New energy station multi-type inspection equipment cooperative control method and comprehensive management and control platform

Abstract

The invention provides a new energy station multi-type inspection equipment cooperative control method and a comprehensive management and control platform, and the method comprises the steps: generating a global inspection path comprising an unmanned plane, a cabin inspection robot and a ground inspection vehicle based on a particle swarm optimization algorithm through obtaining a station topological structure, an equipment state, and meteorological and historical defect data; a priority inspection area is determined through time sequence analysis of equipment state data, and a path is dynamically adjusted according to obstacle information. According to the method, the instruction synchronization and the data sharing of multiple devices are realized by adopting a user-defined New Entry Protocol, and the collision of the devices is avoided by combining time window scheduling and space partitioning. And carrying out data fusion analysis by using a YOLOv5 algorithm and the like, generating a maintenance work order and optimizing an equipment inspection period. And meanwhile, a reinforcement learning model optimization cooperation strategy is constructed, so that multi-equipment'air-ground-cabin 'intelligent cooperation is realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance of new energy power stations, and in particular to a collaborative control method and integrated management and control platform for multiple types of inspection equipment in new energy power stations. Background Technology

[0002] With the rapid development of the new energy industry, the scale of new energy power plants such as wind farms and photovoltaic power stations is constantly expanding. Equipment is widely distributed and the environment is complex, making equipment operation and maintenance management paramount. Inspection work, as a core component, has always been a crucial line of defense for ensuring normal equipment operation and production safety. Traditional manual inspection methods have long faced numerous challenges, including low efficiency, high safety risks, and incomplete coverage, making it difficult to meet the operation and maintenance needs of large-scale new energy power plants.

[0003] Currently, while there are existing inspection solutions that replace manual inspections, such as mobile inspection robots for monitoring equipment appearance and the surrounding environment, and sensor network systems for detecting internal electrical parameters, these inspection devices mostly operate independently, lacking effective coordination. In actual inspection processes, this independent operation mode causes a series of problems. For example, task overlap and data silos exist. With the increase in the number of devices and the expansion of the inspection space, the risk of device collisions also arises. Different inspection devices move along their own planned routes within the inspection area, lacking unified scheduling and obstacle avoidance coordination, thus failing to fully utilize the advantages of each device. Summary of the Invention

[0004] In view of the above problems, this invention proposes a collaborative control method for multiple types of inspection equipment in new energy power stations to overcome or at least partially solve the above problems. This method addresses the issues caused by the independent operation and lack of coordination of equipment in traditional inspection schemes, and realizes intelligent collaborative control and task planning, thereby improving inspection efficiency and accuracy, and reducing operation and maintenance costs and safety risks.

[0005] According to one aspect of the present invention, a method for collaborative control of multiple types of inspection equipment in a new energy power station is provided, the method comprising:

[0006] Acquire topology data, equipment status data, meteorological data, and historical defect data of new energy power stations;

[0007] Based on the aforementioned topology data, equipment status data, meteorological data, and historical defect data, a global inspection path is generated using a particle swarm optimization algorithm, which is then executed collaboratively by multiple types of inspection equipment. The inspection equipment includes drones, cabin inspection robots, and ground inspection vehicles, and the global inspection path includes at least a drone inspection path, a cabin inspection robot inspection path, and a ground inspection vehicle inspection path.

[0008] Identify abnormal data in the equipment status data, determine the priority inspection area of ​​the new energy power station based on the abnormal data, and adjust the global inspection path to prioritize the coverage of the priority inspection area.

[0009] The system monitors obstacle information within the inspection area of ​​the new energy power station in real time and dynamically adjusts the global inspection path based on the obstacle information.

[0010] Optionally, the topology data is acquired through three-dimensional laser scanning or satellite remote sensing technology;

[0011] The equipment status data is collected in real time by various sensors installed on each piece of equipment in the new energy power station. The equipment status data includes equipment operating parameters, including wind turbine vibration value, gearbox temperature, and photovoltaic panel power generation efficiency.

[0012] The meteorological data is obtained through monitoring by meteorological stations deployed at the site, including wind speed, light intensity, and rainfall.

[0013] The historical defect data is stored in the historical defect database, which includes the types, locations, and handling status of previous equipment failures.

[0014] Optionally, based on the topology data, equipment status data, meteorological data, and historical defect data, the global inspection path is generated using a particle swarm optimization algorithm, including:

[0015] Based on the aforementioned topological data, a three-dimensional model of the new energy power station is constructed; the three-dimensional model includes the wind turbine location coordinates, photovoltaic panel array distribution, booster station layout, and obstacle spatial coordinates.

[0016] Based on the equipment status data, each device is assigned an inspection priority level;

[0017] Based on the meteorological data, multi-dimensional inspection time windows are divided; the time windows include at least the UAV high-altitude inspection window, the cabin robot in-cabin operation window, and the ground vehicle inspection window.

[0018] Based on the aforementioned 3D model, inspection priority levels, and multi-dimensional inspection time windows, a global inspection path is generated using a particle swarm optimization algorithm.

[0019] Optionally, dividing the inspection time window based on the meteorological data includes:

[0020] When the wind speed is less than or equal to the first set wind speed and the light intensity is greater than or equal to the first set light intensity, it is designated as the high-altitude inspection window for UAVs, and the flight altitude range of the UAVs within the window is the first set high zone range.

[0021] When the wind speed is greater than the first set wind speed and less than or equal to the second set wind speed, or the light intensity is less than the first set light intensity and greater than or equal to the second set light intensity, a mixed inspection window is defined. The drone's flight altitude within the window is adjusted to the second set high zone, and the ground inspection vehicle's preprocessing path is started simultaneously.

