Multi-scene adaptation method for electricity testing equipment of unmanned aerial vehicle

Abstract

The invention provides a multi-scene adaptation method for an unmanned aerial vehicle electricity testing device, and the method comprises the steps: constructing a whole-process simulation test environment of the unmanned aerial vehicle electricity testing device, carrying out the modeling of a mechanical structure, flight control, sensor perception and a data processing module of the device, simulating different working conditions in a virtual environment, carrying out the evaluation and optimization of the performance of the device, and carrying out the simulation of the whole-process simulation test environment. Verifying and perfecting the multi-scene adaptive knowledge base by utilizing a simulation result; the method comprises the following steps: designing a distributed big data processing architecture, processing massive multi-source heterogeneous data collected by electricity testing equipment by adopting a mode of combining edge computing and cloud computing, carrying out primary processing and compression of the data locally on an unmanned aerial vehicle, carrying out data convergence, storage and deep analysis mining at a cloud end, and updating model parameters in real time by utilizing an incremental learning algorithm. And dynamically expanding the knowledge base.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a multi-scenario adaptation method for unmanned aerial vehicle electrical testing equipment. Background Art

[0002] There are some technical contradictions and challenges when developing UAV electrical testing equipment for different voltage levels and line types. First, different voltage levels put forward different requirements for the insulation performance and withstand voltage level of UAVs, which require targeted design and material selection. Secondly, the difference in line types, such as overhead lines and cable lines, puts forward different requirements for the mechanical structure, flight control algorithm, sensor configuration, etc. of UAVs. In addition, the electrical testing equipment needs to work stably in complex and changing environments, such as electromagnetic interference, meteorological conditions, etc., which poses challenges to the reliability and adaptability of the equipment. At the same time, the amount of data collected by the electrical testing equipment is large and diverse, and there are also high requirements for data transmission, storage, and analysis capabilities. Finally, as a new type of equipment, UAV electrical testing equipment also faces problems such as standards and specifications, safety assessment, and operator training in actual applications. These technical contradictions need to be systematically analyzed and resolved, comprehensively considering factors such as equipment performance, applicable scenarios, and cost, optimizing the design plan, and conducting test verification, so as to finally achieve the reliable application of UAV electrical testing equipment in different voltage levels and line types. Summary of the invention

[0003] The present invention provides a multi-scenario adaptation method for an unmanned aerial vehicle electrical testing device, which mainly includes:

[0004] Construct a multi-scenario adaptive knowledge base for UAV electrical testing equipment, which includes insulation performance and withstand voltage requirements corresponding to different voltage levels, as well as mechanical structures, flight control algorithms and sensor configuration parameters corresponding to different line types. Adopt a hybrid reasoning method based on rules and cases to adaptively generate UAV electrical testing equipment configuration solutions that meet the requirements of the current scenario; construct a full-process simulation test environment for UAV electrical testing equipment, model the mechanical structure, flight control, sensor perception, and data processing modules of the equipment, simulate different working conditions in a virtual environment, evaluate and optimize equipment performance, and use simulation results to verify and improve the multi-scenario adaptive knowledge base; adopt multi-sensor Information fusion technology collects environmental information around the drone in real time through onboard multi-source heterogeneous sensors, performs spatiotemporal registration, feature extraction and semantic mapping of multi-source heterogeneous information, builds a unified environmental perception model, combines historical data and expert experience, and uses deep reinforcement learning algorithms to optimize the flight control strategy and equipment working parameters of the drone; designs a distributed big data processing architecture, uses a combination of edge computing and cloud computing to process the massive multi-source heterogeneous data collected by the electrical testing equipment, performs preliminary data processing and compression locally on the drone, aggregates, stores and conducts deep analysis and mining of data in the cloud, and uses incremental learning algorithms to update model parameters in real time and dynamically expand the knowledge base.

[0005] Furthermore, the construction of a multi-scenario adaptive knowledge base for unmanned aerial vehicle electrical test equipment includes: obtaining the voltage level and line type information of the current electrical test scenario, querying the corresponding insulation performance and withstand voltage level requirements in the knowledge base, and determining the basic configuration parameters of the unmanned aerial vehicle electrical test equipment according to the requirements; for the line type of the current electrical test scenario, obtaining the corresponding mechanical structure, flight control algorithm and sensor configuration parameters from the knowledge base as a supplement to the configuration plan of the unmanned aerial vehicle electrical test equipment; using a decision tree algorithm, according to the voltage level, insulation performance requirements, and withstand voltage level requirements of the current electrical test scenario, searching the knowledge base for historical electrical test cases with a high degree of matching, and obtaining the unmanned aerial vehicle electrical test equipment configuration plan for the case; using a fuzzy logic algorithm, matching each attribute value of the current electrical test scenario with the fuzzy rules stored in the knowledge base, calculating the membership of each configuration parameter, and selecting the parameter value with a higher membership as the configuration plan.

