A UAV-assisted inspection method and system for power transmission lines

Abstract

The present invention discloses a transmission line UAV assisted inspection method and system, which relates to the field of UAV technology, including collecting environmental data from the UAV global positioning system, inertial measurement unit and camera sensor to form an environmental status data set; based on the environmental status data set, using a dynamic environmental perception adaptation algorithm to generate real-time environmental perception data; training the autonomous learning algorithm of the UAV through the high-precision transmission line digital environment model to generate a flight strategy model; real-time fusion of the real-time environmental perception data and the flight strategy model to generate an optimal path planning strategy, so that the UAV can adjust flight parameters according to the real-time environment and optimize the flight path. The present invention integrates multi-sensor data and dynamic environmental perception technology to enable the UAV to self-adjust the flight mode according to the immediate environmental changes, realizes the ability to cope with complex environments, reduces inspection costs, and improves work efficiency and safety.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicles, and in particular to a method and system for unmanned aerial vehicle-assisted inspection of power transmission lines. Background Art

[0002] With the expansion of power infrastructure and the increase in maintenance needs, transmission line inspection has become a key link in ensuring the safe operation of the power grid. Traditional manual inspection methods are not only time-consuming and labor-intensive, but also inefficient when facing complex terrain and severe weather conditions, and there are safety hazards. Although the introduction of drones has initially improved this situation, most existing drone inspection systems lack the ability to dynamically adapt to complex and changing environments, especially in terms of real-time perception of environmental changes, intelligent path planning, and autonomous decision-making. Especially under the influence of environmental factors such as wind speed changes, temporary obstacles, and subtle changes in the terrain around the line, it is difficult to accurately perceive the surrounding conditions and make quick and accurate adjustments to the flight strategy, resulting in inspection efficiency and safety cannot be fully guaranteed. Summary of the invention

[0003] In view of the above existing problems, the present invention is proposed.

[0004] Therefore, the present invention provides a transmission line drone-assisted inspection method and system to solve the problems of existing inspection methods such as insufficient adaptability in complex environments, poor path planning flexibility, and slow response to emergencies.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] In the first aspect, an embodiment of the present invention provides a transmission line UAV-assisted inspection method, which includes collecting environmental data from a UAV global positioning system, an inertial measurement unit, and a camera sensor to form an environmental status data set; based on the environmental status data set, using a dynamic environmental perception adaptation algorithm to generate real-time environmental perception data; using a geographic information system and historical line data to construct a high-precision digital environment model of the transmission line; using the high-precision digital environment model of the transmission line to train the UAV's autonomous learning algorithm to generate a flight strategy model; and real-time fusion of the real-time environmental perception data and the flight strategy model to generate an optimal path planning strategy, so that the UAV can adjust flight parameters according to the real-time environment and optimize the flight path.

[0007] As a preferred solution of the transmission line drone-assisted inspection method described in the present invention, the environmental status data set includes positioning and navigation data, vision and image data, environmental parameters, terrain and landform information, endurance status and sensor health data, and external interference data.

[0008] As a preferred solution of the transmission line drone-assisted inspection method of the present invention, wherein: based on the environmental state data set, a dynamic environmental perception adaptation algorithm is used to generate real-time environmental perception data, and the specific steps are:

[0009] The collected environmental status data set is preprocessed by noise filtering, missing value filling and data standardization;

[0010] Different types of feature extraction are performed on each preprocessed data. The positioning and navigation data are extracted through filtering processing, the visual and image data are extracted through image enhancement and directional gradient histogram algorithm, the environmental parameters are extracted through mean and correlation analysis, the terrain and landform information are extracted through digital elevation model terrain slope and slope aspect analysis, the endurance status and sensor health data are extracted through time series analysis and independent component analysis, and the external interference data are extracted through spectrum analysis and buffer zone analysis. The extracted features are integrated to generate the feature vector expression:

[0011] F=[f1,f2,...,f n ]

[0012] Among them, F is a vector composed of multiple features, f1,f2,...,f n are the characteristic elements in the vector F, and n is the total number of features in the characteristic vector F;

[0013] Based on the feature vector, a dynamic evaluation function is constructed to quantify the degree of dynamic change of the current environment. The expression of the dynamic evaluation function is:

[0014]

[0015] Among them, Φ is the dynamic change degree of the current environment, For each feature f i There is a corresponding weight w i , the square is used as a normalized representation, is the standard deviation of the ith feature, μ i is the mean, f i (t) is the specific value of the i-th feature changing with time t, t is the time index variable, f i is the i-th feature in the feature vector F, i is the index variable, and dt is the integral of the change over time t;

[0016] By adjusting the function to map the dynamic evaluation results to the parameter configuration space, real-time monitoring is performed and the perception parameters are dynamically adjusted. The dynamic adjustment function expression is:

[0017]

[0018] Among them, Γ(F) is the perception parameter value after real-time adjustment, α, β, γ are adjustment coefficients;

[0019] Combined with the perception parameter values ​​adjusted in real time, various sensor data are processed instantly to generate real-time environmental perception data that integrates multi-source information.

