or between different fixed bases. The fast-charging interface connects fixed-wing UAVs, multi-rotor UAVs, or unmanned ground vehicles.

[0027] The task and decision-making layer includes a map construction module and a dynamic scheduling module. The map construction module is used to construct and update a dynamic environmental map reflecting terrain, equipment topology, and communication links based on the environmental map and location information in the sensing data. The dynamic scheduling module is used to periodically calculate and output dynamic task instructions with the dynamic environmental map as input and task completion efficiency, equipment energy consumption, and operational safety as joint optimization objectives.

[0028] The constraints of the joint optimization objective include: the remaining energy of fixed-wing UAVs and multi-rotor UAVs is not lower than the energy threshold, the signal-to-noise ratio of the communication link is not lower than the communication threshold, the driving slope of the ground unmanned vehicle does not exceed the slope threshold, and a minimum safe distance is maintained between fixed-wing UAVs or between multi-rotor UAVs.

[0029] The dynamic scheduling module uses a rolling time-domain optimization method, performing optimization calculations once at set time intervals.

[0030] It also includes an edge-cloud collaboration layer, which comprises an edge computing module and a cloud analytics module. The edge computing module receives and analyzes the perceived data to obtain edge screening results containing suspected abnormal areas and preliminary categories, and feeds the edge screening results back to the task and decision-making layer. The cloud analytics module receives the edge screening results uploaded by the edge computing module, performs parameter inversion and deep diagnosis based on a physical model, and generates cloud analytics results containing anomaly scores and diagnostic reports. The cloud analytics results are used to correct the optimization strategies of the task and decision-making layer and the analysis model of the edge computing module.

[0031] The edge computing module includes: initial hot spot screening based on infrared thermal imaging, crack detection based on visible light images, and vegetation occlusion identification based on near-infrared images. The anomaly score generated in the cloud analysis module is a value between 0 and 1; when the anomaly score is greater than or equal to a first threshold, a detailed inspection work order is triggered based on the cloud analysis results; when the anomaly score is between a second threshold and the first threshold, a review mark is triggered by the cloud analysis results; the first threshold is greater than the second threshold.

[0032] Example 2 like Figure 1As shown, this invention provides a multi-machine collaborative autonomous inspection system suitable for photovoltaic power stations in mountainous areas. The system consists of a field layer, a collaborative perception layer, a task and decision-making layer, and an edge-cloud collaborative layer. The field layer includes fixed-wing UAVs, multi-rotor UAVs, ground-based unmanned vehicles, and several fixed bases, used for long-endurance coverage inspections, low-altitude fine-grained inspections, ground-based short-range detection, and communication relay under complex terrain conditions. The collaborative perception layer consists of airborne multi-mode sensors, sparse artificial anchors, and geological disaster sentinels, forming a multi-source information acquisition and auxiliary positioning system. The task and decision-making layer constructs a terrain-topology-communication map based on a digital elevation model, road network, photovoltaic array vectors, and communication line-of-sight matrix. It generates multi-machine collaborative inspection paths and relay deployment schemes through a joint optimization algorithm and performs online replanning in the event of link fading or weather disturbances. The edge-cloud collaboration layer includes an edge computing module and a cloud analysis module. The edge computing module completes thermal image screening, crack detection, occlusion identification and location fusion. The cloud analysis module performs non-contact electrical parameter inversion based on the thermo-electric-mechanical coupling model, outputs virtual current and voltage curve parameter ranges and generates alarm work orders, thereby realizing full coverage inspection, stable communication, accurate diagnosis and closed-loop management of photovoltaic power stations in mountainous areas.

[0033] Based on the aforementioned overall architecture, to better reflect the system's layered design and functional synergy, this invention further details the composition and function of each layer. Specifically, the multi-machine collaborative autonomous inspection system of this invention can be functionally divided into a field layer, a collaborative perception layer, a task and decision-making layer, and an edge-cloud collaborative layer. Each layer is relatively independent yet closely interconnected, forming bottom-up execution and perception support, as well as top-down decision-making and diagnostic feedback. The specific structure and implementation of each layer are further elaborated below with reference to the accompanying drawings.

[0034] The field layer is the direct execution unit for the inspection tasks of the system in the mountainous photovoltaic power station. It mainly consists of fixed-wing UAVs, multi-rotor UAVs, ground unmanned vehicles and fixed bases. The modules cooperate with each other to realize functions such as coverage inspection, fine inspection, back-side detection and communication relay.