[0006] Furthermore, the construction of a full-process simulation test environment for the UAV electrical test equipment includes: obtaining the characteristics of the mechanical structure, flight control, sensor perception and data processing modules of the UAV electrical test equipment, and constructing corresponding three-dimensional models and mathematical models; integrating the three-dimensional models and mathematical models in the simulation environment; obtaining environmental parameters corresponding to different voltage levels and line types, and loading the environmental parameters into the simulation environment; setting the flight trajectory, speed and altitude parameters of the UAV electrical test equipment in the simulation environment; obtaining environmental models corresponding to meteorological conditions and electromagnetic interference, and loading the environmental models into the simulation environment; performing interference tests on the sensor perception and data processing modules of the UAV electrical test equipment in the simulation environment; and using machine learning algorithms to evaluate the flight control, sensor perception and data processing performance of the UAV electrical test equipment under different working conditions to obtain performance evaluation results.

[0007] Furthermore, the use of multi-sensor information fusion technology includes: obtaining multi-source heterogeneous information of the environment around the drone, the multi-source heterogeneous information including meteorological condition information and electromagnetic interference information; preprocessing the multi-source heterogeneous information to obtain preprocessed multi-source heterogeneous information, the preprocessing including data cleaning and spatiotemporal alignment; extracting key features from the preprocessed multi-source heterogeneous information to obtain a key feature set; using a feature selection algorithm to filter out a feature subset from the key feature set, the features in the feature subset are features that have a greater impact on environmental perception and flight control; semantically mapping the features in the feature subset to obtain a semantic feature set; constructing an environmental perception model based on the semantic feature set; obtaining historical flight data and environmental perception data of the drone to obtain a training data set.

[0008] Furthermore, the use of deep reinforcement learning algorithm to optimize the flight control strategy and equipment operating parameters of the drone includes: using a dual Q learning network, maximizing long-term cumulative rewards as the objective function, taking the training data set as input, and continuously interacting with the environment, using the flight control strategy and equipment operating parameters as the action space, and the output of the environmental perception model as the state space, training the dual Q learning network to obtain a flight control strategy model; deploying the flight control strategy model to the drone flight control system; the drone flight control system obtains the real-time environmental perception model output, and inputs the real-time environmental perception model output into the flight control strategy model to obtain the flight control strategy and equipment operating parameters under the current environment, and controlling the drone flight according to the flight control strategy and equipment operating parameters.

[0009] Furthermore, the design of a distributed big data processing architecture includes: obtaining multi-source heterogeneous data collected by the electrical testing equipment, the multi-source heterogeneous data including equipment operating parameters, environmental information and image data; transmitting the multi-source heterogeneous data to an edge computing node locally deployed on the drone; the edge computing node pre-processes the multi-source heterogeneous data, and obtains pre-processed data through data cleaning and feature extraction; transmitting the pre-processed data to a distributed big data processing platform in the cloud via a wireless network; the distributed big data processing platform aggregates and stores pre-processed data from different drones to build a unified data warehouse; using a distributed computing framework to perform parallel calculations on the data in the data warehouse, and extracting feature information from the data through a designed data mining algorithm.

[0010] Furthermore, the use of an incremental learning algorithm to update model parameters in real time and dynamically expand the knowledge base includes: training a machine learning model based on the feature information to obtain a fault diagnosis and prediction model; using an incremental learning algorithm to update the parameters of the fault diagnosis and prediction model in real time based on newly collected data and dynamically expand the knowledge base; deploying the trained fault diagnosis and prediction model to the cloud and edge computing nodes of the drone; the edge computing nodes and the cloud collaboratively perform reasoning calculations and generate alarm information and maintenance suggestions based on the reasoning results; the edge computing nodes and the cloud collaboratively perform reasoning calculations and generate alarm information if the reasoning results meet preset conditions.

[0011] Furthermore, the method also includes: developing a lightweight intelligent terminal for UAV electrical testing, using an embedded processor and an expandable storage architecture, integrating data acquisition, processing, storage, display and interaction functions; the terminal software adopts a modular design, including equipment status monitoring, fault diagnosis, and remote control modules; supporting multiple communication methods to achieve real-time data interaction and collaborative control with UAVs and cloud platforms; obtaining real-time status data of the UAV, pre-processing it through an embedded processor, eliminating abnormal values, and obtaining normalized status data; storing the normalized status data in an expandable storage architecture, and indexing it according to the timestamp and type of the data; using a machine learning algorithm to model the stored historical status data, and training an equipment fault diagnosis model for real-time judgment of whether the equipment is abnormal; if the diagnosis model determines that the equipment is faulty, a fault alarm is triggered, and the alarm information is sent to the remote control center through the UAV communication module.