[0020] As a preferred solution of the transmission line drone-assisted inspection method of the present invention, wherein: the use of geographic information system and historical line data to construct a high-precision transmission line digital environment model, the specific steps are:

[0021] Data cleaning, coordinate registration and data preprocessing in a unified format are performed on the digital elevation model, land cover type, slope, and aspect terrain data in the geographic information system, as well as the route direction, height, and surrounding environment records in the historical route data, to form a comprehensive terrain and route analysis data set;

[0022] Extract features of digital elevation model, land cover type, and historical route compatibility from the comprehensive terrain and route analysis dataset, generate numerical data of slope degree, normalized vegetation index, category labels of land cover type, and evaluation indicators of route and terrain matching scores;

[0023] Integrate the numerical data, category labels, and evaluation indicators to form a comprehensive terrain route characteristic indicator set;

[0024] Using the comprehensive terrain route characteristic index set, the multi-scale terrain feature fusion expression is constructed as follows:

[0025]

[0026] Among them, H is the fused high-precision terrain adaptability score, S is the number of scales, ω s is the weight of the Sth scale, N s is the characteristic number at the Sth scale, φ si is the normalized value of the i-th feature at the S-th scale, G si is the matching degree between the ith feature and the historical route data at the Sth scale;

[0027] Based on the fused high-precision terrain adaptability score H and combined with historical route data, the adaptive path optimization algorithm expression is constructed as follows:

[0028]

[0029] Among them, P opt For the optimized flight path, To find p that makes the expression in the brackets reach the minimum value, λ1, λ2, λ3 are the first, second, and third weight factors respectively, C(P) is the cost function, D(P,H) is the difference in adaptability between the path and the high-precision terrain, and U(P) is the safety evaluation function;

[0030] The cost function expression is:

[0031] C(P)=w1·L(P)+w2·V(P)+w3·M(P)

[0032] Among them, w1, w2, w3 are weight factors, L(P) is the path length, V(P) is the flight time cost, and M(P) is the maintenance cost;

[0033] The security assessment function expression is:

[0034] E(P)=w4·O(P)+w5·W(P)+w6·K A (P)

[0035] Among them, w4, w5, w6 are weight factors, O(P) is the obstacle risk on the path, W(P) is the impact of weather conditions, K A (P) is the terrain adaptability score;

[0036] The optimized flight path P opt It is superimposed with terrain data and historical line data in the geographic information system in three-dimensional space to build a high-precision digital environment model of the transmission line.

[0037] As a preferred solution of the transmission line drone-assisted inspection method of the present invention, the specific steps of constructing a high-precision transmission line digital environment model are as follows:

[0038] Through the geographic information system, all data of the optimized flight path, terrain data, and historical route data are converted and formatted uniformly;

[0039] Using the known coordinate information, spatially align the terrain data after data conversion and format unification and the optimized flight path, and geo-align the historical route data;

[0040] Based on the digital elevation model data in the geographic information system, the discrete elevation points in the digital elevation model data are converted into a continuous terrain surface through interpolation method and rendered to generate a three-dimensional terrain model;

[0041] Converting the optimized flight path and historical route data into three-dimensional lines and superimposing them on the three-dimensional model;

[0042] The professional 3D visualization software Unity is used to visualize the 3D model and generate a high-precision digital environment model of the transmission line.

[0043] As a preferred solution of the transmission line UAV assisted inspection method of the present invention, wherein: the autonomous learning algorithm of the UAV is trained through the high-precision transmission line digital environment model to generate a flight strategy model, and the specific steps are:

[0044] Extract key features of terrain undulation, line structure, obstacle distribution, and historical data of meteorological conditions from the high-precision digital environment model of power transmission lines, and perform outlier removal and standardized data preprocessing;

[0045] The weighted frequency feature similarity function is constructed through the preprocessed data to dynamically adjust the weight of each feature to reflect the importance change of each factor in the current flight environment. The weighted frequency feature similarity function expression is:

[0046]

[0047] Among them, W j is the weight value of the jth feature, E t,j is the energy value of the jth feature at time t, is the energy mean of the jth feature, σ j is the energy standard deviation of the jth feature, is the square root of the time series length T, j' is the traversal of all feature indexes j, and M is the total number of feature indexes j;

[0048] For the weighted frequency feature similarity function, adaptive multi-objective deep reinforcement learning is used to optimize environmental adaptability, safety and efficiency during flight;

[0049] The comprehensive objective function is used to integrate the adaptive multi-objective deep reinforcement learning into a unified evaluation standard. The expression of the comprehensive objective function is:

[0050]

[0051] Among them, Q * (se,a) is the optimal cumulative discounted reward that can be obtained by taking an action a in a given state, se is the current state, a is the action taken, Q * For optimization, To find the optimal strategy among all strategies π, is the expected value of the expression in brackets, γ t is the discount factor, R(se t ,a t ,s et+1 ) is the immediate reward function, W'j (t) is the W j The weight value of the jth feature in will be dynamically adjusted over time t;

[0052] The instant reward function expression is:

[0053] R(se t ,a t ,s et+1 )=w1R 安全 +w2R 效率 +w3R 任务 +w4R 适应性

[0054] Among them, R(se t ,a t ,s t+1 ) is the immediate reward function, se t is the current state of the drone, a t The current state of the drone t The execution action below, t+1 To perform action a t After that, the drone reaches a new state in response to the environment, w1, w2, w3, w4 are weight factors, R 安全 For drone safety rewards, R 效率 is the drone efficiency bonus, R 任务 The reward for the completion of the drone mission, R 适应性 Bonus for drone adaptability;

[0055] Use real-time data to continuously evaluate the dynamic changes of the flight environment and adjust the feature weights W j To ensure that the flight strategy model can adapt to the new environment in real time and generate a flight strategy model.

[0056] As a preferred solution of the transmission line UAV-assisted inspection method of the present invention, the real-time environmental perception data and the flight strategy model are integrated in real time to generate the optimal path planning strategy. The specific steps are as follows:

[0057] Establishing a data interface to transmit the real-time environment perception data to the flight strategy model processing module;

[0058] Based on the feature vector F=[f1,f2,...,f n ], using the dynamic evaluation function φ(F) to calculate the degree of dynamic change of the environment;

[0059] According to the environmental dynamics assessment result φ, the perception parameters of the UAV are optimized in real time by dynamically adjusting the function Γ(φ);

[0060] Based on real-time optimization of the drone’s perception parameters, adaptive multi-objective deep reinforcement learning and comprehensive objective functions are used to dynamically adjust the feature weights W in the flight strategy model. j , forming an updated flight strategy model;

[0061] The updated flight strategy model is applied to the UAV path planning, and the flight path P is recalculated in combination with the real-time UAV optimized perception data. opt , ensuring that real-time data feedback can immediately adjust the flight strategy and generate the best path planning strategy.

