Transformer station unmanned aerial vehicle three-dimensional route real-time planning and autonomous obstacle avoidance system

Abstract

The invention provides a transformer substation unmanned aerial vehicle three-dimensional route real-time planning and autonomous obstacle avoidance system, and relates to the technical field of non-electrical variable control or adjustment systems. Through cooperative work of a navigation mode switching module and a near-field navigation control module, an electric field of equipment is used as a natural navigation beacon; and high-precision autonomous flight is realized in a strong electromagnetic interference area with RTK signal failure, and the dependence on external positioning is eliminated, so that the inspection blind area is eliminated. Meanwhile, a fusion environment map containing multi-physics field information is constructed, the value of observation points is scientifically quantified according to waypoint inspection efficiency indexes, and a path planning algorithm aims at optimizing a comprehensive target fusing energy consumption, risks and inspection efficiency. The design not only ensures that the unmanned aerial vehicle can safely get close to the core equipment, but also enables the planned route to be the most valuable path rather than the shortest path, and finally fundamentally improves the safety, the operation efficiency and the data validity of substation inspection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of non-electric variable control or regulation system, in particular to a real-time planning and autonomous obstacle avoidance system for unmanned aerial vehicle three-dimensional route in a substation. BACKGROUND

[0002] In the operation and maintenance system of modernized substations, unmanned aerial vehicle (UAV) inspection has become an important means to replace high-risk manual operations. The existing technology mainly relies on GPS / RTK for positioning and plans a collision-free path based on a three-dimensional laser point cloud model. However, in strong electromagnetic interference areas such as large transformers and high-voltage busbars, the GPS / RTK signal will be severely attenuated or even completely invalid, which makes it impossible for the UAV to approach the core equipment for detailed detection, forming a "blind area" for inspection. At the same time, the planning algorithm targeting the "shortest path" or "shortest time" ignores the flight energy consumption, aerodynamic stability and the final inspection image quality, which makes it difficult to guarantee the inspection efficiency and data effectiveness.

[0003] The existing typical implementation is as follows: first, a three-dimensional laser point cloud or a real scene model of the equipment and buildings in the substation is constructed by means of LiDAR or oblique photography, forming a high-precision geometric space representation; then, a path planning algorithm such as graph search algorithm (such as A , Dijkstra) or Rapidly-exploring Random Tree (RRT) is used on this geometric model to generate a collision-free shortest or optimal flight route; finally, the UAV is equipped with an RTK receiver to receive the differential correction signal from the ground base station, achieving centimeter-level precise positioning, so as to accurately execute the preset route and complete the inspection and shooting of the specified target.

[0004] In the prior art, a patent application with the publication number CN115562348A and the title of a UAV image technology method based on a substation includes the following research steps: S1, comprehensively understanding the UAV positioning applicable technology, applicable model, route planning and anti-interference measures in the substation, S2, researching the UAV flight control and gimbal adjustment algorithm to realize effective collection of substation inspection images, planning the route by researching the applicable technology and applicable model of UAV image collection in the substation, positioning the UAV, calculating the parallax, depth and obstacle avoidance deviation angle through binocular distance measurement, avoiding obstacles by adjusting the flight angle, making the UAV flight safer, resisting and identifying the interference to reduce the interference, accurately collecting the required images and identifying hidden dangers, quickly understanding the real-time situation and hidden dangers at different positions in the substation through the images collected by the UAV, and improving the UAV inspection efficiency of the substation.

[0005] The existing deficiencies are as follows: 1. Core navigation capabilities heavily rely on the continuity and high quality of external GNSS signals. Substations, as areas with a high concentration of high-voltage, high-current equipment, are filled with strong electromagnetic fields, especially near large main transformers, reactors, high-voltage busbars, and switchgear, where electromagnetic interference (EMI) is extremely severe. This strong electromagnetic environment causes drastic attenuation, multipath effects, and even periodic signal loss in GNSS signals, leading to frequent cycle slips, decreased accuracy, or complete failure in RTK positioning calculations, resulting in a sharp deterioration in the positioning quality factor. Therefore, in the core equipment areas where precise inspection is most needed, UAVs lose their most basic positioning capabilities, creating so-called "inspection no-go zones." This makes close-range, multi-angle observation of critical equipment nodes an impossible task, severely limiting the depth and breadth of inspections. 2. Existing path planning concepts suffer from significant "cognitive limitations." They abstract the substation environment into a purely static geometric space, with planning objectives largely limited to geometrically optimal metrics such as "shortest path" or "shortest time." This model ignores the essential characteristics of a substation as a multi-physics coupled environment. For example, non-uniformly distributed electromagnetic fields in space can affect the stability of the flight control system, while thermal fields generated by equipment (such as thermal plumes) directly interfere with the aerodynamic stability of the UAV. More importantly, it separates flight path planning from the ultimate goal of inspection operations—obtaining high-quality equipment status data. A geometrically optimal path does not mean that every observation point along it will provide the best shooting angle, lighting conditions, or the most reflective thermal characteristics of equipment anomalies, leading to poor data validity and diagnostic accuracy in inspection operations.