[0012] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0013] The present invention discloses a multi-scenario adaptation method for unmanned aerial vehicle electrical testing equipment. The system establishes a knowledge base containing different voltage levels and line types, and uses a hybrid reasoning method to generate an adaptive configuration plan. Through full-process simulation testing, the performance of the equipment is evaluated and optimized in a virtual environment. Multi-sensor information fusion and deep reinforcement learning are used to perceive complex environments in real time and optimize control strategies. A distributed architecture combining edge computing and cloud computing is used to process massive data and dynamically update the knowledge base. A lightweight intelligent terminal is developed to achieve collaborative control with unmanned aerial vehicles and cloud platforms. The present invention improves the reliability, adaptability and intelligence level of unmanned aerial vehicle electrical testing equipment in complex environments, and provides a strong guarantee for the safe operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0015] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0016] like Figure 1 , this embodiment may specifically include:

[0017] S101. Establish a multi-scenario adaptive knowledge base for UAV electrical inspection equipment, including insulation performance and withstand voltage requirements for different voltage levels, as well as mechanical structures, flight control algorithms, and sensor configuration parameters for different line types. Adopt a hybrid reasoning method based on rules and cases, combined with decision trees and fuzzy logic, to adaptively generate UAV electrical inspection equipment configuration solutions that meet the requirements of the current scenario.

[0018] Obtain the voltage level and line type information of the current electrical test scenario, query the corresponding insulation performance and withstand voltage level requirements in the knowledge base, and determine the basic configuration parameters of the UAV electrical test equipment according to the requirements; for the line type of the current electrical test scenario, obtain the corresponding mechanical structure, flight control algorithm and sensor configuration parameters from the knowledge base as a supplement to the UAV electrical test equipment configuration plan; use the decision tree algorithm to search the knowledge base for the historical electrical test case with the highest matching degree according to the voltage level, insulation performance requirements, and withstand voltage level requirements of the current electrical test scenario, and obtain the UAV electrical test equipment configuration plan for the case; use the fuzzy logic algorithm to match the attribute values ​​of the current electrical test scenario with the fuzzy rules stored in the knowledge base, calculate the membership of each configuration parameter, and select the parameter value with the highest membership degree. as the final configuration scheme; merge the configuration schemes inferred by the decision tree algorithm and the fuzzy logic algorithm, take the average value of the parameters included in both schemes; directly add the parameters that only appear in one scheme to the final scheme; optimize the UAV's external dimensions and material strength according to the mechanical structure parameters corresponding to the line type, and improve its ability to adapt to different line environments; optimize the UAV's flight speed and route planning according to the flight control algorithm parameters, and improve its electrical testing efficiency; match the sensor configuration parameters stored in the knowledge base with the optimized mechanical structure and flight control scheme, determine the type, quantity, and installation location of the sensor, ensure that the equipment cost is reduced while meeting the electrical testing function requirements, and output the final UAV electrical testing equipment configuration scheme.

[0019] Specifically, obtaining the voltage level and line type information of the current power test scene is the first step in configuring the UAV power test equipment. For example, for a 220kV overhead transmission line, the corresponding insulation performance and voltage withstand level requirements need to be queried from the knowledge base. According to national standards, the insulator string of a 220kV line needs to withstand a 460kV power frequency withstand voltage. Based on this requirement, the insulation components of the UAV power test equipment need to reach a withstand voltage level of at least 500kV to ensure a safety margin. For overhead transmission lines, the mechanical structure of the drone needs to consider wind resistance. For example, it can still maintain stable flight in an environment with a wind speed of 10m / s. The flight control algorithm needs to implement an automatic obstacle avoidance function that can identify and bypass obstacles such as insulator strings on the transmission line. In terms of sensor configuration, high-precision infrared thermal imagers and ultraviolet imagers can be used to detect line hot spots and corona discharges. The decision tree algorithm can be used to quickly match historical power test cases. Assume that the current scenario is a 220kV overhead line, with an insulation requirement of 500kV and a withstand voltage level of 550kV. By screening layer by layer, it is possible to match a historical case with a similarity of 95%, and its drone configuration includes parameters such as carbon fiber fuselage, quad-rotor structure, and 1080P high-definition camera. The fuzzy logic algorithm infers the optimal configuration by setting fuzzy rules. For example, the rule "If the voltage level is high and the insulation requirements are strict, a high-performance composite fuselage is selected". It may be concluded that the membership of the carbon fiber fuselage is 0.8 and the membership of the aluminum alloy fuselage is 0.6, thus determining the selection of carbon fiber materials. Fusion of the results of the two algorithms may result in the following configuration: carbon fiber fuselage, quad-rotor structure, 500kV withstand voltage level, 1080P high-definition camera plus infrared thermal imager. This fusion method retains historical experience and takes into account the special needs of the current scenario. When optimizing the mechanical structure according to the line type, a foldable arm design can be considered to adapt to transmission corridors of different widths. In terms of material strength, carbon fiber composite materials with high strength and low density can be selected to improve wind resistance while reducing weight. The optimization of the flight control algorithm may include adaptive cruise speed adjustment. For example, the flight speed can be increased to 10m / s in straight sections, and automatically reduced to 5m / s at corners to ensure safety. Route planning can adopt a zigzag mode to ensure full coverage of the route. Finally, the sensor configuration needs to match the mechanical structure and flight control scheme. For example, a high-definition camera and an infrared thermal imager can be installed on the front of the fuselage, and a UV imager can be installed on the bottom. This configuration can not only meet the needs of all-round detection, but also maintain the balance of the center of gravity of the equipment, which is beneficial to flight stability. By optimizing the number and selection of sensors, such as using an integrated multispectral camera instead of separate visible light and infrared cameras, the equipment cost can be effectively reduced and the system structure can be simplified.