[0035] Fixed-wing UAVs primarily undertake large-area, long-endurance coverage inspection tasks. Their structure utilizes a lightweight composite material fuselage, with a wingspan ranging from 1.8 to 2.4 meters, a cruising speed controlled between 18 and 24 meters per second, and a full-load endurance of no less than 90 minutes. The fixed-wing UAVs are equipped with a mid-wave infrared imaging unit (resolution 640×512@30 Hz) and a high-resolution visible light sensor for large-area, rapid scanning to acquire surface temperature distribution and image information of photovoltaic modules.

[0036] In mountainous environments, fixed-wing UAVs use onboard 3D wind sensors to measure wind speed and turbulence intensity in real time, and combine this with a nonlinear model predictive controller (NMPC) in the flight control system to dynamically correct the flight path. When updrafts are detected on a ridge, the flight control system automatically adjusts the angle of attack and speed to recover some energy, thereby reducing energy consumption per flight path.

[0037] Multi-rotor drones primarily perform low-altitude precision inspection and communication relay tasks. They have a takeoff weight of less than 6 kg, employ quadcopter or hexcopter configurations, and have a single flight time of at least 35 minutes. Equipped with a three-axis stabilized gimbal, a 45-megapixel optical camera, and a 12-bit near-infrared imaging module, these drones fly at an altitude of 6 to 10 meters above photovoltaic modules, maintaining an imaging angle of 35 to 50 degrees, enabling them to acquire high-resolution images for detailed inspection of cracks, obstructions, hot spots, and other defects.

[0038] In communication relay scenarios, multi-rotor UAVs can hover at locations designated by the mission planning layer, serving as mobile relay nodes. When a decrease in the communication signal-to-noise ratio or obstructed line-of-sight is detected, the multi-rotor UAV can perform minor position adjustments to optimize link quality. This capability is particularly important in complex mountainous terrain, significantly reducing backhaul interruptions caused by terrain obstruction.

[0039] The ground-based unmanned vehicles are mainly used for road inspection, inspection of the back of photovoltaic modules, and inspection of combiner boxes. The chassis structure adopts a tracked or four-wheel independent drive, with a climbing ability of more than 35 degrees and a ground clearance of more than 120 mm, adapting to the complex road environment in mountainous photovoltaic power stations.

[0040] The unmanned vehicle is equipped with a 32-line LiDAR and a narrow-range infrared detection module for road obstacle recognition and junction box overheating monitoring; it also carries a visible light camera for identifying loose bolts, loose wiring, and foreign object accumulation on the back of components. The unmanned vehicle collects attitude data in real time during operation, and when the tilt angle exceeds 30°, the system triggers an automatic parking function to prevent the equipment from tipping over.

[0041] The fixed base, installed in the inverter room, key combiner box, and ridge location, is a crucial support unit for the field layer. It integrates an ultra-wideband base station, RTK relay module, and millimeter-wave communication unit to provide positioning reference and high-speed data relay. Simultaneously, the base has a built-in edge computing box for task assignment and preliminary data processing. Furthermore, the base features a power supply interface, enabling rapid recharging of drones and unmanned vehicles, extending their continuous operating capability.

[0042] Through the efficient coverage of fixed-wing UAVs, the low-altitude precision inspection and relay of multi-rotor UAVs, the close-range detection of ground unmanned vehicles, and the communication and power replenishment support of fixed bases, the field layer constitutes a collaborative execution unit covering the air and the ground, providing high-quality raw data and stable execution guarantees for the collaborative perception layer and the mission decision-making layer.

[0043] The collaborative sensing layer is a crucial component of the system in this invention for acquiring multi-source information and providing auxiliary positioning. It mainly consists of three parts: multi-mode sensors, artificial anchors, and array geometric priors, forming an integrated air-ground data sensing and positioning constraint network.

[0044] Multi-mode sensors are installed on fixed-wing UAVs, multi-rotor UAVs, and ground unmanned vehicles to simultaneously collect multi-dimensional inspection information.

[0045] Visible light imaging unit: with a resolution of no less than 45 million pixels, it can clearly acquire detailed images of the surface of photovoltaic modules for the identification of cracks, damage, obstructions and contamination.