[0006] The information disclosed in the background section above is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0007] The purpose of this invention is to provide a real-time three-dimensional flight path planning and autonomous obstacle avoidance system for substation UAVs, in order to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: The substation UAV real-time 3D flight path planning and autonomous obstacle avoidance system specifically includes: Model building module: Before generating a 3D route, it acquires a 3D geometric model, a 3D electromagnetic field intensity distribution map, and a 3D thermal field distribution map in the inspection space containing multiple inspection targets, and then performs spatial registration and data fusion to generate a fused environment map containing multiple physical field attributes. Candidate point evaluation module: Based on the fused environment map, it generates multiple candidate observation points for each inspection target and calculates the waypoint inspection efficiency index for each candidate observation point; Macro-path planning module: used to select the highest waypoint inspection efficiency index from multiple candidate observation points under each inspection target as nodes of the path map, and generate the globally optimal flight route connecting the nodes. The navigation mode switching module is used to monitor the RTK positioning quality factor and the environmental electric field strength scalar of the UAV in real time during flight, and calculate a switching confidence index based on these two parameters to characterize the necessity of mode switching; the mode switching includes switching to a vector navigation mode based on local electric field perception. Near-field navigation control module: After the flight control mode is switched to vector navigation mode, it processes the signals from the airborne electric field sensor array, calculates the local spatial electric field gradient vector of the UAV's location in real time, and generates control commands for each actuator motor of the UAV based on the gradient vector.

[0009] Compared with the prior art, the beneficial effects of the present invention are: 1. Through the navigation mode switching module, the system intelligently judges and automatically switches to the near-field navigation control mode using the switching confidence index; the near-field navigation control module calculates the local electric field gradient and executes the dual-channel control law, using the device's own physical field as a navigation beacon; this avoids dependence on external RTK signals, thus unifying obstacle avoidance and navigation at the physical level; therefore, the UAV can safely and accurately enter the core area of ​​strong electromagnetic interference, eliminating inspection blind spots, achieving reliable flight without blind spots across the entire domain, greatly expanding the depth and breadth of inspection operations, and facilitating data collection; 2. A multiphysics fusion environment map is generated through the model building module, and the waypoint inspection efficiency index, which integrates image quality, information completeness, and flight stability, is calculated by the candidate point evaluation module, quantifying the "value" of each candidate observation point; based on this, the macroscopic path planning module is improved. The algorithm uses a comprehensive transition cost that integrates energy consumption and risk, along with a heuristic function guided by efficiency, to search for paths. This ensures that the planned routes are "most valuable" rather than "shortest." Ultimately, while ensuring safety, it improves the inspection efficiency, data validity, and endurance utilization of a single flight, achieving a fundamental improvement in the overall benefits of inspection tasks. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of the overall system logic block of the present invention; Figure 2 This is a schematic diagram of the model building module of the present invention; Figure 3The logic diagram of the candidate point evaluation module of the application; Figure 4 The logic diagram of the macro path planning module of the application; Figure 5 The logic diagram of the navigation mode switching module of the application; Figure 6 The logic diagram of the near field navigation control module of the application; Figure 7 The real scene collection diagram of the function interface of the system of the application; Figure 8 The path demonstration diagram of the function interface of the system of the application. DETAILED DESCRIPTION

[0011] In order to make the above objectives, characteristics and advantages of the application more obvious and easy to understand, the specific embodiments of the application will be described in detail below with reference to the accompanying drawings.

[0012] In the following description, many specific details are set forth in order to provide a thorough understanding of the application, but the application can also be practiced without other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the concept of the application, therefore the application is not limited to the specific embodiments disclosed below.