[0020] S102. Construct a full-process simulation test environment for UAV electrical testing equipment, and model the mechanical structure, flight control, sensor perception, and data processing modules of the equipment. Simulate different voltage levels, line types, meteorological conditions, and electromagnetic interference conditions in a virtual environment, evaluate and optimize equipment performance, and improve the robustness and practical application capabilities of the equipment. Use simulation results to verify and improve the multi-scenario adaptability knowledge base.

[0021] Obtain the characteristics of the mechanical structure, flight control, sensor perception and data processing modules of the UAV electrical test equipment, and construct corresponding three-dimensional models and mathematical models; integrate the three-dimensional model and mathematical model in a simulation environment; obtain environmental parameters corresponding to different voltage levels and line types, and load the environmental parameters into the simulation environment; set the flight trajectory, speed and altitude parameters of the UAV electrical test equipment in the simulation environment; obtain the environmental model corresponding to meteorological conditions and electromagnetic interference, and load the environmental model into the simulation environment; perform interference tests on the sensor perception and data processing modules of the UAV electrical test equipment in the simulation environment; use machine learning algorithms to evaluate the flight control, sensor perception and data processing modules of the UAV electrical test equipment under different working conditions and data processing performance to obtain a performance evaluation result; based on the performance evaluation result, optimize the flight control, sensor perception and data processing modules of the UAV electrical test equipment to obtain an optimized UAV electrical test equipment model; test the optimized UAV electrical test equipment model in the simulation environment to obtain equipment performance data; store the equipment performance data in a knowledge base and update the knowledge base; use a deep learning algorithm to mine and analyze the equipment performance data in the knowledge base to obtain the optimal working parameters of the UAV electrical test equipment in different scenarios; apply the optimal working parameters to the actual control of the UAV electrical test equipment; compare the test results of the UAV electrical test equipment in the simulation environment with the actual application results, and optimize the simulation model and knowledge base.

[0022] Specifically, the simulation test and optimization of UAV electrical inspection equipment is a complex system engineering, involving the integration of multiple technical fields. First, it is necessary to build a three-dimensional model and mathematical model of the equipment, which includes modules such as mechanical structure, flight control, sensor perception and data processing. Taking the mechanical structure as an example, computer-aided design software can be used to build accurate three-dimensional models of components such as carbon fiber fuselage and foldable arms. The flight control module needs to establish a mathematical model including attitude estimation, trajectory planning, dynamic response, etc. Integrating these models into the simulation environment is the next key step. The simulation environment can use platforms such as MATLAB Simu link, which can organically combine the models of each module to form a complete system simulation platform. On this basis, it is necessary to load environmental parameters corresponding to different voltage levels and line types. For example, for 220kV overhead transmission lines, parameters such as line height, tower spacing, and conductor sag can be set. These parameters directly affect the flight trajectory planning and obstacle avoidance strategy of the drone. Setting the flight trajectory, speed and height parameters is an important part of the simulation test. A variety of typical flight scenarios can be designed, such as straight line inspection, corner flight, obstacle avoidance, etc. For example, a cruising speed of 10m / s can be set in a straight section, and it will automatically drop to 5m / s at a corner. The flight altitude needs to be adjusted according to the line type and the requirements of the electrical inspection task, and is usually kept within 1-3 meters from the conductor. Meteorological conditions and electromagnetic interference are important factors affecting the performance of UAV electrical inspection equipment. Meteorological parameter models such as wind speed, temperature, and humidity, as well as electromagnetic interference models such as corona discharge and power frequency electric field can be loaded in the simulation environment. These models can simulate complex actual working environments and provide more realistic conditions for equipment performance testing. Interference testing of sensor perception and data processing modules in a simulation environment is an important means to evaluate the anti-interference ability of equipment. Electromagnetic interference of different intensities can be simulated to test the imaging quality of sensors such as high-definition cameras and infrared thermal imagers, as well as the anti-noise ability of data processing algorithms. This helps to optimize the shielding design of sensors and signal processing algorithms. Using machine learning algorithms to evaluate equipment performance is an efficient method. Algorithms such as support vector machines (SVM) can be used to comprehensively evaluate indicators such as flight control accuracy, sensor data quality, and fault detection accuracy. Through training with a large amount of simulation data, a model that can quickly evaluate equipment performance can be established. Optimizing the device model based on the evaluation results is a key step to improve performance. For example, if it is found that the flight stability is insufficient under strong wind conditions, the PID parameters in the flight control algorithm can be adjusted, or the mechanical structure design can be improved to enhance wind resistance. For the sensor perception module, it may be necessary to optimize the image processing algorithm to improve the recognition accuracy under adverse lighting conditions. The optimized device model needs to be tested again in a simulation environment to obtain new performance data. This data should be stored in a knowledge base for continuous updating and improvement of the device model.The construction of the knowledge base can adopt a relational database combined with unstructured data storage technology to efficiently manage and retrieve a large amount of heterogeneous data. Deep learning algorithms have unique advantages in mining and analyzing equipment performance data. Convolutional neural networks (CNNs) can be used to process image data, and long short-term memory networks (LSTMs) can be used to analyze time series data to obtain the optimal working parameters in different scenarios. These parameters may include optimal flight altitude, sensor sampling frequency, data fusion weights, etc. Applying the optimal working parameters to actual control is the implementation of simulation optimization results. This requires the establishment of an interface between the simulation environment and the actual control system to ensure that the parameters can be accurately and timely transmitted and updated. At the same time, it is also necessary to consider the unknown factors that may exist in the actual environment and design appropriate fault-tolerant mechanisms. Finally, comparing the simulation results with the actual application results and optimizing the simulation model and knowledge base accordingly is a process of continuous improvement. This may involve fine-tuning the simulation model parameters, optimizing the knowledge base structure, and introducing new data collection and analysis methods. Through this closed-loop optimization, the accuracy and practicality of the simulation can be continuously improved, and the performance of the UAV electrical inspection equipment can be continuously improved.

