Unmanned aerial vehicle intelligent inspection path planning method in construction stage of converter station

Abstract

The invention provides an unmanned aerial vehicle intelligent inspection path planning method in a converter station construction stage, and belongs to the technical field of path planning. Electromagnetic interference source distribution in electrical installation, debugging and other stages is dynamically recorded by establishing a converter station three-dimensional environment model and a stage characteristic database; an improved bat algorithm is adopted, a double-layer game optimization framework is constructed, a dynamic parameter adaptive adjustment mechanism is utilized to adjust a search strategy in real time according to electromagnetic interference intensity and an algorithm convergence state, path risks are evaluated through an electromagnetic interference intensity calculation function, and a multi-band adaptive filtering system is started to perform interference suppression. Stable planning and reliable execution of the unmanned aerial vehicle inspection path in the electromagnetic interference environment are realized, and the technical problems of poor convergence of an unmanned aerial vehicle inspection path planning algorithm and low path execution reliability caused by the electromagnetic interference environment in the processes of electrical installation, debugging and the like in the construction stage of the converter station are solved.

Description

Technical Field

[0001] This invention belongs to the field of path planning technology, and more specifically, relates to a method for intelligent inspection path planning by unmanned aerial vehicles (UAVs) during the construction phase of a converter station. Background Technology

[0002] Unmanned aerial vehicle (UAV) inspection path planning technology during the converter station construction phase is an important component of intelligent operation and maintenance in the power system. Traditional path planning methods mainly employ heuristic algorithms such as genetic algorithms, ant colony optimization, and particle swarm optimization. These methods can generate relatively reasonable inspection paths under ideal environmental conditions and have been successfully applied in relatively stable electromagnetic environments such as conventional substation inspections and transmission line monitoring. However, traditional algorithms are designed primarily based on static constraints and are severely inadequate in adapting to electromagnetic interference environments, exhibiting significant performance degradation under the complex electromagnetic interference conditions at converter station construction sites. In current converter station construction phase inspection practices, due to the frequent start-ups and shutdowns of high-power equipment during the electrical installation phase, the strong electromagnetic fields generated by the energization of high-voltage equipment during the commissioning and acceptance phase, and the complex distribution of electromagnetic interference sources formed by various construction and communication equipment, traditional path planning algorithms are prone to problems such as slow convergence speed, getting trapped in local optima, and unstable planning results. Furthermore, during actual execution, electromagnetic interference affects the UAV's navigation and communication systems, leading to poor path tracking accuracy and a high task failure rate. In other words, existing technologies suffer from low reliability in UAV path planning execution due to electromagnetic interference caused by electrical installation and commissioning processes during the converter station construction phase. Summary of the Invention

[0003] In view of this, the present invention provides a method for intelligent inspection path planning of UAVs during the construction phase of a converter station, which can solve the technical problem of low reliability of UAV path planning execution in the prior art due to electromagnetic interference caused by electrical installation, commissioning and other processes during the construction phase of a converter station.

[0004] This invention is implemented as follows: This invention provides a method for intelligent unmanned aerial vehicle (UAV) inspection path planning during the construction phase of a converter station, comprising the following steps: establishing a three-dimensional environment model of the converter station, constructing three spatial layers: underground pipeline layer, ground equipment layer, and high-altitude line layer; collecting spatial coordinates, obstacle distribution, and equipment location information for each layer to form a multi-layered spatial database; constructing a construction phase feature database, recording the equipment distribution status, construction progress, safety constraints, and electromagnetic interference source distribution for each phase according to the different characteristics of the civil engineering phase, electrical installation phase, and commissioning and acceptance phase, and establishing a dynamically updated set of phase feature parameters; initializing the improved bat algorithm parameters; and establishing a multi-objective optimization function. The algorithm aims to minimize the total path length, maximize equipment coverage, minimize emergency obstacle avoidance response time, and minimize electromagnetic interference false alarm rate. A two-layer game-theoretic optimization framework is constructed. Local path costs and sampling parameters are calculated. Based on the current UAV position and target point position, the maximum sampling distance, the coordinates of the farthest sampling point, and the maximum sampling step size are calculated to form a basic parameter set. Dynamic parameter adaptive adjustment is performed. The first and second rate of change thresholds are calculated using a step function. Based on the current iteration count and convergence status, the pulse emission rate and loudness parameters are dynamically adjusted. Path effectiveness verification and risk assessment are conducted. The electromagnetic interference intensity value of the planned path is calculated using an electromagnetic interference intensity calculation function.

[0005] Specifically, the steps for establishing the multi-level spatial database involve acquiring geometric and equipment information at different spatial levels of the converter station through three-dimensional laser scanning to form a structured spatial coordinate dataset.

[0006] Specifically, the set of stage feature parameters is a dynamically updated parameter vector group that describes the environmental characteristics and constraints of different construction stages, including construction plans, equipment installation status, safety specifications, and electromagnetic interference source location information.

[0007] Specifically, the improved bat algorithm is a swarm intelligence optimization algorithm based on the echolocation behavior of bats. It searches for the optimal solution by simulating bats emitting ultrasonic waves and receiving echoes. During the construction phase of the converter station, the impact of electromagnetic interference generated at different construction stages on the convergence of the algorithm is considered.

[0008] The initialization and improvement of the bat algorithm parameters includes setting convergence criteria based on Lyapunov stability theory. By constructing a Lyapunov function, the convergence criteria of the algorithm are determined to prevent it from getting trapped in a local optimum.

[0009] Specifically, the two-layer game optimization framework includes an upper-layer path optimization model with path efficiency as the objective and a lower-layer parameter adjustment model with parameter adjustment as the objective. The two objective functions influence each other through coupling terms consisting of the product of pulse emission rate and path length weights, the product of loudness and equipment coverage weights, and the product of frequency and obstacle avoidance time weights.

[0010] The objective function of the upper-level path optimization model includes a path length term, a reciprocal term for device coverage, an obstacle avoidance response time term, and a logarithmic term for electromagnetic interference false alarm rate.

[0011] The objective function of the lower-level parameter adjustment model includes a pulse emission rate square term, a loudness linear term, a frequency sine term, and a population diversity index term.

[0012] Specifically, the set of basic parameters is determined based on the UAV's endurance and communication range, with a maximum sampling distance used to limit the search range and step size of a single path planning operation by the UAV.

[0013] Specifically, the step function is a first rate of change threshold and a second rate of change threshold that are dynamically adjusted based on the current iteration number, the path length change rate, and the population convergence index.

[0014] Specifically, the dynamic parameter adaptive adjustment step enhances global exploration capabilities when the path length change rate is less than a first change rate threshold, and strengthens local development capabilities when the path length change rate is greater than a second change rate threshold.

[0015] The global exploration capability refers to the ability of the improved bat algorithm to search for the optimal solution extensively throughout the solution space, by increasing the pulse emission rate and reducing the loudness to expand the search range.

[0016] The local development capability refers to the ability of the improved bat algorithm to perform a fine search near the current optimal solution by reducing the pulse emission rate and increasing the loudness to narrow the search range.

[0017] Specifically, the path validity verification and risk assessment steps involve executing the normal inspection process when the electromagnetic interference intensity value is ∈ [0, 0.3), activating the multi-band adaptive filtering system when the electromagnetic interference intensity value is ∈ [0.3, 0.7), and replanning the path and increasing the safety distance when the electromagnetic interference intensity value is ∈ [0.7, 1.0].