[0013] Embodiment one: Please refer to Figures 1 to 8 , the application provides a technical solution: The three-dimensional flight path real-time planning and autonomous obstacle avoidance system of the unmanned aerial vehicle of the substation, applied to the three-dimensional flight path planning of the unmanned aerial vehicle, specifically comprising: The model construction module is used to acquire the three-dimensional geometric model, the three-dimensional electromagnetic field intensity distribution map and the three-dimensional thermal field distribution map in the inspection space containing a plurality of inspection targets before generating the three-dimensional flight path, and then perform spatial registration and data fusion to generate a fusion environment map containing a plurality of physical field attributes; Further explanation: the fusion environment map takes three-dimensional voxels as units, and each three-dimensional voxel contains geometric, electromagnetic and thermal field attribute information; the fusion environment map is denoted as M fused ; the inspection target specifically refers to the key power equipment in the substation that needs to be observed and data collected by the unmanned aerial vehicle under periodic or special state; the key power equipment includes but is not limited to: power transformer, high-voltage circuit breaker, disconnecting switch, high-voltage bus and lead wire, current / voltage transformer and lightning arrester; The three-dimensional geometric model is used to represent the geometric shape of the space obstacle of the inspection space; The three-dimensional electromagnetic field intensity distribution map is used to represent the electromagnetic interference intensity of the inspection space; The three-dimensional thermal field distribution map is used to characterize the impact of equipment heat dissipation in the inspection space; The inspection space is discretized into multiple three-dimensional voxels, and a voxel comprehensive risk index characterizing the comprehensive flight risk of each three-dimensional voxel is calculated. The voxel comprehensive risk index is calculated based on the electromagnetic interference potential and thermal disturbance intensity within the three-dimensional voxel through a preset risk assessment model. In this embodiment, the three-dimensional geometric model is denoted as G. geom The three-dimensional electromagnetic field intensity distribution map is denoted as M. em The three-dimensional thermal field distribution diagram is denoted as M. th The voxel-based risk index is denoted as I. risk The electromagnetic interference potential is denoted as V. em The intensity of the thermal disturbance is denoted as V. th Each three-dimensional voxel is denoted as v.

[0014] A three-dimensional geometric model G representing the geometry of spatial obstacles is obtained by scanning with an airborne lidar on a drone or by using an existing BIM model. geom .

[0015] By scanning multiple inspection targets within the inspection space using a professional field strength meter before deployment at the substation, or by generating a three-dimensional electromagnetic field intensity distribution map M based on electrical engineering simulation software. em .

[0016] Real-time detection is performed during inspections using an onboard infrared thermal imager on a drone, or a three-dimensional thermal field distribution map is generated based on historical data. th .

[0017] When performing spatial registration, the model building module uses the three-dimensional geometric model as the reference coordinate system. By extracting common feature points from the three-dimensional electromagnetic field intensity distribution map and the three-dimensional thermal field distribution map, the latter two are transformed to the reference coordinate system to achieve consistency of multi-source data in spatial location. For each three-dimensional voxel, the voxel composite risk index I risk The model building module performs the following calculation process: 1.1) Regarding the electromagnetic interference potential V em Determination of: Electromagnetic interference potential V em It is a dimensionless normalized value with a range of [0,1], used to quantify the potential interference of the electromagnetic environment within a three-dimensional voxel v to the UAV navigation and communication system. The larger the value, the stronger the interference. Electromagnetic interference potential V em The acquisition method and numerical determination method are as follows: the model building module obtains the three-dimensional electromagnetic field intensity distribution map M from the input. emIn the process, the original measured value E of the electric field intensity corresponding to the center point of the three-dimensional voxel v is extracted. raw Unit: V / m. E is mapped using a preset exponential mapping function. raw Converted to electromagnetic interference potential V em The expression for the exponential mapping function is: Electromagnetic interference potential V em It equals 1 minus e (-k) e ×E raw ) to the power of, where k e It is a conversion coefficient calibrated based on the electromagnetic interference resistance performance of UAVs; this exponential mapping function ensures that when the original field strength is zero, the potential is zero; as the field strength increases, the electromagnetic interference potential V increases. em It nonlinearly approaches 1.

[0018] It should be noted that: V em The calculation formula originates from the saturation effect model in physics, used to describe the phenomenon that the system response gradually reaches an upper limit as the input increases; in mathematical theory, the electromagnetic interference potential V em The function's range is [0, 1), meaning its value monotonically increases from 0 and approaches 1 infinitely. However, in computer floating-point arithmetic implementations, when the original measured value of the electric field strength E... raw Reaching a sufficiently large saturation threshold causes the exponential term "e" to (-k) e ×E raw When the result of the calculation of "power" is considered 0 by the system due to underflow, the electromagnetic interference potential V em The calculation result can reach 1 precisely. Therefore, from the perspective of engineering implementation and data processing, its effective value range is defined and processed as a closed interval [0,1] including the upper limit; this processing method ensures that the model can accurately represent the entire state range from no disturbance (value 0) to fully saturated disturbance (value 1).