[0022] When the wind speed exceeds the second set wind speed or the rainfall exceeds the set rainfall, a priority window for robots is designated. Only cabin inspection robots and ground inspection vehicles are allowed to operate within this window, and the drone returns to the charging base station.

[0023] Optionally, after the global inspection path is executed collaboratively by multiple types of inspection equipment, the method further includes:

[0024] Acquire inspection result data collected by inspection equipment; the inspection result data includes image data captured by drones, sensor detection data collected by cabin inspection robots, and / or ultrasonic detection data collected by ground inspection vehicles;

[0025] New inspection tasks or maintenance work orders are generated based on the inspection results data, and the equipment health records are dynamically updated based on the inspection results data.

[0026] Optionally, the method further includes:

[0027] Establish a cross-device command synchronization protocol, which includes at least priority queues for drone photography commands, robot sensor activation commands, and ground vehicle path switching commands;

[0028] Based on the priority queue, when the drone detects an equipment abnormality, it generates a trigger command containing geographic coordinates and abnormality type. The trigger command takes priority over non-urgent commands in the current queue.

[0029] After receiving the trigger command, the cabin inspection robot plans an in-cabin navigation path based on the geographic coordinates. The navigation path includes the infrared thermal imager detection path and the vibration sensor scanning points.

[0030] Optionally, the cross-device command synchronization protocol adopts a custom New Energy Protocol; the cross-device command synchronization protocol includes:

[0031] Device identification field, used to identify a unique ID for a drone, robot, or ground vehicle;

[0032] The task type field is used to distinguish between inspection tasks, re-inspection tasks, or emergency tasks;

[0033] The data format field is used to specify the transmission protocol for visible light images, infrared thermal images, vibration data, and ultrasonic data.

[0034] The conflict arbitration field is used to define the avoidance rules when multiple devices simultaneously apply for the same work area.

[0035] Optionally, identifying anomalous data in the device status data includes:

[0036] Perform time-series analysis on the equipment status data to construct a normal distribution model of the equipment operating parameters;

[0037] Calculate the deviation between the real-time parameters and the normal distribution model. When the deviation is greater than a preset threshold, it is marked as abnormal data.

[0038] Optionally, the preset threshold is dynamically updated based on fault parameter samples in the historical defect database, with an update cycle of once a day.

[0039] According to another aspect of the present invention, a comprehensive management and control platform is also provided, the platform comprising a processor and a memory:

[0040] The memory is used to store program code and transmit the program code to the processor;

[0041] The processor is used to execute the collaborative control method for multiple types of inspection equipment in new energy power stations as described above, according to the instructions in the program code.

[0042] This invention relates to a collaborative control method and integrated management platform for multiple types of inspection equipment in new energy power stations. By acquiring data on the station's topology, equipment status, weather conditions, and historical defects, it generates a global inspection path based on a particle swarm optimization algorithm, incorporating drones, cabin inspection robots, and ground inspection vehicles. Priority inspection areas are determined through time-series analysis of equipment status data, and the path is dynamically adjusted based on obstacle information. A custom New Energy Protocol is used to achieve multi-device command synchronization and data sharing, combined with time window scheduling and spatial partitioning to avoid equipment collisions. Data fusion analysis is performed using algorithms such as YOLOv5 to generate maintenance work orders and optimize equipment inspection cycles. Simultaneously, a reinforcement learning model is constructed to optimize collaborative strategies, achieving intelligent "air-ground-cabin" collaboration among multiple devices.

[0043] The technical solution provided by this invention has the following advantages compared with the prior art:

[0044] 1. This invention integrates multi-source data through a unified interface protocol, eliminating data silos. Through intelligent data analysis using algorithms, potential faults are detected promptly, reducing equipment failure rates and maintenance costs. The YOLOv5 algorithm and signal processing methods used in data fusion analysis and feedback optimization enable effective analysis and fault identification of various data types.

[0045] 2. This invention uses time window scheduling and spatial partitioning to avoid equipment collisions, and real-time obstacle avoidance and emergency task prioritization to improve emergency response capabilities. The real-time obstacle avoidance function in global task planning and dynamic allocation, as well as the time and space scheduling strategy in the multi-device collaboration mechanism, ensure the safe and stable operation of inspection work.

[0046] 3. The multi-device collaborative energy management mechanism of this invention rationally utilizes renewable energy, reducing equipment energy consumption and dependence on traditional energy sources. The mechanism's monitoring of equipment power consumption, charging path planning, and the use of renewable energy for power supply at charging base stations achieve green and energy-saving goals.

[0047] 4. This invention utilizes reinforcement learning-based collaborative strategy optimization, enabling the system to automatically optimize strategies according to different site conditions and environments, thereby improving adaptability and intelligence. The reinforcement learning-based collaborative strategy optimization module provides strong support for intelligent operation and maintenance of new energy power plants through continuous learning and iteration.

[0048] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below.

[0049] The above and other objects, advantages and features of the present invention will become more apparent to those skilled in the art from the following detailed description of specific embodiments of the invention in conjunction with the accompanying drawings. Attached Figure Description

[0050] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0051] Figure 1 A schematic diagram of the collaborative control method for multiple types of inspection equipment in a new energy power station according to an embodiment of the present invention is shown. Detailed Implementation

[0052] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0053] This invention provides a collaborative control method for multiple types of inspection equipment in new energy power stations, which can be applied to a comprehensive management and control platform. For example... Figure 1 As shown, the collaborative control method for multiple types of inspection equipment in new energy power stations according to an embodiment of the present invention may include the following steps S1 to S4.