[0018] Specifically, the multi-band adaptive filtering system uses a blind source separation algorithm to identify interference source signals of different frequencies, and processes various electromagnetic interference signals by combining spatial filtering technology and temporal filtering technology.

[0019] Specifically, the electromagnetic interference intensity calculation function calculates the degree of electromagnetic interference experienced by the planned path based on the set of path coordinate points, the location of the electromagnetic interference source, the power of the interference source, and the propagation distance attenuation coefficient.

[0020] This invention establishes a path planning algorithm capable of stable convergence under strong electromagnetic interference (EMI) conditions by employing an improved bat algorithm combined with Lyapunov stability theory, effectively solving the problem of poor convergence of traditional algorithms under EMI conditions. The invention utilizes a dynamic parameter adaptive adjustment mechanism to adjust pulse emission rate and loudness parameters in real time based on changes in EMI intensity and the algorithm's convergence status. A dynamic balance between global exploration and local exploitation capabilities is achieved through a step function-calculated rate-of-change threshold, significantly improving the algorithm's convergence stability and optimization efficiency in complex electromagnetic environments. The path risk assessment and interference suppression mechanism, combining an EMI intensity calculation function with a multi-band adaptive filtering system, effectively suppresses the impact of EMI on UAV communication and navigation systems by real-time monitoring of EMI intensity at each point on the path and implementing corresponding filtering strategies, ensuring the executability of the planned path and the reliability of task completion. The invention's designed two-layer game-theoretic optimization framework treats path quality optimization and algorithm parameter adjustment in layers. Through collaborative optimization of the upper and lower layer models, it achieves multi-objective balance under EMI constraints, overcoming the limitations of traditional single optimization methods that struggle to balance convergence and execution reliability. In summary, this invention solves the technical problem mentioned in the background art, where electromagnetic interference occurs during the construction phase of converter stations, leading to low reliability of UAV path planning execution. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention.

[0022] Figure 2 This is a comparison diagram of the electromagnetic interference intensity distribution in the three construction phases of the embodiment.

[0023] Figure 3 This is an evolution diagram of drone inspection path planning at different construction stages in the embodiment. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0025] like Figure 1 The diagram shown is a flowchart of an intelligent unmanned aerial vehicle (UAV) inspection path planning method for the construction phase of a converter station, provided by this invention. This method includes the following steps: S01. Establish a three-dimensional environment model of the converter station, and construct three spatial layers: underground pipeline layer, ground equipment layer, and high-altitude line layer. Collect spatial coordinates, obstacle distribution, and equipment location information of each layer to form a multi-level spatial database. S02. Construct a construction phase characteristic database. Based on the different characteristics of the civil engineering phase, electrical installation phase, and commissioning and acceptance phase, record the equipment distribution status, construction progress, safety constraints, and electromagnetic interference source distribution of each phase, and establish a dynamically updated set of phase characteristic parameters. S03. Initialize the improved bat algorithm parameters, set the population size to 50, the maximum number of iterations to 200, the initial pulse emission rate to 0.5, the initial loudness to 0.25, and the frequency range to [0, 2]. Set the convergence criteria according to the Lyapunov stability theory. S04. Establish a multi-objective optimization function with the objectives of minimizing the total path length, maximizing equipment coverage, minimizing emergency obstacle avoidance response time, and minimizing electromagnetic interference false alarm rate. Construct a two-layer game optimization framework with an upper-level path optimization model that aims at path efficiency and a lower-level parameter adjustment model that aims at parameter adjustment. S05. Calculate the local path cost and sampling parameters. Based on the current UAV position and target point position, calculate the maximum sampling distance, the coordinates of the farthest sampling point, and the maximum sampling step size to form a set of basic parameters for local path planning. S06. Perform dynamic parameter adaptive adjustment. Calculate the first and second rate of change thresholds through the step function. Dynamically adjust the pulse emission rate and loudness parameters based on the current iteration number and convergence status. When the path length change rate is less than the first rate of change threshold, enhance global exploration capability. When the path length change rate is greater than the second rate of change threshold, strengthen local development capability. S07. Perform path validity verification and risk assessment. Calculate the electromagnetic interference intensity value of the planned path using the electromagnetic interference intensity calculation function. When the electromagnetic interference intensity value ∈ [0, 0.3), execute the normal inspection process. When the electromagnetic interference intensity value ∈ [0.3, 0.7), start the multi-band adaptive filtering system. When the electromagnetic interference intensity value ∈ [0.7, 1.0], replan the path and increase the safety distance.

[0026] Among them, the improved bat algorithm is a swarm intelligence optimization algorithm based on the echolocation behavior of bats. It searches for the optimal solution by simulating bats emitting ultrasonic waves and receiving echoes. During the construction phase of the converter station, it is necessary to consider the impact of electromagnetic interference generated at different construction stages on the convergence of the algorithm. In particular, the strong electromagnetic interference generated by the power-on of high-voltage equipment during the commissioning and acceptance phase will affect the algorithm's global exploration capability and local development capability.

[0027] Lyapunov stability theory is used to ensure the convergence of the algorithm in high-dimensional solution space. By constructing a Lyapunov function, the convergence condition of the algorithm is judged, thus preventing it from getting trapped in a local optimum.

[0028] The two-layer game optimization framework includes an upper-layer path optimization model with path efficiency as the objective and a lower-layer parameter adjustment model with parameter adjustment as the objective. The objective function of the upper-layer path optimization model consists of four components: path length, reciprocal of equipment coverage, obstacle avoidance response time, and logarithm of electromagnetic interference false alarm rate. The objective function of the lower-layer parameter adjustment model consists of four components: square of pulse emission rate, linear of loudness, sine of frequency, and population diversity index. The two objective functions influence each other through coupling terms consisting of the product of pulse emission rate and path length weights, the product of loudness and equipment coverage weights, and the product of frequency and obstacle avoidance time weights.

[0029] The basic parameter set is used to limit the search range and step size of the UAV's single path planning. The maximum sampling distance is determined based on the UAV's endurance and communication distance to avoid planning excessively long paths that could lead to mission failure.

[0030] The step function is used to calculate dynamically adjusted threshold parameters based on the current iteration progress and algorithm convergence status. The inputs include the current iteration number, path length change rate, and population convergence index. The outputs are the first change rate threshold and the second change rate threshold.

[0031] Among them, global exploration capability refers to the ability of the improved bat algorithm to search for optimal solutions extensively in the entire solution space. This is achieved by increasing the pulse emission rate and reducing the loudness to expand the search range and prevent the algorithm from converging to a local optimum too early.

[0032] Among them, local exploitation capability refers to the ability of the improved bat algorithm to perform fine-grained searches near the current optimal solution. This is achieved by reducing the pulse emission rate and increasing the loudness to narrow the search range and improve the search accuracy of the algorithm in local areas.

[0033] Among them, the multi-band adaptive filtering system is used to suppress electromagnetic interference signals at the construction site. It uses a blind source separation algorithm to identify interference source signals of different frequencies and processes various electromagnetic interference signals by combining spatial filtering technology and temporal filtering technology.

[0034] The electromagnetic interference intensity calculation function is used to evaluate the degree of electromagnetic interference on the planned path. The inputs include the set of path coordinate points, electromagnetic interference source location data, interference source power data, and propagation distance attenuation coefficient. The output is the electromagnetic interference intensity value.