[0019] When the electromagnetic interference potential V em The closer E is to 0, the better. raw The smaller the value, the purer the electromagnetic environment in this area, and the less interference it causes to the navigation and communication of drones. When the electromagnetic interference potential V em The closer E is to 1, the better. raw The larger the value, the more saturated the electromagnetic interference in this area, and the greater the interference to the navigation and communication of UAVs. 1.2) For the thermal disturbance intensity V th Determination of: Thermal disturbance intensity V th It is a dimensionless normalized value with a range of [0,1]. It is used to comprehensively evaluate the combined perturbation effect of the updraft and microscale turbulence generated by the heat dissipation of the device within the three-dimensional voxel v on the attitude stability of the UAV. The larger the value, the stronger the perturbation. The micro-scale wind shear coefficient P is characterized by a pre-calculated flow field database; Thermal disturbance intensity V th The acquisition approach and numerical determination method are as follows: the model construction module extracts the temperature T of the three-dimensional voxel v from the input three-dimensional thermal field distribution map M th , and extracts the corresponding micro-scale wind shear coefficient P wind from a pre-calculated flow field database, unit: s -1 .

[0020] The micro-scale wind shear coefficient P wind is acquired in the following manner: In the process of generating the fusion environment map, the pre-calculated flow field database identifies a device with a size greater than a preset threshold as a disturbance source geometry based on the three-dimensional geometric model; and for each three-dimensional voxel v in the inspection space, the spatial relationship between the three-dimensional voxel v and the surface of the nearest disturbance source geometry is calculated, and the input environmental wind speed vector is referred to, and is denoted as V env , to determine a micro-scale wind shear coefficient P wind characterizing the local airflow turbulence intensity in the three-dimensional voxel v.

[0021] To calculate the micro-scale wind shear coefficient P wind of each three-dimensional voxel v, the model construction module performs the following calculation process: The model construction module first traverses all independent device models in the three-dimensional geometric model. By calculating the circumscribed bounding box volume of each device model, a device with a volume greater than a preset "disturbance volume threshold" (including but not limited to a main transformer, a large circuit breaker, etc.) is identified as a disturbance source geometry.

[0022] The environmental wind speed vector V env is a three-dimensional vector representing the direction and speed of the background wind in the open and unobstructed area of the current inspection area; the environmental wind speed vector V env is obtained by querying the local micro-meteorological station data accessed by the system before the task starts; or by the unmanned aerial vehicle hovering at a safe height far from the ground and the device after taking off, and is measured and calibrated by the airborne anemometer. Its value is represented by a vector containing three directional components, with a unit of m / s.

[0023] For the three-dimensional voxel v to be calculated, the model construction module performs the following geometric queries: The nearest disturbance source distance ds: calculate the Euclidean distance from the center point of the three-dimensional voxel v to the nearest point on the surface of all identified disturbance source geometries; the distance is a scalar, with a unit of meters.

[0024] Nearest point surface normal ns: Get the unit outward normal vector of the surface of the perturbation source geometry at the aforementioned nearest point.

[0025] The microscale wind shear coefficient P wind The value is derived from the environmental wind speed scalar U. wind Multiplied by a distance attenuation factor F dist Then multiply by an angle influence factor F angle get.

[0026] Ambient wind speed scalar U wind The ambient wind speed vector V env The modulus, i.e., its size. It is calculated by taking the ambient wind speed vector V... env The sum of the squares of the three components is taken as the square root. This scalar reflects the overall intensity of the background wind force, and the unit is m / s.

[0027] Distance attenuation factor F dist This characterizes the attenuation effect of wind shear intensity with distance; the closer to the disturbance source, the stronger the shear. It is calculated using an exponential decay function, expressed as e raised to the power of (-kd × ds).

[0028] Where ds is the distance to the nearest disturbance source calculated above. kd is a dimensionless "distance decay constant", which is calibrated using a small amount of actual wind tunnel experiments or CFD simulation data to adjust the decay rate.

[0029] Angle Influence Factor F angle This characterizes the effect of the angle between the wind direction and the obstacle surface on wind shear. The disturbance is greatest when the wind blows perpendicularly to the surface; the disturbance is smaller when the wind is parallel to the surface. The calculation method is as follows: taking the ambient wind speed vector V... env The absolute value of the cosine of the angle between the surface normal ns and the nearest point. This cosine is obtained by calculating the dot product of the two unit vectors. In this embodiment, taking the absolute value ensures that a disturbance will occur as long as a normal component exists, regardless of whether it is the windward or leeward side.

[0030] Microscale wind shear coefficient P wind The calculation method conforms to the principles of fluid dynamics: the greater the wind speed, the stronger the wind shear; the closer to a large obstacle, the stronger the wind shear; and the more directly the wind is directed towards the obstacle surface, the stronger the wind shear.