[0062] In the second aspect, the present invention provides a transmission line UAV assisted inspection system, including an environmental data acquisition module, a dynamic environment perception adaptation processing module, a high-precision digital environment modeling module, a flight strategy algorithm training module, and a real-time path planning and flight control fusion module; the environmental data acquisition module is used to collect environmental data from the UAV global positioning system, inertial measurement unit and camera sensor to form an environmental state data set; the dynamic environment perception adaptation processing module is used to generate real-time environmental perception data based on the environmental state data set using a dynamic environment perception adaptation algorithm; the high-precision digital environment modeling module is used to use a geographic information system and historical line data to construct a high-precision digital environment model of the transmission line; the flight strategy algorithm training module is used to train the UAV's autonomous learning algorithm through the high-precision digital environment model of the transmission line to generate a flight strategy model; the real-time path planning and flight control fusion module is used to fuse the real-time environmental perception data and the flight strategy model in real time to generate an optimal path planning strategy, so that the UAV can adjust flight parameters according to the real-time environment and optimize the flight path.

[0063] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the transmission line drone-assisted inspection method as described in the first aspect of the present invention is implemented.

[0064] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the transmission line drone-assisted inspection method as described in the first aspect of the present invention is implemented.

[0065] The beneficial effects of the present invention are as follows: the present invention realizes fine monitoring and real-time analysis of the flight environment by comprehensively utilizing multiple data sources such as global positioning, inertial navigation, visual sensing, etc., combined with a dynamic environment perception algorithm, so as to accurately evaluate the dynamic changes of the environment and automatically adjust the perception parameters. On this basis, the constructed high-precision digital environment model not only integrates geographic information and historical route data, but also generates a flight strategy model through a deep learning algorithm, so that the UAV can make intelligent decisions based on the real-time environment, optimize the flight path, and complete intelligent and adaptive inspection strategies, which significantly improves the ability to cope with complex environments, reduces inspection costs, improves work efficiency and safety, and at the same time ensures the stable operation of power facilities, solving the technical problems of low efficiency, poor safety, and weak adaptability of traditional inspection methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0067] Figure 1 This is a flow chart of the UAV-assisted inspection method for power transmission lines in Example 1.

[0068] Figure 2 This is a flow chart of the optimal path planning strategy in Example 1. DETAILED DESCRIPTION

[0069] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0070] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0071] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0072] Example 1

[0073] Reference Figure 1 and Figure 2, which is the first embodiment of the present invention, provides a transmission line drone-assisted inspection method, comprising the following steps:

[0074] S1, collect environmental data from the drone's global positioning system, inertial measurement unit, and camera sensor to form an environmental status data set.

[0075] Furthermore, the environmental status data set includes positioning and navigation data, vision and image data, environmental parameters, terrain and topography information, endurance status and sensor health data, and external interference data;

[0076] It should be noted that the positioning and navigation data includes the longitude and latitude coordinates and altitude information provided by the drone's global positioning system GPS, as well as the heading angle, pitch angle, roll angle and speed provided by the inertial measurement unit IMU, to ensure that the drone accurately determines its own position and flight direction;

[0077] Visual and image data include high-definition images and video streams captured by the drone’s onboard camera, which are processed to identify visual information such as terrain features, line status, signage, damage signs, etc. Image enhancement technology improves visibility in low light or complex lighting conditions, and the directional gradient histogram algorithm helps understand the scene structure;

[0078] Environmental parameters include recording the surrounding meteorological data such as temperature, humidity, wind speed, wind direction, air pressure, etc., as well as radiation levels, which have a direct impact on the flight performance of the drone. They also include air quality, pollution indicators, and assess visibility and possible impacts on sensors.

[0079] Terrain and geomorphic information includes terrain data based on the digital elevation model (DEM), which provides detailed terrain undulation, slope, and aspect analysis, as well as land cover types such as vegetation, water, rock, and urban building information, which is crucial for route planning and safe flight altitude selection;

[0080] Endurance status and sensor health data include battery power, remaining flight time prediction, and the working status, calibration, and fault warning of each sensor such as GPS, IMU, and camera. Time series analysis and independent component analysis are used to predict and maintain equipment health to ensure reliable operation during the mission period;

[0081] External interference data includes electromagnetic interference such as radar and communication signal strength and spectrum analysis, as well as physical obstacles such as birds and buffer zone analysis of other aircraft to ensure safe flight distance and avoid collision risks. Spectrum analysis and buffer zone analysis help predict and avoid these potential threats.

[0082] S2. Based on the environmental status data set, the dynamic environmental perception adaptation algorithm is used to generate real-time environmental perception data.