[0035] Among them, path validity verification refers to the process of checking the safety and feasibility of the planned path. It involves calculating the electromagnetic interference intensity, obstacle collision risk, and communication signal strength at each point on the path to determine whether the path meets the requirements of the inspection task.

[0036] The blind source separation algorithm is used to separate individual interference source signals from mixed electromagnetic interference signals. The input includes multi-channel received mixed interference signal data and signal spectrum feature data, and the output is the separated individual interference source signals.

[0037] Among them, spatial filtering technology is used to suppress electromagnetic interference signals from a specified direction by utilizing the spatial position information of multiple sensors. Spatial suppression of interference signals is achieved by adjusting the amplitude and phase of the signals from each sensor.

[0038] Among them, time-domain filtering technology is used to filter electromagnetic interference signals in the time domain. By designing different filter parameters, it is possible to suppress interference signals of different frequency components.

[0039] The specific implementation methods of the above steps are described in detail below. Step S01 involves establishing a three-dimensional environment model of the converter station and constructing a multi-level spatial database using a layered modeling principle. First, point cloud data of the converter station construction site is acquired using three-dimensional laser scanning technology. The scanning accuracy is set to 2–5 mm, and the scanning range covers three spatial levels: underground pipelines, ground equipment, and overhead lines. Then, a point cloud registration algorithm is used to spatially align the point cloud data from multiple scans, with the registration error controlled within 10 mm. Next, a three-dimensional reconstruction algorithm is used to convert the point cloud data into a triangular mesh model. The mesh density is adaptively adjusted according to the accuracy requirements of different levels: the mesh side length for the underground pipeline layer is set to 0.1–0.3 meters, for the ground equipment layer to 0.05–0.2 meters, and for the overhead lines layer to 0.2–0.5 meters. Finally, a multi-level spatial database is established using a spatial indexing algorithm. An octree structure is used to store spatial coordinate data, and the number of geometric objects stored in each leaf node is controlled to be 8–16 to improve spatial query efficiency. The purpose of this step is to provide accurate three-dimensional spatial environment data for subsequent path planning.

[0040] The specific implementation of step S02 involves constructing a construction phase feature database, recording dynamic feature parameters of different construction phases based on time-series modeling principles. First, the construction is divided into three main phases based on the construction plan data: civil engineering, electrical installation, and commissioning / acceptance. The time span of each phase is determined according to the actual project progress. Then, a state machine model is used to describe the equipment distribution status of each phase. The civil engineering phase mainly records the construction progress of infrastructure and structures; the electrical installation phase focuses on recording the equipment installation location and connection status; and the commissioning / acceptance phase records the operating status and electromagnetic field distribution of energized equipment in detail. Next, a dynamic update mechanism is used to collect construction progress information in real time, with the update frequency set to once daily or based on major construction milestones. Then, safety constraint modeling technology is used to convert safety specifications into numerical constraints, including parameters such as minimum safe distance, no-fly zone boundaries, and flight altitude restrictions. Finally, an electromagnetic interference source distribution model is established, and an electromagnetic field simulation algorithm is used to calculate the electromagnetic field intensity distribution generated by electrical equipment at different locations and with different power levels, forming a three-dimensional electromagnetic interference intensity field. The purpose of this step is to provide corresponding environmental feature constraints for different construction phases.

[0041] The specific implementation of step S03 involves initializing the parameters of the improved bat algorithm and setting the algorithm control parameters based on the principle of swarm intelligence optimization. First, the population size is determined to be 50 individuals based on the complexity of the path planning problem, with each individual representing a candidate path. Then, the maximum number of iterations is set to 200 to balance the algorithm's convergence speed and solution quality. Next, the pulse emission rate is initialized to 0.5, which controls the algorithm's global search capability and ranges from 0 to 1. Simultaneously, the initial loudness is set to 0.25, which affects the algorithm's local exploitation capability and also ranges from 0 to 1. Then, the frequency range is determined to be 0 to 2, and the frequency parameter adjusts the algorithm's search step size and direction. Next, a convergence criterion function is constructed based on Lyapunov stability theory. The convergence of the algorithm is determined by calculating the rate of change of the objective function value over several consecutive generations; convergence is considered achieved when the rate of change of the objective function value over 10 consecutive generations is less than 0.001. Finally, the population position and velocity vectors are initialized, and the initial solutions are uniformly distributed in the solution space using the Latin hypercube sampling method. The purpose of this step is to provide suitable initial parameter configurations for the improved bat algorithm.

[0042] The specific implementation of step S04 involves establishing a multi-objective optimization function and constructing a multi-objective optimization model using a two-layer game optimization framework. First, the objective function of the upper-layer path optimization model is constructed, including four optimization objectives: minimizing the total path length, maximizing equipment coverage, minimizing emergency obstacle avoidance response time, and minimizing the electromagnetic interference false alarm rate. The path length term uses Euclidean distance to calculate the cumulative straight-line distance between adjacent path points. The equipment coverage term calculates the ratio of the number of devices that the planned path can effectively monitor to the total number of devices. The obstacle avoidance response time term calculates the time required for emergency obstacle avoidance based on the distance from each point on the path to the nearest obstacle and the maximum maneuverability of the UAV. The electromagnetic interference false alarm rate term calculates the interference probability at each point on the path based on the electromagnetic interference intensity distribution. Then, the objective function of the lower-layer parameter adjustment model is constructed, including four parameter optimization objectives: pulse emission rate squared term, loudness linear term, frequency sine term, and population diversity index term. Next, the coupling relationship between the upper and lower layer models is established. Coupling terms are formed by multiplying the pulse emission rate by the path length weight, the loudness by the equipment coverage weight, and the frequency by the obstacle avoidance time weight, thus achieving mutual influence and coordinated optimization between the two optimization models. Finally, a weighted summation method is used to combine multiple optimization objectives into a single objective function, with the weights of each objective set according to the priority of the inspection task. The purpose of this step is to establish an optimization objective system that comprehensively considers multiple performance indicators.

[0043] The specific implementation of step S05 involves calculating the local path cost and sampling parameters, and determining the basic parameters for path planning based on heuristic search principles. First, the straight-line distance between the current UAV position and the target point position is calculated as a benchmark value for path cost evaluation. Then, the maximum sampling distance is determined based on the UAV's endurance and communication distance limitations. The endurance is set to 30-60 minutes, and the communication distance to 2-5 kilometers, with the maximum sampling distance being the smaller of the two. Next, the coordinates of the farthest sampling point are calculated. Within a circular area centered on the current position and with the maximum sampling distance as the radius, the farthest reachable sampling point is determined along the target direction. Then, the maximum sampling step size is determined based on the UAV's flight speed and control accuracy. The flight speed is set to 5-15 meters per second, and the control accuracy to 0.5-2 meters, with the maximum sampling step size set to 2-5 times the control accuracy. Next, the gradient descent method is used to calculate the local path cost gradient to guide the algorithm's search direction. Finally, an adaptive adjustment mechanism for the sampling parameters is established, dynamically adjusting the sampling density and step size based on the obstacle density and electromagnetic interference intensity of the current search area. The purpose of this step is to provide reasonable search range and accuracy control for local path planning.