[0031] Thermal disturbance intensity V th The calculation method is as follows: 1.21) By calculating the temperature T and the ambient reference temperature T base The difference is used to obtain the thermal buoyancy gradient G. th Approximate value; 1.22) The normalized thermodynamic buoyancy gradient Gth With the normalized microscale wind shear coefficient P wind Perform a weighted average and set the thermal buoyancy gradient G for each. th With microscale wind shear coefficient P wind The weighting coefficients are w g and w p Weighting coefficient w g and w p The correlation between attitude adjustment and these two parameters in historical flight data was determined by regression analysis, and w g +w p =1; Thermal disturbance intensity V th The calculation employs the weighted average method from operations research, a classic method for multi-attribute decision-making.

[0032] This embodiment determines through regression analysis that, for a selected UAV model, the thermal buoyancy gradient has a greater impact on attitude stability; therefore, the weight is set to w. g =0.7, w p =0.3. In scenarios with high wind speeds, the weight w of the microscale wind shear coefficient can be adjusted online. p The value was increased to 0.6 to adapt to the risk focus under different meteorological conditions and achieve dynamic optimization.

[0033] When the intensity of thermal disturbance V th The closer the value is to 0, the closer the temperature inside the three-dimensional voxel is to the ambient temperature and the more stable the airflow is, making it a more ideal flight area. When the intensity of thermal disturbance V th The closer the value is to 1, the stronger the updraft or the more severe the wind shear within the three-dimensional voxel. Drones flying in this voxel will need to consume more energy to maintain their attitude, and the higher the risk of instability.

[0034] 1.3) The model building module constructs a risk assessment model using Principal Component Analysis (PCA); the specific construction content includes: 1.31) Data Preparation: Collect a set of data containing (V em V th ) and corresponding historical or simulation data samples of the frequency of risk events that occurred in actual flight (the frequency of risk events includes, but is not limited to, the number of attitude instability events or the duration of communication interruption); 1.32) Principal Component Extraction: Perform principal component analysis on the sample dataset to find the first principal component PC1 that maximizes the explanation of the variance of risk events; this principal component is V em and V th A linear combination of principal components, expressed as: principal component PC1 equals coefficient c1 multiplied by V. em Add the coefficient c2 multiplied by Vth The coefficients c1 and c2 are the loadings automatically calculated by the PCA algorithm according to the intrinsic structure of the data, which reflect the true contribution of the two original risk factors to the comprehensive risk, and are both positive values; 1.33) Index generation: finally, the voxel comprehensive risk index I of the three-dimensional voxel v is generated risk defined as the principal component PC1 value corresponding to the three-dimensional voxel, and normalized. Its calculation expression is: the voxel comprehensive risk index I risk equals the coefficient c1 multiplied by the electromagnetic interference potential V of the three-dimensional voxel em plus the coefficient c2 multiplied by the thermal disturbance intensity V of the three-dimensional voxel th .

[0035] The output voxel comprehensive risk index I risk characterizes the comprehensive threat degree of the three-dimensional voxel to the safe flight of the UAV.

[0036] Since the coefficients c1 and c2 are both positive in this embodiment, it is ensured that the increase of any risk factor (V em or V th ) will lead to a monotonic increase of the final voxel comprehensive risk index I risk ; this embodiment determines the combination coefficient through principal component analysis, avoiding the subjectivity of manually setting the weight, and scientifically solves the problem of how to combine risk factors of different physical sources into a unified risk measure, and the result can more accurately reflect the true comprehensive risk.

[0037] It should be noted that: the calculation of the voxel comprehensive risk index I risk uses the principal component analysis (PCA) in multivariate statistical analysis, aiming to find a linear combination that can explain the variability of the data to the greatest extent through dimensionality reduction.

[0038] Calculation logic and proportional relationship: I risk is a linear combination of (V em , V th ) c1xV em +c2xV th . In the principal component analysis PCA, the signs of the loading coefficients c1 and c2 depend on the data correlation. To ensure the proportional relationship, we select the principal component positively correlated with the risk event frequency to ensure that c1 and c2 are both positive values.

[0039] After normalization, the value range of the voxel comprehensive risk index I risk is limited to the interval [0, 1].

[0040] When the voxel comprehensive risk index I risk tends to 0, it indicates that the comprehensive flight risk of the three-dimensional voxel is smaller, and it is easier to be selected by the planned route. When the voxel comprehensive risk index I risk The closer to 1, the greater the three-dimensional voxel comprehensive flight risk, the less likely to be selected by the planned route.

[0041] In this embodiment, if electromagnetic interference is the main cause of flight accidents, c1 calculated by PCA will be significantly greater than c2, so that the voxel comprehensive risk index I risk V em is more sensitive to the change. This data-driven weight determination method makes the risk assessment model adaptive and objective, and the output voxel comprehensive risk index I risk can more truly reflect the comprehensive flight risk in a specific scenario.