[0083] Furthermore, the collected environmental status data set is subjected to noise filtering, and the Kalman filter advanced filtering algorithm is used to remove random noise in the sensor readings to improve data quality. Missing value filling is performed by using interpolation, nearest neighbor filling, or prediction using machine learning models to reasonably fill in the missing parts of the data set to ensure data continuity and integrity and data standardization preprocessing. Through the z-score standardization method, all feature data are scaled to the same scale to facilitate subsequent feature extraction and comparison;

[0084] Different types of feature extraction are performed on each preprocessed data. The positioning and navigation data are filtered to extract features. The filtering process is used to extract accurate position and speed information from the raw data to provide a basis for autonomous navigation of the UAV. The visual and image data are extracted through image enhancement and directional gradient histogram algorithms. Image enhancement improves image quality. The directional gradient histogram (HOG) algorithm extracts shape features in the image, which helps to identify obstacles and target objects. Environmental parameters are extracted through mean and correlation analysis. Statistical analysis such as mean, standard deviation, and correlation reveals the changing trends and relationships of environmental conditions such as temperature and humidity. Terrain and landform information is obtained through digital elevation models. The terrain slope and digital elevation model (DEM) are analyzed to calculate the slope and aspect of the terrain, providing an important reference for the flight altitude and path planning of the drone. The slope aspect analysis is used for feature extraction. The endurance status and sensor health data are extracted through time series analysis and independent component analysis. Time series analysis captures the change pattern of data over time. Independent component analysis (ICA) separates independent signal sources in sensor signals, which helps to monitor the system status. External interference data is extracted through spectrum analysis and buffer analysis. Spectrum analysis identifies specific frequency interference that may affect sensor readings. Buffer analysis evaluates the impact range of nearby interference sources. The extracted features are integrated to generate feature vector expressions:

[0085] F=[f1,f2,...,f n ]

[0086] Among them, F is a vector composed of multiple features, which summarizes the key features extracted from different dimensions to form a comprehensive description of the environment, f1,f2,...,f n are the characteristic elements in the vector F, and n is the total number of features in the characteristic vector F;

[0087] Based on the feature vector, a dynamic evaluation function is constructed to quantify the degree of dynamic change of the current environment. The expression of the dynamic evaluation function is:

[0088]

[0089] Among them, Φ is the dynamic change degree of the current environment, For each feature f i There is a corresponding weight w i , the square is used as a normalized representation, is the standard deviation of the ith feature, μ i is the mean, f i (t) is the specific value of the i-th feature changing with time t, t is the time index variable, f i is the i-th feature in the feature vector F, i is the index variable, and dt is the integral of the change over time t;

[0090] It should be noted that the weighted Gaussian kernel function integral form is used to reflect the smoothness and amplitude of the feature changes over time. i Representative feature f i The importance of μ i and They represent the expected value and variance of the feature respectively. The function calculates the cumulative effect of the overall change of the feature by integration, and ensures the comparability of the evaluation results through the normalization term in the denominator.

[0091] By adjusting the function to map the dynamic evaluation results to the parameter configuration space, real-time monitoring is performed and the perception parameters are dynamically adjusted. The dynamic adjustment function expression is:

[0092]

[0093] Among them, Γ(F) is the perception parameter value after real-time adjustment, α, β, γ are adjustment coefficients;

[0094] It should be noted that by introducing adjustment coefficients α, β, and γ, the environmental dynamics assessment results are mapped to the appropriate perception parameter range. This function is designed to balance the sensitivity and stability of environmental changes, ensuring that parameter adjustments are timely and not too radical.

[0095] Combined with the perception parameter values ​​adjusted in real time, various sensor data are processed instantly, such as dynamically adjusting the sampling rate and filtering strength, to optimize data quality and processing efficiency, and generate comprehensive and accurate real-time environmental perception data that integrates multi-source information, providing a solid foundation for the intelligent decision-making and adaptive flight control of drones.

[0096] It should also be noted that through precise data preprocessing, in-depth feature extraction, dynamic quantitative evaluation, and flexible parameter adjustment mechanism, it is ensured that the UAV can accurately perceive and quickly respond to complex environmental changes. It is the core technical framework for realizing autonomous, efficient, and safe inspection tasks.

[0097] S3. Use geographic information systems and historical line data to build a high-precision digital environment model of the transmission line.

[0098] Furthermore, data cleaning is performed on the terrain data of digital elevation model, land cover type, slope, aspect in the geographic information system, as well as the route direction, height, and surrounding environment records in the historical route data to remove errors, duplications, or inconsistent information in the original data such as digital elevation model DEM, land cover type, slope, aspect, etc., to ensure the accuracy of subsequent analysis. Coordinate registration is used to unify data from different sources into the same coordinate system to ensure accurate alignment and comparison between data and data preprocessing with unified formats, converting all data into compatible formats to facilitate comprehensive processing and analysis on a single platform, and forming a comprehensive terrain and route analysis data set;

[0099] The digital elevation model, land cover type, and historical route compatibility features of the comprehensive terrain and route analysis dataset are extracted to generate slope degrees, numerical data of the normalized vegetation index, category labels of land cover types, and evaluation indicators of route and terrain matching scores. Terrain features that are critical to route design, such as slope, normalized vegetation index NVDI, and compatibility with historical routes, are identified from the preprocessed dataset, laying the foundation for subsequent adaptability scoring.

[0100] Integrate numerical data such as slope degree, NVDI value, category label, and evaluation indicators to form a comprehensive terrain route characteristic indicator set;

[0101] Using the comprehensive terrain route characteristic index set, the multi-scale terrain feature fusion expression is constructed as follows:

[0102]

[0103] Among them, H is the fused high-precision terrain adaptability score, S is the number of scales, ω s is the weight of the Sth scale, N s is the characteristic number at the Sth scale, φ si is the normalized value of the i-th feature at the S-th scale, G si is the matching degree between the ith feature and the historical route data at the Sth scale;

[0104] It should be noted that the multi-scale terrain feature fusion expression describes how to comprehensively consider the matching of terrain features and historical route data at different scales to calculate the final high-precision terrain adaptability score H. Here ω s is the importance weight of each scale, φ si and G si They represent the normalized value and matching degree of the features respectively, emphasizing the comprehensive consideration of terrain adaptability in multiple dimensions.