[0046] Infrared imaging unit: The mid-wave infrared thermal imager has a resolution of no less than 640×512, which can capture the temperature rise distribution of the components in real time, thereby identifying hot spots, loose solder joints, and potential electrical performance abnormalities.

[0047] Near-infrared sensing unit: Employs a near-infrared camera covering the 700 to 1100 nanometer band, which can be used to distinguish between vegetation cover and component surface defects, improving the accuracy of anomaly detection.

[0048] The aforementioned sensors ensure that the acquisition time error is less than 10 milliseconds through a time synchronization module, thereby supporting the joint analysis of multimodal data.

[0049] Artificial anchors are deployed at key inflection points and typical nodes of the photovoltaic array, with spacing between 80 and 150 meters. These artificial anchors utilize high-reflectivity corner points or visual identification codes for easy identification by the airborne camera. The anchor positions are determined with high precision using RTK, serving as positioning reference points when GNSS signals are obstructed. When the UAV enters areas with weak satellite signals, such as canyons or the back of hillsides, the position of the anchors can be visually detected to correct the flight path calculation and prevent positioning drift.

[0050] By utilizing existing design drawings and construction data of photovoltaic power plants, the row and column indices, azimuth angles, tilt angles, and support node information of photovoltaic modules are extracted to form an array geometric prior database. This geometric information, combined with anchor and sensor observation results, is input into the positioning solution module, providing additional geometric constraints when some anchors are damaged or fail to be identified. For example, when a deviation is detected between the module arrangement direction and the expected geometric model, the system automatically generates new visual feature points to replace the missing anchors, ensuring the long-term stability of the positioning system.

[0051] The collaborative sensing layer achieves multi-dimensional observation of the photovoltaic power station's operating status through airborne multi-mode sensors, and establishes a positioning constraint system under complex mountainous terrain conditions through artificial anchors and array geometric priors. This design not only improves the accuracy of anomaly detection but also enhances positioning robustness in GNSS signal-blocked areas, providing reliable data input for path optimization and diagnostic analysis at the mission and decision-making layers.

[0052] The task and decision-making layer is the core control and scheduling module of this invention, responsible for generating efficient and safe inspection task plans in the complex environment of mountainous photovoltaic power stations. This layer mainly consists of two parts: terrain-topology-communication map and joint optimization and replanning, ensuring reasonable task allocation and dynamic adjustment during multi-machine collaborative operation.

[0053] The terrain-topology-communication map is a multi-dimensional information model that comprehensively describes the operating environment of a photovoltaic power station in a mountainous area. It consists of the following four parts: Digital Elevation Model (DEM): With a resolution of no less than 5 meters, it is used to depict the terrain undulations and slope changes within the power station area, providing a basis for fixed-wing UAV trajectory planning and ground unmanned vehicle accessibility analysis.

[0054] Road network information: includes the location, slope and curvature data of maintenance roads, which are used for path planning and safety constraints of ground-based unmanned vehicles.

[0055] Photovoltaic array vector information: records the arrangement, tilt angle, azimuth angle and density of the components, and supports the generation of fine inspection line strips and component-level diagnostic tasks.

[0056] Communication line-of-sight matrix: Based on DEM and ray casting method, combined with signal propagation model, it reflects the visual link and signal-to-noise ratio prediction results between nodes, and provides constraints for the deployment of relay nodes.

[0057] The above four types of information are modeled uniformly in the system, forming the core inputs for the task and decision-making layers.

[0058] Based on the graph, the system adopts a rolling time-domain optimization method, taking inspection time, energy consumption, link quality and risk cost as joint optimization objective functions.

[0059] Energy Constraint: The aircraft is required to retain at least 25% of its energy reserves at the end of the mission to ensure a safe return.

[0060] Communication constraints: The signal-to-noise ratio of the communication link must be no less than 8 dB to ensure stable transmission of images and diagnostic data.

[0061] Ground vehicle constraints: Limit the slope of autonomous vehicles to no more than 26° to avoid vehicle instability due to steep slopes.

[0062] Airspace safety constraints: The minimum horizontal distance between two aircraft is limited to 30 meters, and the minimum height difference is limited to 20 meters, to prevent aerial conflicts.

[0063] The optimization process is updated every 10 seconds. The system dynamically adjusts the drone's flight path, the multi-rotor relay hovering point, and the ground unmanned vehicle's path to ensure the continuity and robustness of mission execution.