[0023] S103. Use multi-sensor information fusion technology to collect environmental information around the drone in real time, including meteorological conditions and electromagnetic interference, through onboard multi-source heterogeneous sensors. Perform spatiotemporal registration, feature extraction and semantic mapping of multi-source heterogeneous information to build a unified environmental perception model. Combine historical data and expert experience, use deep reinforcement learning algorithms to optimize the flight control strategy and equipment operating parameters of the drone. Use a dual Q learning network structure, with maximizing long-term cumulative rewards as the objective function, and continuously adjust the policy network parameters through interaction with the environment to improve reliability and adaptability in complex environments.

[0024] Acquire multi-source heterogeneous information of the environment around the drone, wherein the multi-source heterogeneous information includes meteorological condition information and electromagnetic interference information; preprocess the multi-source heterogeneous information to obtain preprocessed multi-source heterogeneous information, wherein the preprocessing includes data cleaning and spatiotemporal registration; extract key features from the preprocessed multi-source heterogeneous information to obtain a key feature set; use a feature selection algorithm to filter out a feature subset from the key feature set, wherein the features in the feature subset are features whose impact on environmental perception and flight control is greater than a preset threshold; perform semantic mapping on the features in the feature subset to obtain a semantic feature set; construct an environmental perception model based on the semantic feature set; acquire historical flight data and environmental perception data of the drone, and perform data annotation and Enhancement to obtain a training data set; adopt a double Q learning network, take long-term cumulative reward maximization as the objective function, take the training data set as input, through continuous interaction with the environment, use flight control strategy and equipment working parameters as action space, and use the output of the environment perception model as state space to train the double Q learning network to obtain an optimal flight control strategy model; deploy the optimal flight control strategy model to the UAV flight control system; the UAV flight control system obtains the real-time environment perception model output, and inputs the real-time environment perception model output into the optimal flight control strategy model to obtain the optimal flight control strategy and equipment working parameters under the current environment, and control the flight of the UAV according to the optimal flight control strategy and equipment working parameters.

[0025] Specifically, when the UAV electrical testing equipment works in a complex environment, it needs to obtain multi-source heterogeneous information to accurately perceive the surrounding environment. This information includes meteorological conditions and electromagnetic interference. For example, meteorological condition information may include temperature, humidity, wind speed and wind direction, while electromagnetic interference information may include electromagnetic field strength, frequency and direction. Preprocessing these multi-source heterogeneous information is crucial. Data cleaning can remove outliers and noise, such as eliminating obviously unreasonable temperature readings or electromagnetic field strength data. The spatiotemporal matching criteria ensure that data from different sources are aligned in time and space, such as synchronizing the data from the weather station with the data collected by the drone in real time. Extracting key features from the preprocessed data is the basis for building an environmental perception model. For example, the mean and standard deviation of wind speed, the peak and spectral characteristics of electromagnetic field strength, etc. can be extracted. Feature selection algorithms such as principal component analysis or recursive feature elimination can help screen out the most important feature subsets, such as selecting the wind speed feature that has the greatest impact on flight stability and the electromagnetic interference feature that has the most significant impact on the working accuracy of the equipment. Semantic mapping of the screened features can improve the interpretability of the model. For example, the wind speed feature is mapped to semantic labels such as "strong wind" and "light breeze", and the electromagnetic interference feature is mapped to semantic concepts such as "strong interference" and "weak interference". The environmental perception model constructed in this way is easier to understand and apply. In order to train the flight control strategy model, a large amount of historical flight data and environmental perception data needs to be collected and annotated. Data enhancement techniques such as rotation, scaling, and adding noise can expand the training data set and improve the generalization ability of the model. The use of a dual Q learning network to train the flight control strategy model can effectively solve the over-estimation problem in traditional reinforcement learning. The network uses two Q networks to update alternately, one for selecting actions and the other for evaluating action values, thereby improving the stability and efficiency of learning. After the optimal flight control strategy model obtained by training is deployed to the UAV flight control system, the flight parameters and equipment working parameters can be dynamically adjusted according to the real-time environmental perception results. For example, in a strong wind environment, the model may output control strategies such as lowering the flight altitude and increasing the hovering power; in a strong electromagnetic interference environment, the model may suggest adjusting the sensor sampling frequency or switching to anti-interference mode. This adaptive control method based on environmental perception and reinforcement learning can significantly improve the working efficiency and reliability of UAV electrical testing equipment in complex environments. Through continuous learning and optimization, the system can continuously accumulate experience, cope with various complex working environments, and provide strong guarantees for the safe operation of the power system.