[0105] Based on the fused high-precision terrain adaptability score H and combined with historical route data, the adaptive path optimization algorithm expression is constructed as follows:

[0106]

[0107] Among them, P opt For the optimized flight path, cost, matching difference and safety are introduced as comprehensive evaluation criteria. In order to find p that minimizes the expression in the brackets, λ1, λ2, and λ3 are the first, second, and third weight factors, respectively, to ensure a balance between cost control, terrain adaptability, and safety. C(P) is the cost function, D(P,H) is the difference in matching between the path and the high-precision terrain adaptability, and U(P) is the safety evaluation function.

[0108] The cost function expression is:

[0109] C(P)=w1·L(P)+w2·V(P)+w3·M(P)

[0110] Among them, w1, w2, w3 are weight factors, L(P) is the path length, V(P) is the flight time cost, M(P) is the maintenance cost, and C(P) refines the cost structure of the flight path, including path length, flight time and maintenance cost. By adjusting the weights of w1, w2, w3, different economic or efficiency goals can be met.

[0111] The security assessment function expression is:

[0112] E(P)=w4·O(P)+w5·W(P)+w6·K A (P)

[0113] Among them, w4, w5, w6 are weight factors, O(P) is the obstacle risk on the path, W(P) is the impact of weather conditions, K A (P) is the terrain adaptability score. E(P) takes into account obstacle risks, weather conditions, and terrain adaptability scores, further strengthening the safety considerations of path selection.

[0114] Finally, the optimized flight path P opt The terrain data and historical line data in the geographic information system are superimposed in three-dimensional space to construct a high-precision digital environment model of the transmission line. This model not only provides detailed flight guidance for drones, but also provides powerful visualization and analysis tools for line maintenance, fault prediction and resource management, greatly improving the operating efficiency and reliability of the power system.

[0115] Furthermore, a high-precision digital environment model of the power transmission line is constructed. The specific steps are as follows:

[0116] Through the geographic information system, all data in the optimized flight path, terrain data, and historical line data, which are collected from multiple reliable sources and include but are not limited to flight path planning, terrain and geomorphic features, historical meteorological data, vegetation coverage, and the precise location and structural information of transmission lines, are preliminarily screened to eliminate erroneous or inapplicable information, and data conversion and format unification are performed to ensure seamless connection between data from different sources. Data conversion and format unification may involve operations such as coordinate system conversion, data projection matching, and normalization of attribute fields to facilitate subsequent analysis and processing.

[0117] Using known coordinate information, spatial registration is performed on terrain data after data conversion and format unification, and on optimized flight paths to ensure that they are correctly aligned in the same spatial reference frame. Historical route data is geo-referenced, and its geographic location information is corrected through geo-reference technology to ensure accurate matching with current terrain and path data.

[0118] Based on the digital elevation model data in the geographic information system, the discrete elevation points in the digital elevation model data are converted into continuous terrain surfaces and rendered to generate a three-dimensional terrain model through interpolation methods such as Kriging interpolation and bilinear interpolation. This can greatly enhance the visual realism of the model and the accuracy of terrain details.

[0119] Taking into account the physical characteristics of the route such as height, diameter, material, etc. to ensure the realism of the model, the optimized flight path and historical route data are converted into three-dimensional lines and superimposed on the three-dimensional model to reflect the actual route direction and altitude changes.

[0120] The professional 3D visualization software Unity is used to perform visualization conversion on the 3D model to generate a high-precision digital environment model of the power transmission line. In Unity, lighting effects, texture mapping, weather simulation and other functions can be further added to enhance the realism and interactivity of the scene.

[0121] S4. Through the high-precision digital environment model of the power transmission line, the autonomous learning algorithm of the UAV is trained to generate a flight strategy model.

[0122] Furthermore, terrain undulations that affect flight altitude and stability, line structure, and key features that determine flight path planning and obstacle distribution are extracted from high-precision digital environmental models of power transmission lines for use in obstacle avoidance strategy formulation, historical data on meteorological conditions, and prediction of weather impacts on flight. Outliers are eliminated and data preprocessing is standardized to ensure accurate interpretation of the data by the algorithm and avoid interference of extreme values ​​in model training.

[0123] The weighted frequency feature similarity function is constructed through the preprocessed data to dynamically adjust the weight of each feature to reflect the changes in the importance of each factor in the current flight environment. The weighted frequency feature similarity function expression is:

[0124]

[0125] Among them, W j is the weight value of the jth feature, E t,j is the energy value of the jth feature at time t, is the energy mean of the jth feature, σ j is the energy standard deviation of the jth feature, is the square root of the time series length T, j' is the number of feature indexes j traversed, and M is the total number of feature indexes j

[0126] It should be noted that the formula W j This paper describes a method for dynamically adjusting the importance of each feature based on historical data. By calculating the deviation of each feature energy value relative to its mean and normalizing it, and using an exponential decay function, features that are more significant in the current flight environment are given higher weights. This step ensures that the flight strategy can respond to environmental changes in a timely manner and improves adaptability.

[0127] For the weighted frequency feature similarity function, adaptive multi-objective deep reinforcement learning is used to optimize environmental adaptability, safety and efficiency during flight. The adaptive multi-objective deep reinforcement learning is used to balance safety, efficiency, task completion and environmental adaptability during flight. At this stage, the algorithm not only needs to learn how to make the best decision to maximize the cumulative reward, but also find the optimal solution among multiple competing goals, and gradually optimize the flight strategy through continuous exploration and utilization.

[0128] Furthermore, the comprehensive objective function is used to integrate the adaptive multi-objective deep reinforcement learning into a unified evaluation standard. The comprehensive objective function expression is:

[0129]

[0130] Among them, Q * (se,a) is the optimal cumulative discounted reward that can be obtained by taking an action a in a given state, se is the current state, a is the action taken, Q * For optimization, To find the optimal strategy among all strategies π, is the expected value of the expression in brackets, γ t is the discount factor, R(se t ,a t ,s et+1 ) is the immediate reward function, W'j (t) is W j The weight value of the j-th feature in will be dynamically adjusted over time t.