[0054] S1 acquires topology data, equipment status data, meteorological data, and historical defect data of new energy power stations.

[0055] Topology data can be acquired through technologies such as 3D laser scanning and satellite remote sensing. Equipment status data is collected in real time by various sensors installed on the equipment in the renewable energy power station, such as temperature sensors, vibration sensors, current and voltage sensors. The collected equipment status data can include real-time operating parameters of the equipment, such as wind turbine vibration values, gearbox temperature, and photovoltaic panel power generation efficiency.

[0056] Meteorological data is acquired through monitoring by meteorological stations deployed at the site, including wind speed, light intensity, and rainfall; historical defect data is stored in a historical defect database, including information such as the types, locations, and handling status of past equipment failures.

[0057] S2. Based on the above topology data, equipment status data, meteorological data and historical defect data, a global inspection path is generated using the particle swarm optimization algorithm, so that multiple types of inspection equipment can work together to execute the global inspection path. The inspection equipment includes drones, cabin inspection robots and ground inspection vehicles. The global inspection path includes at least the drone inspection path, the cabin inspection robot inspection path and the ground inspection vehicle inspection path.

[0058] S3 identifies abnormal data in the equipment status data, determines the priority inspection area of ​​the new energy power station based on the abnormal data, and adjusts the global inspection path to prioritize the coverage of the priority inspection area.

[0059] S4 monitors obstacle information in the inspection area of ​​new energy power stations in real time and dynamically adjusts the global inspection path based on the obstacle information.

[0060] The collaborative control method for multiple types of inspection equipment in new energy power stations according to this invention acquires the power station topology, equipment status, meteorological and historical defect data, and generates a global inspection path including UAVs, cabin inspection robots and ground inspection vehicles based on particle swarm optimization algorithm. It determines the priority inspection area by analyzing the time sequence of equipment status data and dynamically adjusts the path according to obstacle information to achieve intelligent collaboration of multiple devices in the air-ground-cabin. However, this method has led to problems such as task overlap, data silos and equipment collision risks, which prevent the full utilization of the advantages of each device.

[0061] In an optional embodiment of the present invention, step S2 above, which generates a global inspection path based on topology data, equipment status data, meteorological data, and historical defect data using a particle swarm optimization algorithm, may include: constructing a three-dimensional model of the new energy power station based on the topology data; the three-dimensional model includes the wind turbine location coordinates, photovoltaic panel array distribution, substation layout, and obstacle spatial coordinates; assigning inspection priority levels to each piece of equipment based on equipment status data; dividing multi-dimensional inspection time windows based on meteorological data; the time windows include at least a UAV high-altitude inspection window, a cabin robot in-cabin operation window, and a ground vehicle inspection window; and generating a global inspection path using a particle swarm optimization algorithm based on the three-dimensional model, inspection priority levels, and multi-dimensional inspection time windows.

[0062] In this embodiment of the invention, a global inspection path is generated by combining a three-dimensional model, inspection priority levels, and multi-dimensional inspection time windows divided by meteorological data with Particle Swarm Optimization (PSO). PSO is a swarm intelligence-based stochastic search algorithm with advantages such as simple structure, fast convergence speed, and ease of implementation. It is suitable for solving complex combinatorial optimization problems such as multi-device collaborative inspection path planning, and can quickly find a better solution among many possible path combinations, thus improving inspection efficiency.

[0063] As mentioned earlier, dividing inspection time windows based on meteorological data can include: when the wind speed is less than or equal to a first set wind speed and the light intensity is greater than or equal to a first set light intensity, it is designated as a high-altitude inspection window for drones, and the drone's flight altitude range within the window is the first set high-altitude zone; when the wind speed is greater than the first set wind speed but less than or equal to a second set wind speed or the light intensity is less than the first set light intensity but greater than or equal to the second set light intensity, it is designated as a mixed inspection window, and the drone's flight altitude within the window is adjusted to the second set high-altitude zone, while the ground inspection vehicle's preprocessing path is simultaneously initiated; when the wind speed is greater than the second set wind speed or the rainfall is greater than the set rainfall, it is designated as a robot priority window, and only the cabin inspection robot and the ground inspection vehicle are allowed to operate within the window, while the drone returns to the charging base station.

[0064] For example, this embodiment uses a combination of time window scheduling and spatial partitioning to divide multi-dimensional inspection time windows based on meteorological data: when the wind speed is ≤5.4m / s and the light intensity is ≥20000Lux, it is designated as a high-altitude inspection window for UAVs, with the UAV flight altitude ranging from 80 to 150 meters. Under these conditions, the meteorological environment is stable and suitable for efficient high-altitude UAV inspections; when the wind speed is >5.4m / s and ≤8.0m / s or the light intensity is <20000Lux and ≥10000Lux, it is designated as a mixed inspection window, with the UAV flight altitude adjusted to 150-200 meters and the ground inspection vehicle preprocessing path initiated simultaneously, leveraging the synergistic advantages of UAVs and ground inspection vehicles under relatively complex meteorological conditions; when the wind speed is >8.0m / s or the rainfall is >5mm / h, it is designated as a robot priority window, allowing only the cabin inspection robot and the ground inspection vehicle to operate, while the UAV returns to the charging base station to ensure the safe operation of the equipment.

[0065] In terms of spatial zoning, a high-altitude inspection layer of over 100 meters is designated for drones, an operating layer inside the wind turbine nacelle (approximately 20-30 meters high) is set up for nacelle inspection robots, and a ground inspection vehicle is designated for a ground inspection layer of 0-5 meters, ensuring that only one type of equipment operates in the same area at the same time.