[0026] S104. Design a distributed big data processing architecture, and use a combination of edge computing and cloud computing to process the massive multi-source heterogeneous data collected by the electrical testing equipment. Perform preliminary data processing and compression locally on the drone to reduce data transmission volume and improve real-time performance. Aggregate, store, and deeply analyze and mine data in the cloud to extract valuable feature information. Use incremental learning algorithms, such as online stochastic gradient descent and adaptive moment estimation, to update model parameters in real time, dynamically expand the knowledge base, and adapt to environmental changes.

[0027] Acquire multi-source heterogeneous data collected by the electrical testing equipment, the multi-source heterogeneous data including equipment operating parameters, environmental information and image data; transmit the multi-source heterogeneous data to the edge computing node deployed locally on the UAV; the edge computing node pre-processes the multi-source heterogeneous data, and obtains the pre-processed data through data cleaning and feature extraction; transmit the pre-processed data to the distributed big data processing platform in the cloud through a wireless network; the distributed big data processing platform aggregates and stores the pre-processed data from different UAVs to build a unified data warehouse; use a distributed computing framework to perform parallel calculations on the data in the data warehouse, and extract feature information from the data through a designed data mining algorithm; train a machine learning model based on the feature information to obtain a fault diagnosis and prediction model; use an incremental learning algorithm to update the parameters of the fault diagnosis and prediction model in real time based on the newly collected data, and dynamically expand the knowledge base; deploy the trained fault diagnosis and prediction model to the edge computing nodes of the cloud and the UAV; the edge computing node and the cloud collaborate to perform reasoning calculations, and generate alarm information and maintenance suggestions based on the reasoning results.

[0028] Specifically, the multi-source heterogeneous data collected by the electrical testing equipment include equipment operating parameters, environmental information and image data. Equipment operating parameters may include indicators such as voltage, current, and temperature; environmental information may include humidity, wind speed, etc.; image data may be photos of the appearance or internal structure of the equipment. These data are wirelessly transmitted to the edge computing node on the drone for preliminary processing. The edge computing node preprocesses the data, including data cleaning and feature extraction. Data cleaning may involve operations such as removing outliers and filling missing values. For example, if the temperature sensor occasionally has obviously unreasonable readings, such as -50℃, it can be removed or replaced with adjacent reasonable values. Feature extraction may include calculating statistics, extracting image features, etc. For example, the average and standard deviation of voltage over a period of time can be calculated, or features such as edges and textures can be extracted from the equipment image. The preprocessed data is transmitted to the distributed big data processing platform in the cloud via a wireless network. The platform aggregates data from multiple drones to build a unified data warehouse. The advantage of this is that the operation status of the entire power system can be comprehensively analyzed and potential systemic problems can be discovered. Distributed computing frameworks such as Hadoop or Spark can perform parallel computing on massive data in the data warehouse. Through the designed data mining algorithm, deeper feature information can be extracted. For example, the typical pattern of equipment operating parameters can be discovered through clustering algorithms, or the correlation between parameters can be discovered through association rule mining algorithms. Based on the extracted feature information, machine learning models can be trained for fault diagnosis and prediction. For example, a support vector machine (SVM) can be used to build a classifier to predict whether the equipment is likely to fail based on the equipment operating parameters and environmental information. Or a long short-term memory network (LSTM) can be used to predict the operating status of the equipment for a period of time in the future. In order to adapt to the dynamic changes in equipment status and environment, an incremental learning algorithm is used to update model parameters in real time. For example, when the newly collected data shows a different pattern from the historical data, the model can be quickly adjusted to adapt to this change, thereby improving the accuracy of diagnosis and prediction. The trained model is deployed on the cloud and edge computing nodes at the same time. This collaborative reasoning method can not only use the powerful computing power of the cloud to handle complex tasks, but also quickly respond to emergencies through edge computing. For example, when the edge node detects a sudden abnormal voltage, an alarm can be issued immediately; while the cloud can combine historical data and the status of other equipment to conduct a deeper analysis of the cause of the abnormality. Finally, the system generates alarm information and maintenance suggestions based on the reasoning results. For example, if the transformer temperature is detected to be continuously rising and exceeding the threshold, the system may issue an overheating warning and recommend checking the cooling system or reducing the load. This information can help operation and maintenance personnel to detect potential problems in a timely manner and take preventive measures, thereby improving the reliability and safety of the power system.