[0131] It should be noted that the comprehensive objective function Q * (se,a) emphasizes the optimality in long-term decision making, taking into account the expected rewards of future states and discounting them by a discount factor γ t Discounting ensures that the proportion of recent rewards is higher than that of long-term rewards. At the same time, this function introduces the feature weight W' that is dynamically adjusted over time. j (t), which enhances the sensitivity of the model to immediate environmental changes.

[0132] The immediate reward function expression is:

[0133] R(se t ,a t ,s et+1 )=w1R 安全 +w2R 效率 +w3R 任务 +w4R 适应性

[0134] Among them, R(se t ,a t ,s t+1 ) is the immediate reward function, se t is the current state of the drone, a t The current state of the drone t The execution action below, t+1 To perform action a t After that, the drone reaches a new state in response to the environment, w1, w2, w3, w4 are weight factors, R 安全 For drone safety rewards, R 效率 is the drone efficiency bonus, R 任务 The reward for the completion of the drone mission, R 适应性 Bonus for drone adaptability.

[0135] It should be noted that the immediate reward function R(se t ,a t ,s et+1 ) quantifies the immediate effect of each decision through four dimensions: safety, efficiency, mission completion, and adaptability. The contribution of each dimension is adjusted by the corresponding weight factors w1, w2, w3, and w4. This design encourages drones to pursue speed and efficiency while ensuring safety and flexibly adapting to environmental changes when performing tasks.

[0136] Use real-time data to continuously evaluate the dynamic changes of the flight environment and adjust the feature weights W jTo ensure that the flight strategy model can adapt to the new environment in real time, the flight strategy model is generated. This closed-loop feedback mechanism ensures the continuous optimization of the flight strategy model, enabling the drone to adapt to the new environment in real time and improve its autonomous decision-making ability in complex scenarios.

[0137] It should also be noted that through precise data processing, intelligent algorithm design and flexible strategy adjustment mechanism, a highly intelligent and adaptive UAV flight control system has been jointly constructed, providing strong technical support for inspection tasks in complex environments.

[0138] The present embodiment also provides a transmission line UAV assisted inspection system, including an environmental data acquisition module, a dynamic environmental perception adaptation processing module, a high-precision digital environment modeling module, a flight strategy algorithm training module, and a real-time path planning and flight control fusion module; the environmental data acquisition module is used to collect environmental data from the UAV global positioning system, inertial measurement unit and camera sensor to form an environmental state data set; the dynamic environmental perception adaptation processing module is used to generate real-time environmental perception data based on the environmental state data set using a dynamic environmental perception adaptation algorithm; the high-precision digital environment modeling module is used to use a geographic information system and historical line data to construct a high-precision digital environment model of the transmission line; the flight strategy algorithm training module is used to train the UAV's autonomous learning algorithm through a high-precision digital environment model of the transmission line to generate a flight strategy model; the real-time path planning and flight control fusion module is used to fuse the real-time environmental perception data and the flight strategy model in real time to generate an optimal path planning strategy, so that the UAV can adjust flight parameters according to the real-time environment and optimize the flight path.

[0139] This embodiment also provides a computer device, which is suitable for the case of a UAV-assisted inspection method for transmission lines, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the UAV-assisted inspection method for transmission lines proposed in the above embodiment.

[0140] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0141] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for implementing the transmission line drone-assisted inspection proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, disk or optical disk.

[0142] In summary, the present invention builds an intelligent UAV inspection system by integrating global positioning and inertial sensor data fusion, multi-dimensional environmental status data analysis, dynamic environmental adaptation algorithm, high-precision terrain modeling and adaptive path planning, and autonomous optimization of deep reinforcement learning strategies. It greatly improves the adaptability to complex environments and inspection efficiency, and has important application value for power maintenance.

[0143] Example 2

[0144] Referring to Table 1, which is the second embodiment of the present invention, in order to further verify the advancement of the present invention, experimental simulation data of the transmission line UAV-assisted inspection method are provided.

[0145] Two drones were prepared, one using the traditional inspection method (control group) and the other optimized according to the method of the present invention (experimental group). Both drones were equipped with standard sensors such as GPS, IMU, and camera. In addition, the experimental group drone was also equipped with dynamic environment perception adaptation algorithm and autonomous learning algorithm module.

[0146] First, the two drones performed inspection tasks in the same power transmission line area under the same weather conditions, and collected global positioning system data, inertial measurement unit data, and camera sensor data at the same time to ensure the comprehensiveness of the environmental status data set. For the experimental group, special attention was also paid to environmental parameters, terrain information, endurance status and sensor health, external interference, etc.

[0147] Secondly, after the experimental group drones collected data, they immediately filtered out noise, filled in missing values, and standardized the data. Then, feature extraction technology was used to conduct in-depth analysis of various types of data. For example, the recognition accuracy of visual and image data was improved through image enhancement and directional gradient histogram algorithms. Subsequently, the degree of environmental change was quantified through a dynamic evaluation function, and the perception parameters were adjusted accordingly to generate real-time environmental perception data.

[0148] Then, the data collected by the experimental group's drones were deeply integrated using the Geographic Information System (GIS) and historical line data. Through preprocessing steps such as data cleaning and coordinate registration, a high-precision digital environment model of the transmission line was constructed. The model includes multi-scale terrain feature fusion expressions and adaptive path optimization algorithms to ensure the efficiency and safety of the flight path.

[0149] Finally, based on the constructed digital environment model, the autonomous learning algorithm of the experimental group of drones was trained to generate a flight strategy model, which dynamically adjusted the weights according to environmental characteristics and continuously optimized the flight strategy using adaptive multi-objective deep reinforcement learning. Finally, combined with real-time environmental perception data, the flight path was adjusted in real time to achieve the best path planning.