[0042] Further description of the model construction module: In order to intuitively display the model construction module, six three-dimensional voxels in different positions are taken as examples for parameter substitution and calculation. Through PCA analysis, the expression of the risk model is determined as: I risk =0.65×V em +0.35×V th ; see the following table for details:

[0043] Voxel A1: both the original measured value of electric field intensity and the thermal disturbance intensity are low, and the calculated voxel comprehensive risk index I risk is 0.06, which is a very safe area and is suitable for being used as a route crossing point.

[0044] For voxels B1 and E1: voxel B1 has strong electromagnetic interference and weak thermal disturbance, while voxel E1 is the opposite, but their final voxel comprehensive risk index I risk is 0.62. This reflects that the model can balance the risks from different sources and unify them into one dimension for comparison, avoiding the confusion of the decision maker between "avoiding electric field" and "avoiding air flow"; Voxel C1: electromagnetic interference is general, but thermal disturbance is very strong, resulting in a voxel comprehensive risk index I risk of 0.41, which is higher than that in a pure area, indicating that this is an area that needs to be passed with caution; the very strong thermal disturbance in this embodiment includes the case above a large heat sink; Voxel D1: electromagnetic interference is close to saturation, even if the thermal disturbance is not strong, its voxel comprehensive risk index I risk is as high as 0.74, which is the highest risk among all voxels and should be marked as a high-cost area. Voxel F1: both risks are at a medium level, and its final voxel comprehensive risk index I risk is also in the middle.

[0045] In specific embodiments of the present application, in order to adapt to the limited precision of digital processing system and improve the computational efficiency, the system further comprises a processing step: setting the calculated voxel comprehensive risk index I risk to 0 if the voxel comprehensive risk index I risk is less than the risk neglect threshold value, which is set to 0.01. risk This processing method is a common technical means for those skilled in the art when implementing numerical algorithms, which aims to effectively process parameters that theoretically tend to approach the target value in engineering, thereby ignoring the extremely small values that have no actual impact on the final result.

[0046] The candidate point evaluation module is used to generate a plurality of candidate observation points for each inspection target based on the fusion environment map, and calculate a waypoint inspection efficiency index of each candidate observation point. Further description: the candidate point evaluation module includes various performance components: for each candidate observation point, the expected image clarity representing its imaging quality, the key feature visibility representing its diagnostic value, and the attitude control margin representing its flight stability are calculated respectively; the expected image clarity, the key feature visibility and the attitude control margin are represented as performance components; and are represented in turn as j=1, 2, 3. The candidate point evaluation module determines the objective weight of each performance component by calculating the dispersion degree of each performance component based on the entropy weight method, and performs weighted summation on the expected image clarity, the key feature visibility and the attitude control margin to obtain the waypoint inspection efficiency index. In this embodiment, the expected image clarity is denoted as V clarity , the key feature visibility is denoted as V feature , the attitude control margin is denoted as M control , and the waypoint inspection efficiency index is denoted as I eff . When generating the plurality of candidate observation points, the candidate point evaluation module generates a set of initial point sets uniformly distributed in space outside the safe distance range according to the preset spherical coordinate system rules with the three-dimensional model of the inspection target as the center, as the candidate observation points. To calculate the waypoint inspection efficiency index I eff of each candidate observation point, the candidate point evaluation module performs the following calculation process: the index of the candidate observation point is denoted as p. 2.1) Expected image clarity V clarityis a dimensionless normalized value, whose value range is [0, 1], used to quantify the contrast and detail performance ability when taking an image at the candidate observation point p due to the light and camera perspective, the larger the value, the clearer the expected image.