[0066] The navigation path execution mechanism of the cabin inspection robot in this embodiment of the invention is as follows:

[0067] 1. Path planning and execution process

[0068] Command reception and parsing: The robot receives trigger commands (including geographic coordinates, such as the location of gearbox No. 3 in wind turbine A nacelle) through the New Energy Protocol.

[0069] Cabin map matching: Based on a pre-built 3D cabin map (including equipment layout and passageway location), plan the shortest path from the current location to the target point (such as avoiding cable trays and reserving maintenance space).

[0070] Sensor task allocation:

[0071] Infrared thermal imager detection path: Thermal imaging is performed along 5 preset scanning points on the gearbox housing (such as the bearing positions of the input shaft and output shaft) to detect abnormal temperatures;

[0072] Vibration sensor scanning points: Vibration spectrum data are collected at four fixed mounting points on the gearbox housing (such as the top and side).

[0073] Data feedback: After the detection is completed, the infrared images and vibration data are encrypted and transmitted to the integrated management and control platform, where the YOLOv5 algorithm and Fourier transform are used to analyze whether there are any faults (such as bearing wear or abnormal gear meshing).

[0074] 2. Example Description

[0075] When the drone detects a crack in the blade of wind turbine A (geographic coordinates X = 100m, Y = 200m, Z = 90m), the triggered command includes these coordinates and the type "re-inspection of blade root connection components". The nacelle robot, located at the entrance of the wind turbine A nacelle, receives the command and then:

[0076] Planned path: Inlet → Blade hub connection channel → Root connection component inspection point;

[0077] The infrared thermal imager scans the area of ​​the connecting bolts along a preset path to detect any abnormal heating (such as exceeding 60°C).

[0078] Vibration sensors collect vibration data at three points near the connecting parts to analyze whether there are high-frequency abnormal noises (>500Hz);

[0079] If the bolt temperature is detected to be 75℃ and an abnormal peak appears in the vibration spectrum, the platform will automatically generate a maintenance work order requiring the bolt to be replaced.

[0080] In an optional embodiment of the present invention, step S3, identifying abnormal data in the equipment status data, may include: performing time-series analysis on the equipment status data to construct a normal distribution model of the equipment operating parameters; calculating the deviation between the real-time parameters and the normal distribution model; and marking the data as abnormal when the deviation exceeds a preset threshold. Optionally, the preset threshold is dynamically updated based on fault parameter samples in a historical defect database, with an update cycle of once daily.

[0081] When equipment status data is abnormal, time-series analysis can be performed on the data to construct a normal distribution model of the equipment operating parameters. By calculating the deviation between real-time parameters and the normal distribution model, when the deviation exceeds a preset threshold (which is dynamically updated daily based on fault parameter samples in the historical defect database), it is marked as abnormal data, and a priority inspection area is determined. The global inspection path is then adjusted to prioritize coverage of this area.

[0082] Taking the temperature of a fan gearbox as an example, the process of time series analysis and normal distribution modeling of equipment status data can be as follows:

[0083] 1. Historical data collection and modeling

[0084] Data acquisition: Collect temperature data of a fan gearbox for the past 30 days (once per minute, 43,200 data points in total). During normal operation, the temperature is concentrated between 40 and 60°C, and can reach 70°C during peak load.

[0085] Time series analysis: Plotting the temperature change curve over time reveals that the temperature exhibits a periodic pattern (higher during the day and lower at night) and conforms to the characteristics of a normal distribution.

[0086] Normal distribution modeling: Calculate the mean μ = 55℃, standard deviation σ = 5℃, and construct the probability density function:

[0087]

[0088] Of these, 95% of the data fell within the range of μ±2σ (45–65℃), and 99.7% fell within the range of μ±3σ (40–70℃).

[0089] 2. Real-time data deviation calculation and anomaly marking

[0090] Real-time data: At 14:00 on a certain day, the gearbox temperature was detected to be 75℃.

[0091] Deviation calculation: Deviation = |75-μ| / σ = |75-55| / 5 = 4 (i.e., deviation from the mean by 4 standard deviations).

[0092] Threshold comparison: The preset threshold is 2.5 (corresponding to outside the 99.7% confidence interval). If 4 > 2.5, it is marked as abnormal data.

[0093] Dynamic updates: If multiple high-temperature data (such as above 70℃) are detected later in the day, the historical defect database will be updated with fault samples, and the threshold may be adjusted to 2.8 the next day to improve the sensitivity of anomaly identification.

[0094] Taking photovoltaic panel power generation efficiency as an example, the time series analysis and deviation calculation process when equipment malfunctions can be as follows:

[0095] 1. Historical data modeling

[0096] The power generation efficiency data of a photovoltaic array over the past month (9:00 to 15:00 daily, with a light intensity ≥20000 Lux) has a mean μ = 18.5%, a standard deviation σ = 0.8%, and 95% of the data fall between 16.9% and 20.1%.

[0097] 2. Anomaly Detection Process

[0098] Real-time data: At 11:00 AM on a certain day, when the illuminance was 25,000 Lux, the power generation efficiency was only 15.2%.

[0099] Deviation calculation: Deviation = |15.2-18.5| / 0.8≈4.125.

[0100] Dynamic threshold: The preset threshold is updated daily based on historical data. When there are no anomalies the previous day, the threshold is 2.5. If 4.125 > 2.5, it is marked as an anomaly.

[0101] Cause identification: The platform, based on infrared thermal imaging data, discovered hot spots on the photovoltaic panel (local temperature 80℃, normal ≤50℃), which was determined to be a damaged battery cell, triggering the ground vehicle to prioritize inspection of this area.