[0044] The specific implementation of step S06 involves performing dynamic parameter adaptive adjustment, achieving online optimization of algorithm parameters based on the principle of adaptive control. First, a step function is designed to calculate the rate of change threshold. This function takes the current iteration number, the path length change rate, and the population convergence index as input parameters. The first rate of change threshold is set to 0.05–0.1 to determine whether to enhance global exploration capabilities. The second rate of change threshold is set to 0.15–0.25 to determine whether to strengthen local exploitation capabilities. Then, a parameter adjustment strategy is established. When the path length change rate is less than the first rate of change threshold, the global exploration capability is enhanced by increasing the pulse emission rate to 0.7–0.9 and decreasing the loudness to 0.1–0.2, expanding the search range and avoiding premature convergence. When the path length change rate is greater than the second rate of change threshold, the local exploitation capability is strengthened by decreasing the pulse emission rate to 0.2–0.4 and increasing the loudness to 0.6–0.8, performing a refined search near the current optimal solution. Next, fuzzy rule base for parameter adjustment is established using fuzzy control theory, determining the corresponding parameter adjustment range based on different convergence states and iteration progress. Finally, a parameter smoothing transition mechanism is implemented to avoid abrupt changes during parameter adjustment that could affect algorithm stability. The step size for parameter adjustment is controlled within 10% to 20% of the original value. The purpose of this step is to achieve intelligent adaptive adjustment of algorithm parameters.

[0045] The specific implementation of step S07 involves verifying the effectiveness of the path and conducting a risk assessment, establishing a path safety verification mechanism based on risk assessment theory. First, an electromagnetic interference (EMI) intensity calculation function is used to assess the electromagnetic environment safety of the planned path. This function takes the set of path coordinate points, EMI source location data, EMI source power data, and propagation distance attenuation coefficient as input parameters, and outputs the EMI intensity value for each path point. The EMI intensity value is normalized and ranges from 0 to 1. Then, a graded response mechanism is established. When the EMI intensity value is in the range of 0 to 0.3, it is considered a low-interference environment, and normal inspection procedures are performed without additional measures. When the EMI intensity value is in the range of 0.3 to 0.7, it is considered a medium-interference environment, and a multi-band adaptive filtering system is activated to suppress interference. This system uses a blind source separation algorithm to identify interference source signals of different frequencies and processes various EMI signals through a combination of spatial filtering and temporal filtering techniques. When the EMI intensity value is in the range of 0.7 to 1.0, it is considered a high-interference environment, requiring path replanning and an increase in the safety distance, which is set to 1.5 to 2 times the original distance. Next, an obstacle collision risk assessment is performed, calculating the distance from each point on the path to the nearest obstacle. Segments with distances less than the safety threshold of 2-5 meters are marked as high-risk. Then, the communication signal strength is evaluated to ensure that the signal strength throughout the entire path meets remote control requirements; the signal strength threshold is set to -80dBm to -60dBm. Finally, the path validity verification results and corresponding risk level assessment report are output. The purpose of this step is to ensure the safety and feasibility of the planned path.

[0046] It should be noted that the first key technical concept of this invention is multi-level spatial modeling and stage feature fusion technology. This technology constructs a three-dimensional environmental model with three spatial levels: underground pipelines, ground equipment, and overhead lines. Combined with a dynamic feature database from the construction phase, it achieves the organic integration of spatial and temporal information. Compared to traditional single-level modeling methods, this technology can more comprehensively describe the complex spatial structure and dynamic changes of converter stations, significantly improving the environmental adaptability and accuracy of path planning. Through the principle of layered modeling, the geometric accuracy and update frequency of different spatial levels can be differentiated according to actual needs, ensuring the modeling accuracy of key areas while improving overall modeling efficiency.

[0047] The second key technical approach is to improve the two-layer game-theoretic optimization framework of the Bat Algorithm. This technique transforms the traditional single-layer optimization problem into a two-layer game problem involving upper-layer path optimization and lower-layer parameter adjustment. Lyapunov stability theory ensures algorithm convergence, achieving coordinated optimization of path planning and algorithm parameters. Compared to traditional fixed-parameter optimization algorithms, this framework can dynamically adjust algorithm parameters based on feedback information during the optimization process, effectively avoiding local optimum traps and significantly improving global optimization capabilities. The two-layer game mechanism enables mutual influence between the upper and lower layers through coupling terms, allowing path quality and algorithm performance to mutually promote each other during optimization, achieving a better overall effect.

[0048] The third key technological approach is an adaptive sensing and hierarchical response mechanism for electromagnetic interference. Based on the complexity and dynamics of the electromagnetic environment during the converter station construction phase, this technology establishes an electromagnetic interference intensity calculation model and a multi-band adaptive filtering system, enabling intelligent identification and hierarchical processing of electromagnetic interference of varying intensities. Compared to traditional fixed-threshold interference processing methods, this mechanism can dynamically adjust the response strategy according to the actual electromagnetic environment, ensuring both the safety of the inspection mission and maximizing inspection efficiency. Through blind source separation algorithms and joint spatial-temporal filtering techniques, the system can effectively suppress multi-source interference signals in complex electromagnetic environments, significantly improving the operational capabilities of UAVs in environments with strong electromagnetic interference.

[0049] The fourth key technological approach is a dynamic parameter adaptive adjustment and risk assessment verification mechanism. This technology achieves online intelligent adjustment of algorithm parameters through step functions and fuzzy control theory, and combined with multi-dimensional risk assessment, ensures the reliability and security of path planning results. Compared with traditional static parameter setting methods, this mechanism can adjust the algorithm behavior in real time according to the optimization process and environmental changes, effectively balancing the relationship between global exploration and local development, and improving the convergence speed and solution quality of the algorithm.

[0050] The synergistic effect of these four key technological approaches forms a complete intelligent inspection path planning system. Multi-level modeling provides a precise environmental basis for two-level optimization, two-level optimization provides the optimal path scheme for electromagnetic interference handling, electromagnetic interference sensing provides environmental feedback information for dynamic adjustment, and the dynamic adjustment mechanism provides the entire system with adaptive capabilities. This synergistic effect has significant advantages over existing technologies: First, it achieves integrated design of spatial modeling, path optimization, interference handling, and risk control, avoiding performance losses caused by independent design of each module; second, the integration of multiple technologies significantly improves the robustness and adaptability of the system, enabling it to cope with the complex and ever-changing environmental conditions during the converter station construction phase; and third, it establishes a closed-loop feedback optimization mechanism, enabling the system to continuously improve and optimize itself during execution, achieving truly intelligent inspection path planning.

[0051] It should be noted that this invention also solves the following technical problem: the difficulty in achieving real-time dynamic adjustment of UAV path planning in the ever-changing construction environment of converter station construction sites. During converter station construction, factors such as the uncertainty of construction progress, frequent changes in temporary obstacles, and dynamic updates to equipment installation status lead to a continuously changing inspection environment. Traditional static path planning methods, once a fixed path is generated, struggle to adapt to environmental changes, requiring manual replanning to address new constraints. This passive adjustment method is not only inefficient but may also miss critical inspection opportunities. This invention constructs a multi-level spatial database to update spatial information at three levels: underground pipelines, ground equipment, and overhead lines. Combined with a stage-specific feature parameter set, it dynamically records environmental characteristic changes at different construction stages, providing complete dynamic environmental perception capabilities for path planning. Simultaneously, by utilizing the swarm intelligence characteristics and dynamic parameter adjustment mechanism of the improved bat algorithm, it achieves automatic adaptation of path planning strategies to environmental changes. When a significant change in environmental constraints is detected, it can quickly re-search for the optimal path, effectively solving the problem of real-time dynamic adjustment of path planning in ever-changing construction environments.