[0029] S105. Research and develop lightweight intelligent terminals for UAV electrical testing, using low-power embedded processors and scalable storage architecture, integrating data acquisition, processing, storage, display and interaction functions. The terminal software adopts a modular design, including equipment status monitoring, fault diagnosis, and remote control modules. Support multiple communication methods, such as narrowband Internet of Things and low-power wide area network, to achieve real-time data interaction and collaborative control with UAVs and cloud platforms, and improve operational efficiency and intelligence.

[0030] The real-time status data of the drone, including parameters such as voltage, current, and temperature, is obtained. The data is preprocessed through a low-power embedded processor to remove outliers and obtain normalized status data. The normalized status data is stored in an extensible storage architecture and indexed according to the timestamp and type of the data to facilitate subsequent query and analysis. The stored historical status data is modeled using a machine learning algorithm, and a device fault diagnosis model is trained through algorithms such as support vector machines to determine in real time whether the device is abnormal. If the diagnostic model determines that the device is faulty, a fault alarm is triggered, and the alarm information is sent to the remote control center through the drone communication module to notify the operation and maintenance personnel to handle it in time. After receiving the fault alarm of the drone, the remote control center sends control instructions to the drone through communication methods such as low-power wide area networks to adjust its operating parameters or perform emergency operations to ensure the safety of the equipment. According to the actual operation requirements of the drone, the data collection strategy of the intelligent terminal is dynamically configured. By adjusting the sampling frequency of the sensor, data compression ratio and other parameters, the transmission and storage overhead are reduced while ensuring data quality. A two-way communication link is established between the intelligent terminal and the cloud platform, and the collected data is regularly uploaded to the cloud platform for summary and analysis. At the same time, the latest operation tasks and control strategies are obtained from the cloud platform to achieve end-cloud collaboration and improve operation efficiency and intelligence level.

[0031] The above embodiment is only one of the preferred implementation modes of the present invention and should not be used to limit the protection scope of the present invention. Any changes or modifications that are made to the main design concept and spirit of the present invention and have no substantive significance, and the technical problems they solve are still consistent with the present invention, should be included in the protection scope of the present invention.

Claims

1. A multi-scenario adaptation method for unmanned aerial vehicle electrical testing equipment, characterized in that: include: Construct a multi-scenario adaptive knowledge base for UAV electrical testing equipment. The knowledge base includes insulation performance and withstand voltage requirements corresponding to different voltage levels, as well as mechanical structures, flight control algorithms, and sensor configuration parameters corresponding to different line types. A hybrid reasoning method based on rules and cases is used to adaptively generate UAV electrical testing equipment configuration solutions that meet the requirements of the current scenario. Construct a full-process simulation test environment for UAV electrical testing equipment, model the mechanical structure, flight control, sensor perception, and data processing modules of the equipment, simulate different working conditions in a virtual environment, evaluate and optimize equipment performance, and use simulation results to verify and improve the multi-scenario adaptability knowledge base; Adopt multi-sensor information fusion technology, collect environmental information around the drone in real time through multi-source heterogeneous sensors on board, perform spatiotemporal registration, feature extraction and semantic mapping of multi-source heterogeneous information, build a unified environmental perception model, combine historical data and expert experience, and use deep reinforcement learning algorithms to optimize the flight control strategy and equipment working parameters of the drone; A distributed big data processing architecture is designed, and a combination of edge computing and cloud computing is used to process the massive multi-source heterogeneous data collected by the electrical testing equipment. The data is initially processed and compressed locally on the drone, and the data is aggregated, stored, and deeply analyzed and mined in the cloud. The incremental learning algorithm is used to update the model parameters in real time and dynamically expand the knowledge base.

2. The method according to claim 1, characterized in that The multi-scenario adaptability knowledge base of the drone electrical testing equipment is constructed, including: Obtain the voltage level and line type information of the current electrical testing scene, query the corresponding insulation performance and withstand voltage level requirements in the knowledge base, and determine the basic configuration parameters of the UAV electrical testing equipment according to the requirements; According to the line type of the current electrical inspection scenario, the corresponding mechanical structure, flight control algorithm and sensor configuration parameters are obtained from the knowledge base as a supplement to the configuration scheme of the UAV electrical inspection equipment; A decision tree algorithm is used to search for historical electrical testing cases with a high matching degree in the knowledge base according to the voltage level, insulation performance requirements, and withstand voltage level requirements of the current electrical testing scenario, and a configuration plan for the UAV electrical testing equipment of the case is obtained; A fuzzy logic algorithm is used to match the attribute values ​​of the current electrical testing scenario with the fuzzy rules stored in the knowledge base, the membership degree of each configuration parameter is calculated, and the parameter value with a higher membership degree is selected as the configuration solution.