[0102] In an optional embodiment of the present invention, a cross-device command synchronization protocol can also be established. The protocol includes at least a priority queue of drone photography commands, robot sensor activation commands, and ground vehicle path switching commands. Based on the priority queue, when the drone detects a device abnormality, it generates a trigger command containing geographic coordinates and the type of abnormality. The trigger command takes priority over non-urgent commands in the current queue. After receiving the trigger command, the cabin inspection robot plans an in-cabin navigation path based on geographic coordinates. The navigation path includes the infrared thermal imager detection path and vibration sensor scanning points.

[0103] Optionally, the cross-device command synchronization protocol adopts a custom New Energy Protocol; the cross-device command synchronization protocol includes: a device identification field, used to identify the unique ID of the drone, robot or ground vehicle; a task type field, used to distinguish between inspection tasks, re-inspection tasks or emergency tasks; a data format field, used to standardize the transmission protocol of visible light images, infrared thermal images, vibration data and ultrasonic data; and a conflict arbitration field, used to define the avoidance rules when multiple devices apply for the same work area at the same time.

[0104] The integrated management and control platform uses a custom New Energy Protocol to achieve multi-device command synchronization and data sharing. This protocol is compatible with common protocols such as ROS (Robot Operating System) and Modbus.

[0105] The New Energy Protocol includes fields for device identification, task type, data format, and conflict arbitration. The device identification field uses a unique ID to identify drones, robots, or ground vehicles, ensuring commands are accurately sent to the corresponding device. The task type field distinguishes between inspection tasks, re-inspection tasks, and emergency tasks, facilitating appropriate platform scheduling based on task type. The data format field standardizes the transmission protocols for visible light images, infrared thermal images, vibration data, and ultrasonic data, ensuring accurate data transmission and parsing between different devices and platforms. The conflict arbitration field defines avoidance rules when multiple devices simultaneously apply for the same work area, preventing device conflicts.

[0106] In an optional embodiment of the present invention, the priority order of the cross-device instruction priority queue from high to low can be as follows:

[0107] Level 1: Emergency Mission Trigger Command: Commands generated when the drone detects emergency anomalies such as smoke or severe vibration in the equipment must immediately interrupt non-emergency missions and be executed with priority.

[0108] Level 2: Re-inspection task instruction: Re-inspection instruction of the cabin robot triggered by UAV or ground inspection vehicle (such as blade crack confirmation, gearbox abnormal positioning).

[0109] Level 3: Equipment Fault Warning Command: Warning based on real-time data deviation exceeding a threshold (such as a sudden rise in gearbox temperature or a sudden drop in photovoltaic panel power generation efficiency).

[0110] Level 4: Routine Inspection Task Instructions: Planned tasks such as UAV high-altitude inspections and ground vehicle path inspections, which are assigned according to preset time windows.

[0111] Level 5: Equipment status query command: Non-urgent data acquisition or status confirmation command (such as periodic sensor data feedback).

[0112] For example, when a drone detects a crack in a wind turbine blade using YOLOv5 during an inspection (emergency trigger command), the command will immediately jump to the top of the queue, taking precedence over the routine inspection path switching command being executed by the ground vehicle, ensuring that the cabin robot immediately starts the re-inspection task.

[0113] The aforementioned non-emergency instructions refer to routine task instructions that do not involve equipment failure warnings or safety risks, including: Planned inspection instructions: such as daily fixed-time ground vehicle inspections of photovoltaic panels, and weekly drone aerial photography of the entire site. Periodic data collection instructions: such as hourly collection of routine parameters like wind turbine nacelle temperature and photovoltaic inverter voltage. Equipment status query instructions: non-emergency equipment status queries manually initiated by maintenance personnel (e.g., querying the battery level of a specific robot). Route switching instructions (routine tasks): normal scheduling instructions for ground vehicles to move from area A to area B according to a preset route.

[0114] The comparison scenarios are as follows:

[0115] Emergency command: The drone detected violent swaying of the wind turbine blades (potential risk of breakage), triggering an immediate return to base and initiation of a robot re-inspection command.

[0116] Non-emergency instructions: Path switching instructions for ground vehicles to move from the east side to the west side of the photovoltaic area for routine inspections as planned.

[0117] The platform achieves command synchronization and data sharing between devices through a custom New Energy Protocol. In an optional embodiment of the invention, after multiple types of inspection devices collaboratively execute the global inspection path, the method further includes: acquiring inspection result data collected by the inspection devices; the inspection result data includes image data captured by a drone, sensor detection data collected by a cabin inspection robot, and / or ultrasonic detection data collected by a ground inspection vehicle; generating a new inspection task or a maintenance work order based on the inspection result data.

[0118] For example, after a drone completes an inspection of a certain area, it transmits image data to the platform. The platform analyzes the data and, if ground inspection is required, sends instructions to the ground inspection vehicle. In terms of time window scheduling and spatial zoning, different inspection time windows are divided according to weather conditions, assigning different operating levels and time windows to each piece of equipment. For instance, drones conduct inspections at altitudes above 100 meters from 8:00 AM to 12:00 PM, nacelle inspection robots operate inside wind turbine nacelles from 2:00 PM to 6:00 PM, and ground inspection vehicles conduct inspections at ground levels between 0 and 5 meters from 8:00 PM to midnight, thus avoiding equipment conflicts.

[0119] Furthermore, the YOLOv5 target detection algorithm can be used to analyze images captured by drones. YOLOv5 is a highly efficient target detection algorithm characterized by fast detection speed and high accuracy. The algorithm is optimized for the characteristics of new energy power station equipment, such as adjusting the network structure and enhancing training data, enabling it to accurately identify anomalies such as cracks and hot spots in photovoltaic panels and damage to wind turbine blades. For vibration sensor inspection data collected by the nacelle inspection robot, signal processing methods such as Fourier transform are used to analyze the frequency components of the signal to determine whether there is wear in the gearbox or damage to the bearings; ultrasonic detection data from the ground inspection vehicle is processed to detect internal defects in the equipment.