[0052] Furthermore, this invention addresses the technical problems of objective conflict and difficulty in determining weights that traditional path planning methods often encounter when handling multi-objective optimization during the converter station construction phase. UAV inspections during converter station construction require simultaneous consideration of multiple mutually constraining optimization objectives, such as minimizing path length, maximizing equipment coverage, minimizing emergency obstacle avoidance response time, and minimizing electromagnetic interference false alarm rate. Traditional weighted summation methods or constraint handling methods struggle to accurately reflect the importance relationships between objectives, and the setting of weight coefficients often relies on empirical judgment, lacking theoretical basis, leading to optimization results biased towards certain objectives while neglecting other important indicators. The two-layer game-theoretic optimization framework designed in this invention transforms the multi-objective optimization problem into a game process between an upper-layer path optimization model and a lower-layer parameter adjustment model. The upper-layer model focuses on the coordinated optimization of path-related indicators, while the lower-layer model focuses on the dynamic balance of algorithm control parameters. The two models achieve information exchange and strategy coordination through coupling terms, avoiding the difficulty in determining weight coefficients in traditional methods. Simultaneously, the game mechanism automatically finds the balance point between objectives, achieving adaptive weight allocation for multi-objective optimization and effectively solving the technical challenges of objective conflict and weight determination.

[0053] Specifically, the principle of this invention is as follows: The fundamental principle that enables this invention to solve the technical problems of poor convergence and low reliability of path planning algorithms for UAV inspections under electromagnetic interference environments during the construction phase of converter stations lies in establishing a closed-loop control mechanism for electromagnetic interference perception and adaptive algorithm optimization. First, the construction of a phase feature database enables dynamic modeling of the distribution and intensity of electromagnetic interference sources at different construction phases. By recording key information such as the start-up and shutdown status of equipment during the electrical installation phase and the energization status of high-voltage equipment during the commissioning and acceptance phase, accurate electromagnetic environment constraints are provided for the algorithm, allowing path planning to consider the impact of electromagnetic interference factors in advance. Second, the improved bat algorithm combined with Lyapunov stability theory is the core technology for solving the convergence problem. By constructing a Lyapunov function to establish sufficient conditions for algorithm convergence, it ensures that the algorithm will not fall into local optima or diverge due to fluctuations in the objective function caused by electromagnetic interference in the high-dimensional complex solution space. Simultaneously, the bio-inspired characteristics of the bat algorithm are utilized to achieve adaptive search in the electromagnetic interference environment. Third, the dynamic parameter adaptive adjustment mechanism is a key element in improving algorithm stability. By monitoring the path length change rate and comparing it with a set threshold, it automatically adjusts the pulse emission rate and loudness parameters when electromagnetic interference causes abnormal convergence speed. This enhances global exploration capabilities by expanding the search range to avoid local convergence and strengthens local exploration capabilities by improving search accuracy to ensure the quality of the optimal solution, achieving adaptive matching between algorithm parameters and electromagnetic interference intensity. Fourth, the electromagnetic interference intensity calculation function accurately assesses the electromagnetic interference risk at each point on the planned path by comprehensively considering factors such as interference source location, power, and propagation attenuation, providing a quantitative assessment basis for path execution reliability. Fifth, the multi-band adaptive filtering system uses a blind source separation algorithm to identify electromagnetic interference signals of different frequencies. Combined with spatial and temporal filtering techniques, it targets and suppresses various types of interference, effectively protecting the UAV's communication and navigation systems from electromagnetic interference and ensuring the reliability of path execution. Sixth, the two-layer game optimization framework decomposes the complex optimization problem under electromagnetic interference constraints into two sub-problems: path efficiency optimization and parameter adaptive adjustment. This avoids the optimization difficulties caused by conflicting multiple constraints in a single model, achieving a synergistic improvement in convergence and reliability.

[0054] The following provides a specific embodiment 1 of the present invention, and the specific implementation of each step in this embodiment 1 is described in detail below.

[0055] The specific implementation of step S01 involves establishing a three-dimensional environment model of the converter station and constructing a multi-level spatial database using the principle of layered modeling. The formula for calculating point cloud registration error is as follows: ; In the formula, The total registration error of the point cloud; The first point in the source cloud The coordinate vector of each point; For the target point cloud The coordinate vectors of the corresponding points; To determine the number of registration point pairs; For the first Rotation matrix for secondary registration; To determine the number of registration iterations; The smoothing factor ranges from 0.01 to 0.05. The adaptive grid density adjustment formula is expressed as follows: ; In the formula, For the first Layer mesh density; The base grid density is set to 100-500 grid cells per square meter; For the first The complexity coefficient of the layer; For the first Geometric complexity metrics for layers; The values ​​are 1, 2, and 3, corresponding to the underground pipeline layer, the surface equipment layer, and the overhead line layer, respectively. The method for obtaining these parameters is as follows: and The data was acquired using a 3D laser scanning device with a scanning accuracy of 2–5 millimeters. The results, including the rate of curvature change and boundary density, are calculated using a geometric feature extraction algorithm. The accuracy is set according to the different levels: 0.3 for underground pipelines, 0.5 for ground equipment, and 0.2 for overhead lines.

[0056] The specific implementation of step S02 involves constructing a construction phase feature database, recording dynamic feature parameters for different construction phases based on the principle of time-series modeling. The formula for calculating the electromagnetic field intensity distribution is as follows: ; In the formula, For position At any moment The electromagnetic field strength; This represents the total number of electromagnetic interference sources. For the first The power of each interference source; For the first One interference source at time The gain function; For the observation point to the th The distance between the interference sources; For the first Attenuation coefficient of each interference source; For the first The angular frequency of the interference source; For the first The initial phase of each interference source. The parameter acquisition method is as follows: Obtain the power rating from the equipment nameplate or the actual measured power. The status is determined based on the equipment's operating status; a shutdown status is 0, and an operating status is 1. The value is determined based on the characteristics of the transmission medium, and is taken as 0.001 to 0.01 in air; and Acquired through measurements using a spectrum analyzer.

[0057] The specific implementation of step S03 involves initializing the parameters of the improved bat algorithm and setting the algorithm control parameters based on the principle of swarm intelligence optimization. The algorithm convergence criterion function is expressed as follows: ; In the formula, For the first The Lyapunov function value in the next iteration; The length of the sliding window is set to 10. For the first The optimal objective function value in the next iteration. When the algorithm converges, it is determined that the algorithm has converged. The convergence threshold is set to 0.001. It is obtained by comparing the objective function values ​​of individuals in the population during each iteration; Set according to the accuracy requirements of path planning.

[0058] The specific implementation of step S04 involves establishing a multi-objective optimization function and constructing a multi-objective optimization model using a two-layer game optimization framework. The objective function of the upper-level path optimization model is expressed as follows: ; In the formula, The objective function is optimized for the upper layer; These are the weighting coefficients; This represents the total number of path nodes. For the first The three-dimensional coordinate vectors of each path node; Total number of devices; The number of devices covered; For the first The obstacle avoidance response time of each node; For the first The electromagnetic interference intensity of each node. The objective function of the lower-level parameter adjustment model is expressed as follows: ; In the formula, To optimize the objective function for the lower level; This is the adjustment coefficient; Pulse emission rate; Loudness; For frequency; This is an indicator of population diversity. The parameters are obtained as follows: The default values ​​are set according to the priority of the inspection task, with values ​​of 0.3, 0.25, 0.25, and 0.2 respectively. The default values ​​are set according to the algorithm performance requirements, with values ​​of 0.4, 0.3, 0.2, and 0.1 respectively. It is obtained by calculating the average Euclidean distance between individuals in the population.