[0064] During inspections, if sudden wind disturbances, link fading, or road obstructions are detected, the system will trigger an online replanning mechanism. The replanning process uses the latest environmental awareness data to quickly correct flight paths and relay deployments. To avoid system instability caused by frequent adjustments, the system employs a "risk threshold triggering mechanism," initiating replanning only when link quality continuously declines or energy consumption predictions exceed safety boundaries. After replanning is completed, a safety assessment is performed to ensure that the adjustment plan does not introduce new airspace conflicts or energy shortage risks.

[0065] like Figure 2 As shown, the inputs include a digital elevation model, road network, photovoltaic array vector information, and communication line-of-sight matrix. These data are first used to construct a terrain-topology-communication map. Subsequently, the system performs joint optimization scheduling, outputting an initial inspection path and relay deployment plan. During the inspection process, if the environment changes, the system enters an online replanning phase to dynamically correct the plan, ultimately outputting an updated inspection task plan.

[0066] By constructing and jointly optimizing the terrain-topology-communication map, the task and decision-making layers achieved global coordination of multiple platforms and tasks, which not only ensured the inspection efficiency and communication stability in the complex mountain environment, but also improved the security and adaptability of task execution.

[0067] The edge-cloud collaboration layer is a crucial component in the system of this invention, enabling rapid on-site processing and in-depth remote diagnosis. It mainly consists of an edge computing module and a cloud analysis module, which form a closed-loop mechanism through bidirectional communication, ensuring the real-time nature of data processing and the comprehensiveness of diagnosis.

[0068] The edge computing module is deployed in a fixed base and has an inference capability of no less than 8 TFLOPS. This module receives multi-mode sensor data from fixed-wing UAVs, multi-rotor UAVs, and ground-based unmanned vehicles, and primarily performs the following functions: Initial thermal imaging screening: Convolutional neural networks are used to quickly segment infrared images to detect hot spots and large-area temperature rise anomalies; Crack detection: Based on high-resolution visible light imaging, identify hidden cracks and damage on the surface of components; Occlusion identification: Distinguish between vegetation occlusion and component defects using near-infrared spectral information to reduce false alarm rate; Positioning fusion: It combines UWB ranging, visual anchor recognition, and array geometric constraints into an input factor graph optimization model to generate a high-precision pose solution; Wind field estimation: Combining airborne wind speed sensor data and flight attitude information, a local wind field model is constructed to provide a reference for trajectory correction for mission and decision-making levels.

[0069] The edge computing module performs rapid processing on-site and feeds back the initial screening results and location data to the task and decision-making levels in real time.

[0070] The cloud-based analytics module boasts enhanced computing power and data storage capabilities, primarily used for in-depth diagnostics and results archiving. Its core function is parameter inversion of the thermo-electric-mechanical coupling model, specifically including: Input infrared temperature rise distribution, incident irradiance, component shielding ratio, and support thermal stress index; The range of series resistance, parallel resistance, and photocurrent parameters of the component is inverted using the least squares fitting method; Virtual current-voltage curves are generated based on the inversion results for non-contact electrical performance analysis. A score is calculated based on the abnormal characteristics. When the score is greater than or equal to 0.7, an A-level work order is triggered and the multi-rotor is required to perform a low-altitude fine inspection. When the score is between 0.5 and 0.7, it is included in the review list. When it is less than 0.5, it is recorded as normal.

[0071] The cloud module also has model evolution capabilities, which periodically update the edge-end algorithm through federated learning and model distillation techniques, thereby improving edge diagnostic capabilities over time.

[0072] A high-bandwidth communication link is established between the edge computing module and the cloud analytics module, forming a closed-loop collaborative mechanism: Initial screening is completed quickly at the edge, ensuring immediate response capability on site; Deep diagnostics and parameter inversion are completed in the cloud, and the updated diagnostic model is pushed to the edge. After the work order is generated, it is issued to multi-rotor drones or ground unmanned vehicles through the task and decision-making level to perform secondary fine inspection or close-range review. After the verification results are returned, the cloud model parameters are further corrected to form an adaptive optimization process.