[0047] expected image clarity V clarity The acquisition approach and value determination method are: obtaining the current sun position vector and the three-dimensional geometric model of the inspection target; calculating the included angle θ view between the observation vector from the candidate observation point p to the target surface and the surface normal vector, and the included angle θ light between the sun position vector and the surface normal vector. The expected image clarity V clarity is calculated by a light model function, and the expression of the function is: the expected image clarity V clarity is equal to the sum of a basic clarity value and a light gain term. The light gain term is in direct proportion to the cosine function value of the included angle θ view and θ light , and ensures that the result is normalized. The model aims to select a position that avoids backlight and highlights the surface texture at an angle; 2.2) Key feature visibility V feature is a dimensionless normalized value, whose value range is [0, 1], used to quantify the number and completeness of the preset key equipment features that can be observed at the same time at the candidate observation point p, the larger the value, the richer the diagnostic information; the preset key equipment features include but are not limited to insulator strings, heat sinks, connecting bolts; Key feature visibility V feature The acquisition approach and value determination method are: reading a set of three-dimensional coordinates of key feature regions from the equipment information library associated with the inspection target. For each candidate observation point p, the line-of-sight occlusion is judged by the three-dimensional geometric model, and the number of visible key feature regions from the candidate observation point p is counted. The calculation method of the key feature visibility V feature is to divide the number of visible key feature regions by the total number of key feature regions to obtain a basic ratio; 2.3) Attitude control margin M control is a dimensionless normalized value, whose value range is [0, 1], used to quantify the remaining percentage of the control output required to resist environmental disturbances when the unmanned aerial vehicle hovers at the candidate observation point p, the larger the value, the more stable and safer the flight; Attitude control margin M control The acquisition approach and value determination method are: querying the voxel comprehensive risk index I riskAccording to the dynamic model of the UAV, there is a preset function relationship to map the voxel comprehensive risk index I risk to an expected control consumption C expected ; the calculation method of the attitude control margin M control is to subtract the expected control consumption C expected from 1, that is, 1-C expected . This calculation method intuitively reflects the "idle resources" of the flight control system at this point. The higher the risk, the lower the margin; 2.4) The candidate point evaluation module constructs an m x 3 decision matrix with the three performance components (V clarity , V feature , M control ) of the m candidate observation points; first, normalize the data in each column of the decision matrix to ensure that all values are on the same scale; For the jth performance component, the candidate point evaluation module calculates its information entropy E j ; its calculation expression is: the information entropy E j is equal to a negative constant k1 multiplied by the sum of the normalized values P ij of all candidate points under this performance component and ln(P ij ); P ij represents the proportion of the value of the ith candidate point under the jth performance component to the sum of the component.

[0048] 2.5) The objective weight W j of the jth performance component is determined by its information entropy E j . Its calculation expression is: the objective weight W j is equal to (1-E j ) divided by the sum of (1-E k ) of all performance components; where E k represents the information entropy of each performance component in turn; the basic principle is: the greater the numerical difference of a performance component among all candidate points, that is, the smaller the information entropy, the more effective information the performance component provides, and it should be given a higher weight.

[0049] 2.6) For each candidate observation point p, the waypoint inspection performance index I eff is equal to the normalized value of the expected image clarity V clarity multiplied by its weight W1, plus the normalized value of the key feature visibility V feature multiplied by its weight W2, plus the normalized value of the attitude control margin M control multiplied by its weight W3. All input performance components and the final index are in a positive proportional relationship.

[0050] Further supplement to the candidate point evaluation module: the information entropy calculation formula is originally derived from the Shannon entropy formula in information theory, which is used to measure the uncertainty of information; in the embodiment, the information entropy is used to measure the dispersion degree of an efficiency component over all candidate points; Objective weight W j The calculation formula is based on the inverse application of information entropy, and the logic is that the smaller the information entropy, the more information the index provides, and the greater the weight should be; In this embodiment, the candidate point evaluation module generates 6 candidate observation points in the inspection task of the main transformer of the substation. The candidate point evaluation module calculates three efficiency components for each candidate observation point, and determines the objective weight by the entropy weight method, and finally obtains the waypoint inspection efficiency index of each candidate observation point.

[0051] Scenario one, regular inspection: when the numerical distribution of each candidate observation point in the three efficiency components is relatively uniform, the weight calculated by the entropy weight method is also relatively balanced, and W1=0.35, W2=0.33, W3=0.32. This indicates that in this task, the clarity, information integrity and flight safety are equally important; Scenario two, fine diagnosis under special weather conditions (such as light fog, insufficient light): if the weather conditions cause the expected image clarity V clarity of all candidate observation points to be generally low and the numerical difference to be small, while the flight stability M control is generally high due to no strong wind and the difference is small, but since the inspection target is to find a small crack, the key feature visibility V feature under different observation angles is large in numerical difference. At this time, the entropy weight method will automatically allocate most of the weight to V feature , W1=0.15, W2=0.70, W3=0.15. The system can automatically identify the core contradiction of the current task, so as to preferentially select the candidate observation point that can maximize the exposure of the key feature, and realize the intelligent and adaptive decision-making.

[0052] The data table of the embodiment is as follows:

[0053] The above table shows a typical calculation example. Through calculation, the information entropy E2 of the key feature visibility V feature is 0.951, which is significantly lower than the information entropy of the other two components (0.988 and 0.991), which indicates that the key feature visibility V feature has the largest degree of differentiation among the 6 candidate observation points. Therefore, the entropy weight method allocates a high weight of 0.806 to it, while the other two weights are relatively small. The final waypoint inspection efficiency index I eff The calculation result shows that the I eff= 0.942 and I of candidate observation point 3 = 0.938 are the two optimal candidates. This is consistent with the intuitive judgment: these two points not only have high key feature visibility V eff = 0.942 and I of candidate observation point 3 = 0.938 are the two optimal candidates. This is consistent with the intuitive judgment: these two points not only have high key feature visibility V feature = 0.942 and I of candidate observation point 3 = 0.938 are the two optimal candidates. This is consistent with the intuitive judgment: these two points not only have high key feature visibility V