[0120] Based on the above inspection results data, automatic analysis can be performed to generate maintenance work orders. The work orders contain detailed information such as the location, type, and severity of the fault, and are promptly fed back to maintenance personnel via SMS, email, and other means.

[0121] Meanwhile, equipment health records are dynamically updated based on inspection results, recording historical fault information, inspection time, and maintenance status. Based on equipment health status, subsequent task cycles are optimized; for example, inspection intervals are shortened and inspection frequency increased for fans with more defects, while inspection intervals are appropriately extended for equipment in good condition, thus allocating resources rationally.

[0122] In an optional embodiment of the present invention, participating in step S4 above can also enable real-time monitoring of obstacle information within the inspection area of ​​the new energy power station, and dynamically adjust the global inspection path based on the obstacle information. For example, real-time monitoring of obstacle information such as birds and temporary structures, and using sensor technology (such as lidar, cameras, etc.) to obtain information such as the location, shape, and movement status of obstacles, and dynamically adjusting the global inspection path based on this information to avoid equipment collisions.

[0123] In addition, this embodiment can also support emergency task queueing. When an emergency such as a serious equipment failure occurs, the current inspection task can be interrupted immediately and the emergency task can be executed first.

[0124] Taking the detection of anomalies in wind turbine blades by a drone as an example, the drone's high-definition camera and image recognition module take pictures and perform preliminary analysis of the blades. When anomalies such as cracks or wear are detected, a trigger command containing geographic coordinates and anomaly type is automatically generated. This command is transmitted to the integrated management platform via a wireless communication module. The platform prioritizes this trigger command over non-urgent commands in the current queue based on a priority queue, and then sends the command to the cabin inspection robot. After receiving the command, the robot plans its cabin navigation path based on geographic coordinates using its own navigation system (such as a navigation system based on SLAM technology). The path includes the infrared thermal imager detection path and vibration sensor scanning points. The robot then enters the target cabin for close-range re-inspection and diagnosis, forming a closed-loop inspection mode of "air + cabin".

[0125] This invention also provides an integrated management and control platform, which includes a processor and a memory: the memory is used to store program code and transmit the program code to the processor; the processor is used to execute the collaborative control method for multiple types of inspection equipment in new energy power stations according to the instructions in the program code described above.

[0126] Taking a large wind farm as an example, 3D laser scanning technology is used to acquire the wind farm's topology data, forming a 3D model that includes information such as wind turbine locations and substation layout. Various sensors on the wind turbines collect real-time equipment status data, such as generator speed sensors, gearbox temperature sensors, and vibration sensors. Meteorological stations monitor meteorological data such as wind speed, wind direction, light intensity, and rainfall. A historical fault database stores past fault information.

[0127] Based on this data, the platform uses the Particle Swarm Optimization (PSO) algorithm to generate an initial inspection path. An example of how the PSO algorithm generates inspection paths in this embodiment of the invention is as follows:

[0128] I. Data Preparation and Model Building

[0129] Taking a 100MW wind farm as an example:

[0130] 1. Three-dimensional model: The coordinates of wind turbines (e.g., 10 wind turbines distributed in a 5km×3km area, with a height of 80-120 meters), the location of the booster station, and the spatial coordinates of obstacles (e.g., trees, buildings) are obtained through laser scanning.

[0131] 2. Priority level: Set according to equipment operating parameters:

[0132] The gearbox temperature of fan A remains above 80℃ (priority level 1, requiring priority inspection);

[0133] The vibration value of fan B is within the normal range (priority level 3, routine inspection).

[0134] 3. Time window: Divided according to meteorological data:

[0135] 08:00~12:00: Wind speed ≤5.4m / s, designated as the window for high-altitude drone inspection (height 80~150 meters);

[0136] 14:00~18:00: Wind speed 5.6~7.8m / s, designated as a mixed window (UAV altitude 150~200 meters, ground vehicle simultaneous inspection).

[0137] II. PSO Algorithm Execution Steps

[0138] 1. Particle definition: Each particle represents a set of inspection paths (such as a combination of drone paths, robot paths, and ground vehicle paths). Particle position parameters include the inspection sequence of each device, the coordinates of the dwell point, and the time allocation.

[0139] 2. Fitness function: comprehensively considers inspection efficiency (shortest total path length), priority coverage (high priority equipment inspection time accounts for ≥40%), and time window compliance (drones do not operate during wind speed restriction periods).

[0140] 3. Iterative optimization process:

[0141] Initialize 100 random paths for particles, such as a drone starting from the booster station and inspecting wind turbines 1-5 in random order, and a ground vehicle inspecting the edge of the photovoltaic area.

[0142] Calculate the fitness value of each particle. For example, particle A has a total path length of 12km, covering all level 1 priority wind turbines, and has a fitness score of 90; particle B has a path length of 15km, does not cover wind turbine A, and has a score of 60.

[0143] 4. Update particle velocity and position: Adjust the path based on the individual optimal (the particle's own historical highest score) and the global optimal (the particle with the highest score in the current iteration). For example, if particle B learns from particle A, it will prioritize inspecting fan A and shorten the path.

[0144] 5. After 100 iterations, the optimal path is generated: the drones inspect the wind turbines in the order of "wind turbine A → wind turbine 3 → wind turbine 5" (total length 10km), the ground vehicles simultaneously inspect the central part of the photovoltaic area, and the robots are on standby during the mixed window period.