[0073] like Figure 3As shown, infrared, visible light, and near-infrared data are first collected at the edge, and the edge computing module completes the initial thermal image screening, crack detection, and occlusion identification. Subsequently, the data is uploaded to the cloud analysis module for thermo-electrical-mechanical parameter inversion, obtaining anomaly scores and virtual IV curves. The scoring results trigger work order generation, which is divided into two categories: Level A detailed inspection and review checklists. Finally, a drone or unmanned vehicle completes the closed-loop feedback, feeding new data back into the model for updates.

[0074] The edge-cloud collaboration layer enables a working mode that combines rapid on-site screening with remote in-depth diagnosis through the division of labor between the edge and the cloud. This layer not only ensures the real-time performance and security of inspection tasks, but also continuously optimizes the diagnostic model through closed-loop feedback, thereby improving the stability and intelligence level of the entire system in long-term operation.

[0075] In summary, the present invention provides a multi-machine collaborative autonomous inspection system suitable for photovoltaic power stations in mountainous areas. It combines a field layer consisting of fixed-wing UAVs, multi-rotor UAVs, ground unmanned vehicles and fixed bases, a collaborative perception layer consisting of multi-mode sensors, artificial anchors and array geometric priors, a task and decision layer based on terrain-topology-communication maps, and an edge-cloud collaborative layer that combines edge rapid processing and cloud deep diagnosis, forming a complete inspection system that integrates air and ground, multi-mode information fusion and autonomous optimization.

[0076] This system addresses the challenges of complex terrain, communication obstruction, weather disturbances, and dispersed equipment distribution in mountainous photovoltaic power stations by proposing innovative technical means such as a positioning and perception constraint mechanism, a rolling optimization and replanning mechanism, and an edge-cloud closed-loop diagnostic mechanism. In practical applications, the system not only achieves efficient coordination between large-scale coverage and localized precision inspection, but also maintains the stability and accuracy of inspection tasks under sudden environmental changes and multi-source data interference.

[0077] Therefore, the technical solution of this invention effectively overcomes the shortcomings of existing mountain photovoltaic power station inspection methods, such as reliance on manual labor, incomplete coverage, unstable communication links, and inaccurate diagnostic results. It has high intelligence, strong adaptability and scalability, providing a reliable guarantee for the safe and stable operation of mountain photovoltaic power stations, and also providing a solution that can be promoted for the inspection of new energy equipment in other complex terrain environments.

[0078] Example 3 This embodiment provides a multi-machine collaborative autonomous inspection method suitable for photovoltaic power stations in mountainous areas. The method is applied to a multi-machine collaborative autonomous inspection system suitable for photovoltaic power stations in mountainous areas and includes the following steps: Based on the initial environmental map, initial task allocation and global path planning are performed in a multi-machine collaborative manner, generating initial scheduling instructions and issuing them to the field execution layer. Each device in the field execution layer moves according to the scheduling instructions and collects raw inspection data. Based on the collected raw inspection data, combined with the preset spatial distribution information of artificial anchors and the geometric prior information of the photovoltaic array, the raw inspection data is fused and the positioning is calculated to obtain perception data containing positioning information and target status. Based on the perception data and the pre-constructed environmental map, dynamic task instructions containing inspection paths and / or detection tasks are generated. The environmental map includes digital elevation model information, road network, photovoltaic array vector, and communication line-of-sight information.

[0079] Example 4 In another embodiment of the present invention, a computer-readable storage medium is provided as a storage component within a terminal device, the function of which is to store programs and data. It should be noted that the computer-readable storage medium here encompasses not only the built-in storage components of the terminal device but also extended storage components supported by the device. Essentially, it is a tangible medium capable of containing or storing programs that can be invoked by or in conjunction with an instruction execution system, device, or apparatus. This storage medium provides storage areas for the terminal's operating system and stores one or more instructions suitable for processor loading and execution, which can constitute one or more computer programs containing program code.

[0080] Specifically, examples of computer-readable storage media (a non-exclusive list) include: electrical connections with one or more wires, portable disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, optical fibers, portable optical disc read-only memory, optical storage devices, magnetic storage devices, or any reasonable combination of the above types.

[0081] The storage medium may also include data signals propagated as part of a baseband portion or a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any reasonable combination of both. Furthermore, computer-readable storage medium may also refer to other readable media besides conventional readable storage media, capable of sending, propagating, or transmitting programs for use or operation by an instruction execution system, apparatus, or device. Program code on the storage medium can be transmitted via any suitable medium, including but not limited to wireless, wired, optical fiber, or any reasonable combination thereof.