[0059] The specific implementation of step S05 involves calculating the local path cost and sampling parameters, and determining the basic parameters for path planning based on heuristic search principles. The formula for calculating the maximum sampling distance is as follows: ; In the formula, Maximum sampling distance; The flight speed of the drone; For battery life; The maximum sampling step size is calculated using the following formula, representing the communication distance: ; In the formula, This is the maximum sampling step size; This is the step size coefficient, with a value ranging from 2 to 5; To control accuracy, the parameter acquisition method is as follows: The value is determined based on the model and specifications of the drone, and ranges from 5 to 15 meters per second. The time range is calculated based on battery capacity and power consumption, and is between 30 and 60 minutes. The range is determined based on the specifications of the communication equipment, and is between 2 and 5 kilometers. The values ​​were obtained through flight testing and range from 0.5 to 2 meters.

[0060] The specific implementation of step S06 involves performing dynamic parameter adaptive adjustment, achieving online optimization of algorithm parameters based on the principle of adaptive control. The threshold for the rate of change calculated using the step function is expressed as follows: ; ; In the formula, For the first The first rate of change threshold for the next iteration; For the first The second rate of change threshold for the next iteration; and The base thresholds were set to 0.05 and 0.15 respectively. As a regulating factor; This represents the maximum number of iterations. For the first The convergence index for the next iteration. The formula for dynamically adjusting the parameters is as follows: ; ; In the formula, and The first The pulse emission rate and loudness of the next iteration; and To adjust the step size; and This is the adjustment coefficient; For the first The rate of change of path length in each iteration. The parameter acquisition method is as follows: Set to 0.1 and 0.15; Set to 0.2 and 0.25; and Set to 10% to 20% of the original value; and Set to 2.0 and 1.5.

[0061] The specific implementation of step S07 involves verifying the effectiveness of the path and conducting a risk assessment, establishing a path safety verification mechanism based on risk assessment theory. The electromagnetic interference intensity calculation function is expressed as follows: ; In the formula, Path node Normalized electromagnetic interference intensity at the location; This represents the total number of interference sources. For the first Normalized power of each interference source; For the first The directivity coefficient of each interference source; For the first The position vectors of the interference sources; For the first The influence range parameters of each interference source. The method for obtaining these parameters is as follows: Obtained by power measurement and normalization to the 0-1 range; The value is determined based on the equipment's radiation pattern and ranges from 0.5 to 1.0. The value is determined based on the equipment type and power level, and ranges from 10 to 100 meters.

[0062] The implementation method for using a multi-level spatial database to store geometric and equipment information at different spatial levels of the converter station is the same as described above and will not be repeated in detail here. The implementation method for using a stage feature parameter set to describe the environmental characteristics and constraints at different construction stages is the same as described above and will not be repeated in detail here. The implementation method for the improved bat algorithm, a swarm intelligence optimization algorithm based on bat echolocation behavior, is the same as described above and will not be repeated in detail here. The implementation method for using Lyapunov stability theory to guarantee the convergence of the algorithm in high-dimensional solution space is the same as described above and will not be repeated in detail here.

[0063] The principle and effect of the formula are explained below: Point cloud registration error formula Based on the principle of least squares, accurate registration is achieved by minimizing the sum of squared Euclidean distances between the source point cloud and the target point cloud. At the same time, a smoothing term is introduced to constrain the continuous change of the rotation matrix, which effectively avoids the local optimum problem in traditional registration algorithms. Compared with existing technologies, this method significantly improves the registration accuracy and stability of point clouds scanned multiple times, providing high-quality basic data for subsequent 3D modeling.

[0064] Mesh density adaptive adjustment formula Based on the principle of exponential decay, the mesh density is dynamically adjusted according to the geometric complexity of different spatial levels. While ensuring modeling accuracy, the computational complexity is effectively controlled. Compared with the fixed density modeling method, it achieves the optimal balance between accuracy and efficiency, and is particularly suitable for the modeling needs of multi-level complex spatial structures such as converter stations.

[0065] Formula for calculating electromagnetic field intensity distribution Based on electromagnetic field propagation theory, this paper comprehensively considers the spatial attenuation, time-varying characteristics, and frequency features of multi-source interference. It simulates the propagation loss of electromagnetic waves in space through an exponential attenuation term and describes the time-varying interference signal through a cosine term. Compared with the static electromagnetic field model, it can more accurately predict the dynamic changes of complex electromagnetic environment during the construction phase and provide reliable electromagnetic interference early warning for path planning.

[0066] Lyapunov function Based on stability theory, the algorithm's convergence state is determined by calculating the rate of change of the objective function value through a sliding window. This effectively avoids the oscillation problem of traditional convergence determination methods and ensures stable convergence of the algorithm in high-dimensional complex solution spaces. Compared with the stopping criterion of a fixed number of iterations, it significantly improves optimization efficiency and solution quality.

[0067] Two-level game optimization function and Based on the principles of game theory, path optimization and parameter adjustment are decoupled into two mutually influential decision problems. The upper-level function comprehensively considers four key performance indicators: path length, equipment coverage, obstacle avoidance time, and electromagnetic interference. The lower-level function achieves intelligent adjustment of algorithm parameters through nonlinear combination. Compared with traditional single-objective optimization methods, this method achieves multi-objective collaborative optimization and significantly improves the overall performance of path planning.

[0068] Sampling parameter calculation formula and Based on the principle of constrained optimization, the search range is determined by constraints on endurance and communication distance, and the search step size is determined by constraints on control accuracy. This effectively prevents the algorithm from searching beyond the operational capabilities of the UAV, and compared with unconstrained search methods, it ensures the executability and security of the planning results.

[0069] Dynamic threshold adjustment formula and Based on the principle of adaptive control, the threshold is dynamically adjusted by a nonlinear combination of iterative progress and convergence state. The parameter adjustment formula uses the hyperbolic tangent function to achieve a smooth transition. Compared with the fixed threshold method, it can intelligently adjust the algorithm behavior according to the real-time state of the optimization process, which significantly improves the algorithm's adaptability and robustness.

[0070] Electromagnetic Interference Intensity Calculation Function Based on the principle of field strength superposition, this study comprehensively considers the power contribution, spatial attenuation, and directional effects of multiple interference sources. By simulating the near-field effect through an exponential attenuation term, it can more accurately assess the interference intensity distribution in complex electromagnetic environments compared to a simple distance-inverse proportional model. This provides a scientific quantitative basis for the graded response mechanism and ensures the safe execution of inspection tasks under different electromagnetic environments.

[0071] It should be noted that the variables involved in this invention are explained in detail in Tables 1 and 2.

[0072] Table 1. Variable Explanation Table (Part 1)

[0073] Table 2. Variable Explanation Table (Part Two)

[0074] To better understand and implement this invention, the following is a specific application scenario example 2: A technical team undertook an intelligent unmanned aerial vehicle (UAV) inspection task during the construction phase of an ±800kV ultra-high voltage converter station, which covers an area of ​​approximately 150,000 square meters. The system includes key electrical equipment such as 12 converter transformers, 24 smoothing reactors, and 48 sets of valve hall equipment. The project team employed the path planning method of this invention to optimize the design of drone inspection paths for the three key stages of the converter station's construction.