3. The method according to claim 1, characterized in that The construction of the full-process simulation test environment of the UAV electrical inspection equipment includes: Obtain the characteristics of the mechanical structure, flight control, sensor perception and data processing modules of the UAV electrical testing equipment, and build the corresponding three-dimensional model and mathematical model; Integrating the three-dimensional model and the mathematical model in a simulation environment; Acquire environmental parameters corresponding to different voltage levels and line types, and load the environmental parameters into the simulation environment; Setting the flight trajectory, speed and altitude parameters of the UAV electrical testing device in the simulation environment; Acquire an environmental model corresponding to meteorological conditions and electromagnetic interference, and load the environmental model into the simulation environment; Performing interference testing on the sensor perception and data processing modules of the UAV electrical testing equipment in the simulation environment; A machine learning algorithm is used to evaluate the flight control, sensor perception and data processing performance of the UAV electrical testing equipment under different working conditions to obtain performance evaluation results.

4. The method according to claim 1, characterized in that The multi-sensor information fusion technology includes: Acquire multi-source heterogeneous information of the environment around the drone, wherein the multi-source heterogeneous information includes meteorological condition information and electromagnetic interference information; Preprocessing the multi-source heterogeneous information to obtain preprocessed multi-source heterogeneous information, wherein the preprocessing includes data cleaning and spatiotemporal registration; Extract key features from the preprocessed multi-source heterogeneous information to obtain a key feature set; Using a feature selection algorithm to select a feature subset from the key feature set, the features in the feature subset are features that have a greater impact on environmental perception and flight control; Performing semantic mapping on the features in the feature subset to obtain a semantic feature set; Building an environment perception model according to the semantic feature set; Obtain the drone’s historical flight data and environmental perception data to obtain a training data set.

5. The method according to claim 1, characterized in that The use of deep reinforcement learning algorithms to optimize the flight control strategy and equipment operating parameters of the drone includes: A dual Q learning network is adopted, and the long-term cumulative reward maximization is used as the objective function. The training data set is used as input, and the flight control strategy and the equipment working parameters are used as the action space, and the output of the environment perception model is used as the state space to train the dual Q learning network to obtain a flight control strategy model through continuous interaction with the environment; Deploy the flight control strategy model into the UAV flight control system; The UAV flight control system obtains the real-time environment perception model output, and inputs the real-time environment perception model output into the flight control strategy model, obtains the flight control strategy and equipment operating parameters under the current environment, and controls the UAV flight according to the flight control strategy and equipment operating parameters.

6. The method according to claim 1, characterized in that The design of the distributed big data processing architecture includes: Acquire multi-source heterogeneous data collected by an electrical testing device, wherein the multi-source heterogeneous data includes device operating parameters, environmental information, and image data; Transmitting the multi-source heterogeneous data to an edge computing node locally deployed on the drone; The edge computing node preprocesses the multi-source heterogeneous data to obtain preprocessed data through data cleaning and feature extraction; Transmitting the preprocessed data to a distributed big data processing platform in the cloud via a wireless network; The distributed big data processing platform aggregates and stores pre-processed data from different drones to build a unified data warehouse; A distributed computing framework is used to perform parallel computing on the data in the data warehouse, and characteristic information in the data is extracted through a designed data mining algorithm.

7. The method according to claim 6, characterized in that The method of using an incremental learning algorithm to update model parameters in real time and dynamically expand the knowledge base includes: According to the feature information, a machine learning model is trained to obtain a fault diagnosis and prediction model; Adopting an incremental learning algorithm, the parameters of the fault diagnosis and prediction model are updated in real time according to the newly collected data, and the knowledge base is expanded dynamically; Deploy the trained fault diagnosis and prediction models to the cloud and the edge computing nodes of drones; The edge computing node and the cloud collaborate to perform reasoning calculations and generate alarm information and maintenance suggestions based on the reasoning results; The edge computing node and the cloud collaborate to perform reasoning calculations, and if the reasoning results meet preset conditions, an alarm message is generated.

8. The method according to claim 1, characterized in that The method further comprises: Develop a lightweight intelligent terminal for UAV electrical testing, using an embedded processor and scalable storage architecture, integrating data acquisition, processing, storage, display and interaction functions; The terminal software adopts modular design, including equipment status monitoring, fault diagnosis, and remote control modules; Support multiple communication methods to achieve real-time data interaction and collaborative control with drones and cloud platforms; Obtain the real-time status data of the drone, pre-process it through the embedded processor, remove abnormal values, and obtain standardized status data; Store normalized state data in a scalable storage architecture, indexed by timestamp and type of data; Use machine learning algorithms to model the stored historical status data and train equipment fault diagnosis models to determine in real time whether equipment is abnormal; If the diagnostic model determines that the equipment is faulty, a fault alarm is triggered and the alarm information is sent to the remote control center through the drone communication module.

Images