[0082] The program code used to implement the operations of this invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C. The execution modes of the program code include: running entirely on the user's computing device, running partially on the user's device as a standalone software package, running partially in a distributed manner on both the user's device and a remote computing device, or running entirely on a remote computing device or server. When a remote computing device is involved, the device can be connected to the user's computing device via any type of network such as a local area network (LAN) or a wide area network (WAN), or connected to an external computing device via the Internet through an Internet service provider.

[0083] The processor is capable of loading and executing one or more instructions stored in a computer-readable storage medium to implement the multi-machine collaborative autonomous inspection system for mountain photovoltaic power stations described in Example 1.

[0084] Example 5 Figure 4 This is a schematic diagram of a computer device provided according to an embodiment of the present invention.

[0085] Please see Figure 4 The terminal device is a computer device. In this embodiment, the computer device 60 includes a processor 61, a memory 62, and a computer program 63 stored in the memory 62 and executable on the processor 61. When executed by the processor 61, the computer program 63 implements a multi-machine collaborative autonomous inspection system suitable for mountainous photovoltaic power stations, as described in this embodiment. To avoid repetition, details are omitted here. Alternatively, when executed by the processor 61, the computer program 63 implements the functions of each model / unit in the computational system constituting the multi-machine collaborative autonomous inspection system suitable for mountainous photovoltaic power stations, as described in this embodiment. To avoid repetition, details are omitted here.

[0086] Computer device 60 can be a desktop computer, laptop, handheld computer, cloud server, or other computing device. Computer device 60 may include, but is not limited to, a processor 61 and a memory 62. Those skilled in the art will understand that... Figure 4 This is merely an example of computer device 60 and does not constitute a limitation on computer device 60. It may include more or fewer components than shown, or combine certain components, or different components. For example, computer device may also include input / output devices, network access devices, buses, etc.

[0087] The processor 61 may be a central processing unit (CPU), or other general-purpose processors, CPUs, graphics processing units (GPUs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, quantum computing-based data processing logic units, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0088] The memory 62 can be an internal storage unit of the computer device 60, such as a hard disk or RAM of the computer device 60. The memory 62 can also be an external storage device of the computer device 60, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on the computer device 60.

[0089] Furthermore, the memory 62 may include both internal storage units of the computer device 60 and external storage devices. The memory 62 is used to store computer programs and other programs and data required by the computer device. The memory 62 can also be used to temporarily store data that has been output or will be output.

[0090] Any references to memory, databases, or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (Read-Only Memory). Memory includes ROM, magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0091] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

Claims

1. A multi-machine collaborative autonomous inspection system suitable for photovoltaic power stations in mountainous areas, characterized in that, It includes a task and decision-making layer that communicates sequentially, a collaborative perception layer, and an on-site execution layer; The field execution layer is used to execute corresponding inspection actions based on dynamic task instructions and collect raw inspection data. The collaborative perception layer is used to receive the raw inspection data collected by the field execution layer, and based on the preset spatial distribution information of artificial anchors and the geometric prior information of the photovoltaic array, to fuse and locate the raw inspection data to obtain perception data containing location information and target status. The task and decision layer is used to generate dynamic task instructions containing inspection paths and / or detection tasks based on the perception data output by the collaborative perception layer and in combination with a pre-constructed environmental map; the environmental map includes digital elevation model information, road network, photovoltaic array vector and communication line-of-sight information.

2. The multi-machine collaborative autonomous inspection system for photovoltaic power stations in mountainous areas according to claim 1, characterized in that, The field execution layer includes fixed-wing UAVs, multi-rotor UAVs, and unmanned ground vehicles; The fixed-wing UAV is used to respond to the coverage inspection command in the dynamic mission command, perform a cruise within a set mountain range and collect the first raw data; The multi-rotor UAV is used to respond to the fine inspection command or communication relay command in the dynamic mission command, perform low-altitude hovering or fine inspection flight and collect second raw data. The unmanned ground vehicle is used to respond to the near-ground inspection command in the dynamic task command, travel along the inspection road and collect third raw data.