[0075] First, the technical team established a 3D environment model of the converter station. Using a lidar scanning system, they acquired 3D point cloud data of the entire converter station area, achieving a scanning accuracy of ±5mm, and obtaining approximately 2.8×... Valid point cloud data. Based on equipment height characteristics, the entire converter station space is divided into three layers: underground pipeline layer (-3m to 0m), ground equipment layer (0m to 25m), and overhead line layer (25m to 55m). The underground pipeline layer includes infrastructure such as cable trenches, drainage systems, and grounding networks, with a total of 1580 key node coordinates. The ground equipment layer covers major electrical equipment such as converter transformers, reactors, and switchgear, marking 336 inspection target points. The overhead line layer includes high-voltage equipment such as DC transmission lines, lightning protection wires, and insulator strings, identifying 258 key inspection locations.

[0076] When constructing the construction phase characteristic database, the technical team divided the construction process into three overlapping periods based on the project schedule: the civil engineering phase (0-18 months), the electrical installation phase (12-36 months), and the commissioning and acceptance phase (30-42 months). Characteristic parameters for the civil engineering phase include foundation pouring completion rate, site flatness, and construction machinery distribution density, with the foundation pouring completion rate gradually increasing from 5% to 98%. The electrical installation phase focuses on recording parameters such as equipment availability rate, installation progress, and temporary power distribution, with the equipment availability rate increasing from 15% to 100%. The core parameters for the commissioning and acceptance phase include equipment power-on status, electromagnetic interference source intensity distribution, and safety protection areas, with the number of electromagnetic interference sources increasing from zero to 84 effective interference points.

[0077] When initializing and improving the bat algorithm parameters, the technical team set the population size to 50 individuals and the maximum number of iterations to 200. The initial pulse emission rate was set to 0.5, the initial loudness to 0.25, and the frequency search range to the [0,2] Hz interval. Based on Lyapunov stability theory, a convergence criterion function was constructed, which determines that the algorithm has converged when the rate of change of the optimal solution is less than 0.001 in 15 consecutive iterations. To prevent getting trapped in local optima, a population diversity maintenance mechanism was implemented, automatically increasing the intensity of random perturbations when the population diversity index falls below 0.15.

[0078] In establishing the multi-objective optimization function, the technical team constructed an objective function containing four optimization objectives. The objective of minimizing the total path length requires that the single inspection path not exceed 12km to ensure the drone's endurance. The objective of maximizing equipment coverage requires that a single inspection cover no less than 85% of key equipment points. The objective of minimizing emergency obstacle avoidance response time sets the response time to no more than 0.8s to ensure that the drone can avoid sudden obstacles in a timely manner. The objective of minimizing the electromagnetic interference false alarm rate controls the false alarm rate to within 5%, improving the reliability of inspection data. In the two-layer game optimization framework, the weight coefficients of the objective function of the upper-layer path optimization model are as follows: path length term weight 0.35, equipment coverage reciprocal term weight 0.28, obstacle avoidance response time term weight 0.22, and electromagnetic interference false alarm rate logarithm term weight 0.15. The lower-layer parameter adjustment model achieves adaptive parameter adjustment through a combination of pulse emission rate square term, loudness linear term, frequency sine term, and population diversity index term.

[0079] During the calculation of local path cost and sampling parameters, the technical team determined key constraints based on the technical parameters of the industrial-grade UAV used. The maximum flight radius of the UAV was set to 8km, the maximum flight altitude was limited to 120m, and the inspection speed was controlled within the range of 5-15m / s. The maximum sampling distance was dynamically calculated based on the current battery level and wind speed conditions, typically varying between 3-6km. The coordinates of the farthest sampling point were determined using three parameters: current position, target direction angle, and maximum sampling distance. The maximum sampling step size was adaptively adjusted based on terrain complexity and obstacle density, set at 50m in flat areas and reduced to 15m in complex areas.

[0080] The dynamic parameter adaptive adjustment mechanism uses a piecewise step function to achieve intelligent parameter adjustment. The first rate of change threshold is set to 0.02, and the second rate of change threshold is set to 0.08. When the path length change rate is less than 0.02, the algorithm determines that the convergence is too fast, automatically increases the pulse emission rate to the range of 0.7-0.9, and decreases the loudness to the range of 0.1-0.2 to enhance global exploration capabilities. When the path length change rate is greater than 0.08, the algorithm determines that the convergence is too slow, decreases the pulse emission rate to the range of 0.2-0.4, and increases the loudness to the range of 0.4-0.6 to strengthen local exploration capabilities. In the first 30% of the iteration process, global exploration is mainly performed; in the middle 40% of the process, global and local searches are balanced; and in the last 30% of the process, local fine-tuning search is the focus.

[0081] In the path validity verification and risk assessment phase, the technical team developed an electromagnetic interference (EMI) intensity calculation function. This function comprehensively considers factors such as interference source power, propagation distance, frequency characteristics, and environmental attenuation. During the civil construction phase, since large electrical equipment was not yet powered on, the EMI intensity value was mainly in the range of [0, 0.15], and standard inspection procedures were performed at this time. During the electrical installation phase, some equipment began trial operation, and the EMI intensity value rose to the range of [0.25, 0.45], at which point the system automatically activated the multi-band adaptive filtering mechanism. During the commissioning and acceptance phase, all equipment operated at full power, and the EMI intensity value in some areas reached the range of [0.75, 0.95]. The system replanned the path and increased the safety distance from the standard 30m to 80m.

[0082] like Figure 2 As shown, the technical team compared and analyzed the electromagnetic interference distribution characteristics of the three construction phases. During the civil engineering phase, interference sources mainly came from construction machinery and equipment, and were relatively few in number and low in intensity. During the electrical installation phase, the number of interference sources increased significantly, mainly concentrated in the substation area. During the commissioning and acceptance phase, interference sources covered the entire station, reaching peak intensity. The multi-band adaptive filtering system used a blind source separation algorithm to process mixed interference signals, successfully identifying interference components in different frequency bands, including power frequency 50Hz, switching frequency 2-5kHz, and high-frequency harmonics 20-50kHz. Spatial filtering technology achieved directional interference suppression through an array of eight distributed sensors, achieving a suppression ratio of 25dB. Time-domain filtering technology used an adaptive notch filter, achieving a suppression depth of over 30dB for interference at specified frequencies.

[0083] like Figure 3 As shown, the drone inspection paths exhibit distinct evolutionary characteristics across different construction phases. During the civil engineering phase, the paths are relatively simple, primarily involving structural inspections along the building's perimeter, with an average single inspection path length of 6.8 km and a coverage rate of 92%. The electrical installation phase sees increased path complexity, requiring detailed inspections by traversing between equipment areas, resulting in an average path length of 9.2 km, while maintaining equipment coverage above 88%. The commissioning and acceptance phase presents the most complex path planning, requiring avoidance of areas with high electromagnetic interference, leading to an average path length of 10.5 km. However, through algorithm optimization, an 86% equipment coverage rate was still achieved.