[0145] If analysis reveals a high recent failure rate of wind turbines in a certain area, a normal distribution model of equipment operating parameters is constructed by performing time-series analysis on the equipment status data of that area. The deviation of real-time parameters is calculated to determine this area as a priority inspection zone, and the inspection path is adjusted. The drone inspects according to the planned path, and when a suspected crack is detected on a wind turbine blade, a nacelle inspection robot is triggered for re-inspection and diagnosis.

[0146] The platform utilizes the YOLOv5 target detection algorithm to analyze drone images and identify anomalies such as wind turbine blade cracks and solar panel hot spots. It also performs Fourier transform analysis on vibration sensor data from the nacelle inspection robot to determine the operating status of equipment such as gearboxes.

[0147] When abnormal vibration of the wind turbine gearbox is detected, a maintenance work order is automatically generated, and the operation and maintenance personnel are notified via SMS. The equipment health record is also updated, and the inspection interval of the wind turbine is shortened.

[0148] The integrated management and control platform monitors the battery status of each inspection device in real time. For drones, when the battery level drops below 30%, path planning algorithms (such as Dijkstra's algorithm or A algorithm) are used to automatically plan a path back to the nearest charging base station. Simultaneously, other drones with sufficient battery power are scheduled to take over unfinished inspection tasks, ensuring inspection continuity. Dijkstra's algorithm and A algorithm can quickly find the shortest path from the current location to the charging base station in map data, improving drone return efficiency. For cabin inspection robots, since charging inside the cabin is inconvenient, the platform uses its power consumption prediction model and inspection task priority to reasonably schedule each working time to no more than 4 hours, ensuring the completion of inspection tasks and return to the charging area. The charging base station utilizes renewable energy sources such as solar and wind power generated by the new energy power station itself. Through energy conversion and storage devices (such as solar panels, wind turbines, and batteries), the energy is converted into electrical energy and stored, providing green and energy-saving energy supply for the equipment.

[0149] The platform monitors drone battery levels in real time. When a drone's battery level drops below 30%, it uses Dijkstra's algorithm to plan a path back to the nearest solar charging station and schedules other drones to take over the task. For cabin inspection robots, based on power consumption predictions, each work session is scheduled to last no more than 4 hours. The charging stations utilize solar panels from wind farms to provide green energy for the equipment.

[0150] The platform constructs a reinforcement learning model, using inspection efficiency, fault detection accuracy, and equipment energy consumption as evaluation metrics. Reinforcement learning is a machine learning method that involves an agent interacting with its environment, continuously trying and learning the optimal strategy, and is suitable for solving decision-making problems in complex and dynamic environments.

[0151] In the scenario of new energy power plant inspection, the model generates different collaborative strategies based on the current power plant status (equipment distribution, weather conditions, equipment status, etc.) and historical data, such as equipment scheduling order, inspection path planning, and task allocation ratio. By evaluating the evaluation index values ​​of each strategy, feedback is provided according to the reward function, with high rewards given to strategies that perform well and low rewards given to those that perform poorly. After multiple iterations of training, the model gradually learns to select the optimal collaborative strategy according to different scenarios, improving the system's adaptability and intelligence level.

[0152] The reinforcement learning model built on the platform uses inspection efficiency, fault detection accuracy, and equipment energy consumption as evaluation metrics. During training, the model continuously simulates different equipment scheduling and task allocation strategies, such as trying different equipment inspection sequences and path planning schemes. Based on the evaluation metric values ​​of each strategy, feedback is given according to the reward function. Strategies that improve inspection efficiency, reduce energy consumption, and ensure fault detection accuracy are given high rewards, while those that do not are given low rewards.

[0153] After multiple iterations of training, the model learns to select the optimal equipment combination and inspection strategy based on different weather conditions and equipment status. For example, in windy weather, it prioritizes the operation of cabin inspection robots and ground inspection vehicles, thereby improving the system's adaptability and intelligence.

[0154] This invention's multi-device collaborative control achieves cross-device command synchronization and conflict arbitration through a custom New Energy Protocol, establishing a priority queue and a "air-cabin-ground" closed-loop inspection mode. This embodiment can dynamically adapt to the environment, dividing multi-dimensional time windows based on meteorological data, and dynamically optimizing paths using the PSO algorithm and reinforcement learning, supporting priority inspection of abnormal areas and obstacle avoidance.

[0155] This embodiment integrates YOLOv5 image recognition and Fourier transform signal processing to automatically generate maintenance work orders and update health records, continuously optimizing strategies through reinforcement learning. This embodiment dynamically schedules charging tasks, utilizes renewable energy for power supply, and combines path planning algorithms (Dijkstra / A*) to reduce energy consumption, thus achieving energy management.

[0156] Those skilled in the art will clearly understand that the specific working process of the systems, devices, modules and units described above can be referred to the corresponding process in the foregoing method embodiments. For the sake of brevity, it will not be repeated here.

[0157] Furthermore, the functional units in the various embodiments of the present invention can be physically independent of each other, or two or more functional units can be integrated together, or all functional units can be integrated into one processing unit. The integrated functional units described above can be implemented in hardware, or in software or firmware.

[0158] Those skilled in the art will understand that if the integrated functional unit is implemented in software and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or all or part of it, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computing device (e.g., a personal computer, server, or network device) to execute all or part of the steps of the methods described in the embodiments of the present invention when running the instructions. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0159] Alternatively, all or part of the steps of the foregoing method embodiments can be implemented by hardware (such as a computing device, personal computer, server, or network device) related to program instructions. The program instructions can be stored in a computer-readable storage medium. When the program instructions are executed by the processor of the computing device, the computing device executes all or part of the steps of the methods described in the various embodiments of the present invention.