3. The multi-machine collaborative autonomous inspection system for photovoltaic power stations in mountainous areas according to claim 2, characterized in that, The field execution layer also includes a fixed base, which is installed in the inverter room and key combiner box in the mountain ridge, including a broadband positioning base station, an RTK differential signal relay unit, a millimeter wave communication unit and a fast charging interface; The broadband positioning base station is used to provide relative positioning signals for fixed-wing drones, multi-rotor drones or ground unmanned vehicles that enter the positioning coverage area. The RTK differential signal relay unit is used to receive and forward GNSS differential correction signals to correct the absolute positioning accuracy of the fixed-wing UAV, multi-rotor UAV or ground unmanned vehicle. The millimeter-wave communication unit is used to establish high-speed data transmission links between fixed-wing UAVs, multi-rotor UAVs, or ground unmanned vehicles, or between different fixed bases. The fast charging interface is used to connect the fixed-wing UAV, multi-rotor UAV, or ground unmanned vehicle.

4. The multi-machine collaborative autonomous inspection system for photovoltaic power stations in mountainous areas according to claim 1, characterized in that, The task and decision-making layer includes a graph construction module and a dynamic scheduling module; The map construction module is used to construct and update a dynamic environmental map reflecting terrain, equipment topology and communication links based on the environmental map and the positioning information in the sensing data. The dynamic scheduling module is used to periodically calculate and output the dynamic task instructions, taking the dynamic environment map as input and taking task completion efficiency, equipment energy consumption and operation safety as joint optimization objectives.

5. The multi-machine collaborative autonomous inspection system for photovoltaic power stations in mountainous areas according to claim 4, characterized in that, The constraints of the joint optimization objective include: the remaining energy of fixed-wing UAVs and multi-rotor UAVs is not lower than the energy threshold, the signal-to-noise ratio of the communication link is not lower than the communication threshold, the driving slope of the ground unmanned vehicle does not exceed the slope threshold, and a minimum safe distance is maintained between fixed-wing UAVs or between multi-rotor UAVs.

6. The multi-machine collaborative autonomous inspection system for photovoltaic power stations in mountainous areas according to claim 4, characterized in that, The dynamic scheduling module employs a rolling time-domain optimization method, performing optimization calculations once at set time intervals.

7. The multi-machine collaborative autonomous inspection system for photovoltaic power stations in mountainous areas according to claim 4, characterized in that, It also includes an edge-cloud collaboration layer, which comprises an edge computing module and a cloud analytics module; The edge computing module is used to receive the sensing data, analyze the sensing data to obtain edge screening results including abnormal suspected areas and preliminary categories, and feed the edge screening results back to the task and decision layer. The cloud analysis module is used to receive the edge screening results uploaded by the edge computing module, and perform parameter inversion and deep diagnosis based on the physical model to generate cloud analysis results containing anomaly scores and diagnostic reports; the cloud analysis results are used to correct the optimization strategies of the task and decision layers and the analysis model of the edge computing module.

8. The multi-machine collaborative autonomous inspection system for photovoltaic power stations in mountainous areas according to claim 4, characterized in that, The edge computing module includes: initial screening of hot spots based on infrared thermal imaging, crack detection based on visible light images, and vegetation occlusion identification based on near-infrared images.

9. The multi-machine collaborative autonomous inspection system for photovoltaic power stations in mountainous areas according to claim 7, characterized in that, The anomaly score generated in the cloud analysis module is a value between 0 and 1; when the anomaly score is greater than or equal to the first threshold, a detailed inspection work order is triggered based on the cloud analysis result. When the abnormal score is between the second threshold and the first threshold, the cloud analysis result triggers a review flag; The first threshold is greater than the second threshold.

10. A method for a multi-machine collaborative autonomous inspection system for photovoltaic power stations in mountainous areas, applicable to any one of claims 1 to 9, characterized in that, The method includes: Based on the initial environment map, initial task allocation and global path planning are performed for multi-machine collaboration, and initial scheduling instructions are generated and sent to the field execution layer; Each device in the field execution layer moves and collects raw inspection data according to the scheduling instructions. Based on the collected raw inspection data, combined with the preset spatial distribution information of artificial anchors and the geometric prior information of the photovoltaic array, the raw inspection data is fused and located to obtain perception data containing positioning information and target status. Based on the perceived data and combined with the pre-constructed environmental map, dynamic task instructions containing inspection paths and / or detection tasks are generated; the environmental map includes digital elevation model information, road network, photovoltaic array vectors, and communication line-of-sight information.

Images