[0084] The technical team accumulated a wealth of inspection data through 12 months of practical application. A total of 286 inspection flights were completed, covering a total distance of 2458 km, and high-resolution images of 8.7× were obtained. Zhang, infrared thermal image 3.2× Zhang. 127 equipment defects and construction quality issues were identified, including 15 major safety hazards, all of which were addressed promptly. The drone obstacle avoidance success rate reached 99.7%, with no collisions occurring. The electromagnetic interference false alarm rate was controlled at 3.2%, far below the design target of 5%.

[0085] Table 3 shows the typical problems found during the inspection: Table 3. Statistical Table of Typical Problems Found During Inspections

[0086] Table 2 shows that equipment installation deviation is the most significant problem type, accounting for 33.9% of all issues found. These problems mainly occur during the electrical installation phase and can be promptly detected and corrected using the high-precision positioning and measurement capabilities of drones. Insulator contamination issues are primarily discovered during the commissioning and acceptance phase, accounting for 22.0%, and these problems can be accurately identified using infrared thermal imaging detection.

[0087] Table 4 shows a comparison of inspection efficiency at different construction stages: Table 4 Comparison of Inspection Efficiency at Different Construction Stages

[0088] Table 3 reflects that as the construction phase progresses, the complexity and time consumption of inspections gradually increase, but the problem detection rate also increases accordingly, demonstrating the adaptability and effectiveness of the method of this invention.

[0089] The technical team also conducted an in-depth analysis of the algorithm's convergence performance. Under a maximum iteration limit of 200, the improved bat algorithm converged on average on the 138th iteration, achieving a convergence stability of 95.8%. Compared to the traditional genetic algorithm, which requires an average of 185 iterations, the convergence efficiency was improved by 25.4%. The algorithm demonstrated good robustness in handling path optimization problems under multiple constraints, maintaining stable optimization performance even in environments with high electromagnetic interference.

[0090] This invention represents a significant technological advancement compared to traditional manual inspection and simple UAV inspection methods. First, by establishing a multi-level spatial database and a dynamically updated set of stage-specific characteristic parameters, it achieves accurate modeling of the complex construction environment of converter stations, overcoming the limitations of traditional methods in adapting to dynamic changes during the construction phase. Second, the improved bat algorithm, combined with Lyapunov stability theory, ensures global convergence in the high-dimensional solution space, avoiding the pitfalls of traditional heuristic algorithms that easily get trapped in local optima. Third, the two-layer game-theoretic optimization framework, through the coupling of upper and lower layer models, achieves coordinated optimization of path planning and parameter adjustment, improving the algorithm's adaptability and optimization effect. Fourth, the dynamic parameter adaptive adjustment mechanism intelligently adjusts the search strategy based on the algorithm's convergence state, achieving an optimal balance between global exploration and local exploitation. Finally, the multi-band adaptive filtering system, through blind source separation and space-time joint filtering techniques, effectively suppresses complex electromagnetic interference at the construction site, ensuring the accuracy and reliability of inspection data. These technological innovations fundamentally solve the key technical challenges faced by UAV inspection during the converter station construction phase, such as complex path planning, poor environmental adaptability, and weak anti-interference capabilities.

[0091] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent unmanned aerial vehicle (UAV) inspection path planning during the construction phase of a converter station, characterized in that, Includes the following steps: A three-dimensional environmental model of the converter station was established, constructing three spatial layers: underground pipeline layer, ground equipment layer, and overhead line layer. Spatial coordinates, obstacle distribution, and equipment location information of each layer were collected to form a multi-layered spatial database. A construction phase feature database was also constructed, recording the equipment distribution status, construction progress, safety constraints, and electromagnetic interference source distribution of each phase according to the different characteristics of the civil engineering phase, electrical installation phase, and commissioning and acceptance phase, and establishing a dynamically updated set of phase feature parameters. The parameters of the improved bat algorithm were initialized. A multi-objective optimization function was established, with the objectives of minimizing the total path length, maximizing equipment coverage, minimizing emergency obstacle avoidance response time, and minimizing the electromagnetic interference false alarm rate, and a two-layer game optimization framework was constructed. Calculate local path cost and sampling parameters. Based on the current UAV position and target point position, calculate the maximum sampling distance, coordinates of the farthest sampling point, and maximum sampling step size to form a basic parameter set. Perform dynamic parameter adaptive adjustment. Calculate the first and second rate of change thresholds through a step function. Dynamically adjust the pulse emission rate and loudness parameters based on the current iteration number and convergence status. Perform path validity verification and risk assessment. Calculate the electromagnetic interference intensity value of the planned path using an electromagnetic interference intensity calculation function.

2. The method for intelligent inspection path planning by unmanned aerial vehicles during the construction phase of a converter station according to claim 1, characterized in that, The steps for establishing the multi-level spatial database are as follows: Specifically, geometric information and equipment information of different spatial levels of the converter station are obtained through three-dimensional laser scanning to form a structured spatial coordinate dataset.

3. The method for intelligent inspection path planning by unmanned aerial vehicles during the construction phase of a converter station according to claim 2, characterized in that, The set of stage feature parameters is specifically a dynamically updated set of parameter vectors describing the environmental characteristics and constraints of different construction stages, including construction plans, equipment installation status, safety specifications, and electromagnetic interference source location information.

4. The method for intelligent inspection path planning by unmanned aerial vehicles during the construction phase of a converter station according to claim 3, characterized in that, The improved bat algorithm is specifically a swarm intelligence optimization algorithm based on the echolocation behavior of bats. It searches for the optimal solution by simulating bats emitting ultrasonic waves and receiving echoes. During the construction phase of the converter station, the impact of electromagnetic interference generated at different construction stages on the convergence of the algorithm is considered.

5. The method for intelligent inspection path planning by unmanned aerial vehicles during the construction phase of a converter station according to claim 4, characterized in that, The initialization step of the improved bat algorithm parameters also includes setting convergence criteria based on Lyapunov stability theory.

6. The method for intelligent inspection path planning by unmanned aerial vehicles during the construction phase of a converter station according to claim 5, characterized in that, The two-layer game optimization framework specifically includes an upper-layer path optimization model with path efficiency as the objective and a lower-layer parameter adjustment model with parameter adjustment as the objective. The two objective functions influence each other through coupling terms consisting of the product of pulse emission rate and path length weights, the product of loudness and equipment coverage weights, and the product of frequency and obstacle avoidance time weights.

7. The method for intelligent inspection path planning by unmanned aerial vehicles during the construction phase of a converter station according to claim 6, characterized in that, The objective function of the upper-level path optimization model includes a path length term, a reciprocal term for device coverage, an obstacle avoidance response time term, and a logarithmic term for electromagnetic interference false alarm rate.

8. The method for intelligent inspection path planning by unmanned aerial vehicles during the construction phase of a converter station according to claim 7, characterized in that, The objective function of the lower-level parameter adjustment model includes a pulse emission rate square term, a loudness linear term, a frequency sine term, and a population diversity index term.

9. The method for intelligent inspection path planning by unmanned aerial vehicles during the construction phase of a converter station according to claim 8, characterized in that, The set of basic parameters is specifically determined based on the UAV's endurance and communication range, with the maximum sampling distance used to limit the search range and step size of a single path planning operation by the UAV.

10. The method for intelligent inspection path planning by unmanned aerial vehicles during the construction phase of a converter station according to claim 9, characterized in that, The step function is specifically calculated and dynamically adjusted based on the current iteration number, the path length change rate, and the population convergence index, using a first change rate threshold and a second change rate threshold.

Images