[0150] The details are shown in the following table:

[0151] Table 1 Experimental record table

[0152]

[0153] From the data in the table above, we can see that compared with the control group, the experimental group of drones showed significant improvements in multiple key parameters:

[0154] The experimental group's positioning accuracy improved by 1.3 meters, which means that the drone can fly along the predetermined route more accurately and reduce the risk of accidentally entering dangerous areas.

[0155] The image recognition rate of the experimental group increased by 15%, which is directly related to the accuracy of inspections and fault identification capabilities, and can detect potential problems more quickly.

[0156] The experimental group's path planning efficiency increased by 22%, indicating that the optimized flight strategy model can more effectively plan obstacle avoidance paths and reduce unnecessary detours.

[0157] The experimental group's flight safety score increased by 2 points, thanks to dynamic environment perception and real-time path adjustment, which significantly reduced the risk of collision.

[0158] The experimental group’s flight time was extended by 5 minutes, which means that a larger inspection area can be covered in a single flight, improving work efficiency.

[0159] The transmission line UAV-assisted inspection method of the present invention has been fully demonstrated through detailed comparison of the embodiments in terms of its innovation and advantages in improving inspection efficiency, enhancing safety, and optimizing resource utilization. This series of improvements not only reflects a significant improvement over the existing technology, but also brings technological innovation to the field of UAV inspection.

[0160] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A transmission line drone-assisted inspection method, characterized in that: include, Collect environmental data from the drone's global positioning system, inertial measurement unit, and camera sensors to form an environmental status data set; Based on the environmental state data set, using a dynamic environmental perception adaptation algorithm, generating real-time environmental perception data; Use geographic information systems and historical line data to build a high-precision digital environment model of the transmission line; The autonomous learning algorithm of the UAV is trained by using the high-precision digital environment model of the power transmission line to generate a flight strategy model; The real-time environment perception data and the flight strategy model are integrated in real time to generate an optimal path planning strategy, so that the UAV can adjust the flight parameters according to the real-time environment and optimize the flight path; The environmental status data set includes positioning and navigation data, vision and image data, environmental parameters, terrain and landform information, endurance status and sensor health data, and external interference data; Based on the environmental state data set, a dynamic environmental perception adaptation algorithm is used to generate real-time environmental perception data. The specific steps are as follows: The collected environmental status data set is preprocessed by noise filtering, missing value filling and data standardization; Different types of feature extraction are performed on each preprocessed data. The positioning and navigation data are extracted through filtering processing, the visual and image data are extracted through image enhancement and directional gradient histogram algorithm, the environmental parameters are extracted through mean and correlation analysis, the terrain and landform information are extracted through digital elevation model terrain slope and slope aspect analysis, the endurance status and sensor health data are extracted through time series analysis and independent component analysis, and the external interference data are extracted through spectrum analysis and buffer zone analysis. The extracted features are integrated to generate the feature vector expression: F=[f1,f2,...,f n ] Among them, F is a vector composed of multiple features, f1,f2,...,f n are the characteristic elements in the vector F, and n is the total number of features in the characteristic vector F; Based on the feature vector, a dynamic evaluation function is constructed to quantify the degree of dynamic change of the current environment. The expression of the dynamic evaluation function is: Among them, Φ is the dynamic change degree of the current environment, For each feature f i There is a corresponding weight w i , the square is used as a normalized representation, is the standard deviation of the ith feature, μ i is the mean, f i (t) is the specific value of the i-th feature changing with time t, t is the time index variable, f i is the i-th feature in the feature vector F, i is the index variable, and dt is the integral of the change over time t; By adjusting the function to map the dynamic evaluation results to the parameter configuration space, real-time monitoring is performed and the perception parameters are dynamically adjusted. The dynamic adjustment function expression is: Among them, Γ(F) is the perception parameter value after real-time adjustment, α, β, γ are adjustment coefficients; Combined with the perception parameter values ​​adjusted in real time, various sensor data are processed immediately to generate real-time environmental perception data that integrates multi-source information; The specific steps of constructing a high-precision digital environment model of power transmission lines by using geographic information system and historical line data are as follows: Data cleaning, coordinate registration and data preprocessing in a unified format are performed on the digital elevation model, land cover type, slope, and aspect terrain data in the geographic information system, as well as the route direction, height, and surrounding environment records in the historical route data, to form a comprehensive terrain and route analysis data set; Extract features of digital elevation model, land cover type, and historical route compatibility from the comprehensive terrain and route analysis dataset, generate numerical data of slope degree, normalized vegetation index, category labels of land cover type, and evaluation indicators of route and terrain matching scores; Integrate the numerical data, category labels, and evaluation indicators to form a comprehensive terrain route characteristic indicator set; Using the comprehensive terrain route characteristic index set, the multi-scale terrain feature fusion expression is constructed as follows: Among them, H is the fused high-precision terrain adaptability score, S is the number of scales, ω s is the weight of the Sth scale, N s is the characteristic number at the Sth scale, φ si is the normalized value of the i-th feature at the S-th scale, G si is the matching degree between the ith feature and the historical route data at the Sth scale; Based on the fused high-precision terrain adaptability score H and combined with historical route data, the adaptive path optimization algorithm expression is constructed as follows: Among them, P opt For the optimized flight path, To find p that makes the expression in the brackets reach the minimum value, λ1, λ2, λ3 are the first, second, and third weight factors respectively, C(P) is the cost function, D(P,H) is the difference in adaptability between the path and the high-precision terrain, and U(P) is the safety evaluation function; The cost function expression is: C(P)=w1·L(P)+w2·V(P)+w3·M(P) Among them, w1, w2, w3 are weight factors, L(P) is the path length, V(P) is the flight time cost, and M(P) is the maintenance cost; The security assessment function expression is: E(P)=w4·O(P)+w5·W(P)+w6·K A (P) Among them, w4, w5, w6 are weight factors, O(P) is the obstacle risk on the path, W(P) is the impact of weather conditions, K A (P) is the terrain adaptability score; The optimized flight path P opt It is superimposed with terrain data and historical line data in the geographic information system in three-dimensional space to build a high-precision digital environment model of the transmission line.