[0160] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that within the spirit and principles of the present invention, modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the corresponding technical solutions to depart from the protection scope of the present invention.

Claims

1. A method for collaborative control of multiple types of inspection equipment in a new energy power station, characterized in that, The method includes: Acquire topology data, equipment status data, meteorological data, and historical defect data of new energy power stations; Based on the aforementioned topology data, equipment status data, meteorological data, and historical defect data, a global inspection path is generated using a particle swarm optimization algorithm, which is then executed collaboratively by multiple types of inspection equipment. The inspection equipment includes drones, cabin inspection robots, and ground inspection vehicles, and the global inspection path includes at least a drone inspection path, a cabin inspection robot inspection path, and a ground inspection vehicle inspection path. Identify abnormal data in the equipment status data, determine the priority inspection area of ​​the new energy power station based on the abnormal data, and adjust the global inspection path to prioritize the coverage of the priority inspection area. The system monitors obstacle information within the inspection area of ​​the new energy power station in real time and dynamically adjusts the global inspection path based on the obstacle information.

2. The method according to claim 1, characterized in that, The topology data was acquired through three-dimensional laser scanning or satellite remote sensing technology; The equipment status data is collected in real time by various sensors installed on each piece of equipment in the new energy power station. The equipment status data includes equipment operating parameters, including wind turbine vibration value, gearbox temperature, and photovoltaic panel power generation efficiency. The meteorological data is obtained through monitoring by meteorological stations deployed at the site, including wind speed, light intensity, and rainfall. The historical defect data is stored in the historical defect database, which includes the types, locations, and handling status of previous equipment failures.

3. The method according to claim 1, characterized in that, Based on the aforementioned topology data, equipment status data, meteorological data, and historical defect data, the global inspection path is generated using the particle swarm optimization algorithm, including: Based on the aforementioned topological data, a three-dimensional model of the new energy power station is constructed; the three-dimensional model includes the wind turbine location coordinates, photovoltaic panel array distribution, booster station layout, and obstacle spatial coordinates. Based on the equipment status data, each device is assigned an inspection priority level; Based on the meteorological data, multi-dimensional inspection time windows are divided; the time windows include at least the UAV high-altitude inspection window, the cabin robot in-cabin operation window, and the ground vehicle inspection window. Based on the aforementioned 3D model, inspection priority levels, and multi-dimensional inspection time windows, a global inspection path is generated using a particle swarm optimization algorithm.

4. The method according to claim 3, characterized in that, Based on the aforementioned meteorological data, multi-dimensional inspection time windows are divided as follows: When the wind speed is less than or equal to the first set wind speed and the light intensity is greater than or equal to the first set light intensity, it is designated as the high-altitude inspection window for UAVs, and the flight altitude range of the UAVs within the window is the first set high zone range. When the wind speed is greater than the first set wind speed and less than or equal to the second set wind speed, or the light intensity is less than the first set light intensity and greater than or equal to the second set light intensity, a mixed inspection window is defined. The drone's flight altitude within the window is adjusted to the second set high zone, and the ground inspection vehicle's preprocessing path is started simultaneously. When the wind speed exceeds the second set wind speed or the rainfall exceeds the set rainfall, a priority window for robots is designated. Only cabin inspection robots and ground inspection vehicles are allowed to operate within this window, and the drone returns to the charging base station.

5. The method according to claim 1, characterized in that, After the global inspection path is executed collaboratively by multiple types of inspection equipment, the method further includes: Acquire inspection result data collected by inspection equipment; the inspection result data includes image data captured by drones, sensor detection data collected by cabin inspection robots, and / or ultrasonic detection data collected by ground inspection vehicles; New inspection tasks or maintenance work orders are generated based on the inspection results data, and the equipment health records are dynamically updated based on the inspection results data.

6. The method according to claim 1, characterized in that, The method further includes: Establish a cross-device command synchronization protocol, which includes at least priority queues for drone photography commands, robot sensor activation commands, and ground vehicle path switching commands; Based on the priority queue, when the drone detects an equipment abnormality, it generates a trigger command containing geographic coordinates and abnormality type. The trigger command takes priority over non-urgent commands in the current queue. After receiving the trigger command, the cabin inspection robot plans an in-cabin navigation path based on the geographic coordinates. The navigation path includes the infrared thermal imager detection path and the vibration sensor scanning points.

7. The method according to claim 6, characterized in that, The cross-device command synchronization protocol adopts a custom NewEnergy Protocol. The cross-device command synchronization protocol includes: Device identification field, used to identify a unique ID for a drone, robot, or ground vehicle; The task type field is used to distinguish between inspection tasks, re-inspection tasks, or emergency tasks; The data format field is used to specify the transmission protocol for visible light images, infrared thermal images, vibration data, and ultrasonic data. The conflict arbitration field is used to define the avoidance rules when multiple devices simultaneously apply for the same work area.

8. The method according to any one of claims 1-7, characterized in that, Identifying abnormal data in the device status data includes: Perform time-series analysis on the equipment status data to construct a normal distribution model of the equipment operating parameters; Calculate the deviation between the real-time parameters and the normal distribution model. When the deviation is greater than a preset threshold, it is marked as abnormal data.

9. The method according to claim 8, characterized in that, The preset threshold is dynamically updated based on fault parameter samples in the historical defect database, with an update cycle of once a day.

10. A comprehensive management and control platform, characterized in that, The integrated management and control platform includes a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the collaborative control method for multiple types of inspection equipment in new energy power stations according to any one of the instructions in the program code.

Images