2. The transmission line drone-assisted inspection method according to claim 1, characterized in that: The specific steps of constructing a high-precision digital environment model of a power transmission line are as follows: Through the geographic information system, all data of the optimized flight path, terrain data, and historical route data are converted and formatted uniformly; Using the known coordinate information, spatially align the terrain data after data conversion and format unification and the optimized flight path, and geo-align the historical route data; Based on the digital elevation model data in the geographic information system, the discrete elevation points in the digital elevation model data are converted into a continuous terrain surface through interpolation method and rendered to generate a three-dimensional terrain model; Converting the optimized flight path and historical route data into three-dimensional lines and superimposing them on the three-dimensional model; The professional 3D visualization software Unity is used to visualize the 3D model and generate a high-precision digital environment model of the transmission line.

3. The transmission line drone-assisted inspection method according to claim 2, characterized in that: The autonomous learning algorithm of the UAV is trained by using the high-precision digital environment model of the power transmission line to generate a flight strategy model. The specific steps are as follows: Extract key features of terrain undulation, line structure, obstacle distribution, and historical data of meteorological conditions from the high-precision digital environment model of power transmission lines, and perform outlier removal and standardized data preprocessing; The weighted frequency feature similarity function is constructed through the preprocessed data to dynamically adjust the weight of each feature to reflect the importance change of each factor in the current flight environment. The weighted frequency feature similarity function expression is: Among them, W j is the weight value of the jth feature, E t,j is the energy value of the jth feature at time t, is the energy mean of the jth feature, σ j is the energy standard deviation of the jth feature, is the square root of the time series length T, j′ is the number of traversed feature indexes j, and M is the total number of feature indexes j; For the weighted frequency feature similarity function, adaptive multi-objective deep reinforcement learning is used to optimize environmental adaptability, safety and efficiency during flight; The comprehensive objective function is used to integrate the adaptive multi-objective deep reinforcement learning into a unified evaluation standard. The expression of the comprehensive objective function is: Among them, Q * (se,a) is the optimal cumulative discounted reward that can be obtained by taking an action a in a given state, se is the current state, a is the action taken, Q * For optimization, To find the optimal strategy among all strategies π, is the expected value of the expression in brackets, γ t is the discount factor, R(se t ,a t ,s et+1 ) is the immediate reward function, W′ j (t) is the W j The weight value of the jth feature in will be dynamically adjusted over time t; The instant reward function expression is: R(se t ,whose t ,is t+1 )=w1R 安全 +w2R 效率 +w3R 任务 +w4R 适应性 Among them, R(se t ,a t ,s t+1 ) is the immediate reward function, se t is the current state of the drone, a t The current state of the drone t The execution action below, t+1 To perform action a t After that, the drone reaches a new state in response to the environment, w1, w2, w3, w4 are weight factors, R 安全 For drone safety rewards, R 效率 is the drone efficiency bonus, R 任务 The reward for the completion of the drone mission, R 适应性 Bonus for drone adaptability; Use real-time data to continuously evaluate the dynamic changes of the flight environment and adjust the feature weights W j To ensure that the flight strategy model can adapt to the new environment in real time and generate a flight strategy model.

4. The transmission line drone-assisted inspection method according to claim 3, characterized in that: The real-time environment perception data and the flight strategy model are integrated in real time to generate the best path planning strategy. The specific steps are as follows: Establishing a data interface to transmit the real-time environment perception data to the flight strategy model processing module; Based on the feature vector F=[f1,f2,...,f n ], using the dynamic evaluation function φ(F) to calculate the degree of dynamic change of the environment; According to the environmental dynamics assessment result φ, the perception parameters of the UAV are optimized in real time by dynamically adjusting the function Γ(φ); Based on real-time optimization of the drone’s perception parameters, adaptive multi-objective deep reinforcement learning and comprehensive objective functions are used to dynamically adjust the feature weights W in the flight strategy model. j , forming an updated flight strategy model; The updated flight strategy model is applied to the UAV path planning, and the flight path P is recalculated in combination with the real-time UAV optimized perception data. opt , ensuring that real-time data feedback can immediately adjust the flight strategy and generate the best path planning strategy.

5. A transmission line drone-assisted inspection system, based on the transmission line drone-assisted inspection method according to any one of claims 1 to 4, characterized in that: Including, environmental data acquisition module, dynamic environment perception adaptation processing module, high-precision digital environment modeling module, flight strategy algorithm training module, real-time path planning and flight control fusion module; The environmental data acquisition module is used to collect environmental data from the drone's global positioning system, inertial measurement unit, and camera sensor to form an environmental status data set; The dynamic environment perception adaptation processing module is used to generate real-time environment perception data based on the environment state data set using a dynamic environment perception adaptation algorithm; The high-precision digital environment modeling module is used to construct a high-precision digital environment model of the power transmission line using the geographic information system and historical line data; The flight strategy algorithm training module is used to train the autonomous learning algorithm of the UAV through the high-precision digital environment model of the power transmission line to generate a flight strategy model; The real-time path planning and flight control fusion module is used to fuse the real-time environment perception data and the flight strategy model in real time to generate an optimal path planning strategy, so that the UAV can adjust flight parameters according to the real-time environment and optimize the flight path.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the transmission line drone-assisted inspection method according to any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the transmission line drone-assisted inspection method according to any one of claims 1 to 4 are implemented.

Images