[0054] The output value range of all performance components (V clarity ,V feature ,M control ) and the final waypoint inspection performance index I eff are standardized or limited in the interval [0, 1] through formula design; When the output of waypoint inspection performance index I eff tends to 0, it indicates that the comprehensive performance of the candidate observation point is lower, and the expected return of the inspection task is lower and the risk is higher; When the output of waypoint inspection performance index I eff tends to 0, it means that each term of the weighted sum tends to 0. This is because the value of one or more performance components (V clarity ,V feature ,M control ) itself is close to 0. For example, V clarity of a candidate observation point tends to 0 (severe backlight or blur), V feature tends to 0 (key parts are blocked), or M control tends to 0 (in a strong air flow disturbance area, the flight risk is extremely high). Regardless of the weight allocation, the final output value of the waypoint inspection performance index I eff is low. The system will preferentially avoid such waypoints; When the output of waypoint inspection performance index I eff tends to 1, it indicates that the comprehensive performance of the candidate observation point is higher, and the expected return of the inspection task is higher and the risk is lower.

[0055] When the output of waypoint inspection performance index I eff tends to 1, it requires the result of the weighted sum to tend to 1. Since the sum of the weights is 1, it requires the value of all performance components (V clarity ,V feature ,M control ) with significant weights to tend to 1. For example, in a weight allocation scenario of (0.3, 0.4, 0.3), to make the waypoint inspection performance index I eff close to 1, the values of the three components must be close to 1. This indicates that the candidate observation point can achieve clear imaging, complete diagnostic information, and stable flight attitude. This precisely achieves the technical goal of the invention "selecting the optimal observation point".

[0056] macroscopic path planning module: for selecting several nodes with the highest inspection efficiency index from a plurality of candidate observation points under each inspection target, and generating a global optimal flight route connecting the nodes; Further description: the macroscopic path planning module is further used for: using an improved A search algorithm, wherein the path segment connecting any two nodes is defined as a comprehensive transition cost obtained by weighted summation of geometric path length, predicted flight energy consumption and path environment risk integral; The A The heuristic function of the search algorithm is defined as an efficiency-guided heuristic value, which is calculated according to the remaining distance from the current node to the end point and the average inspection efficiency index of all unvisited nodes; Further, the macroscopic path planning module, before executing the A search algorithm, pre-constructs a completely connected graph with the selected several candidate observation points as intermediate nodes and the preset unmanned aerial vehicle take-off point and task end point as the starting node and the terminal node respectively; The calculation process of the macroscopic path planning module is as follows: 3.1) For the path segment of any two nodes in the completely connected graph, the macroscopic path planning module calculates its comprehensive transition cost and records it as C trans ; any two nodes in this embodiment are the starting node and the target node; The geometric path length is recorded as L geom ; the geometric path length represents the straight-line Euclidean distance of any two nodes in three-dimensional space; the unit is meter (m); L geom The acquisition approach is to directly read the three-dimensional coordinates of any two nodes in the fused environment map and calculate it through the space two-point distance formula; 3.2) The predicted flight energy consumption is recorded as E pred ; the predicted flight energy consumption E pred characterizes the energy consumed by the unmanned aerial vehicle to travel from any two nodes at a preset cruising speed; the unit is joule (J); E pred The acquisition approach is to discretize the straight-line path of any two nodes into a series of equidistant evaluation points; For each evaluation point, the thermal updraft gradient and microscale wind shear coefficient of the three-dimensional voxel where it is located are queried from the fused environment map; according to the unmanned aerial vehicle dynamics model, the additional compensation power required to maintain the cruising speed under the environmental disturbance is calculated, and the instantaneous total power at this point is obtained by combining the basic hovering power; the final value of the predicted flight energy consumption E pred is the result of numerical integration of the instantaneous total power of all evaluation points on the entire path. 3.3)The path environmental risk integral is denoted as R path ; the path environmental risk integral R path comprehensively reflects the environmental danger degree of the path traversal space between any two nodes; R path The acquisition approach is to query the voxel comprehensive risk index I of the three-dimensional voxel where all discrete evaluation points on the path are located through the macroscopic path planning module risk ; the path environmental risk integral R path The final value is the arithmetic mean of these voxel comprehensive risk indexes; 3.4)The geometric path length L geom , the predicted flight energy consumption E pre , and the path environmental risk integral R path are normalized respectively in the numerical set of all path segments; the comprehensive transition cost C trans is calculated as follows: the comprehensive transition cost C trans is equal to the normalized geometric path length multiplied by its weight w L , plus the normalized predicted flight energy consumption multiplied by its weight w E , plus the normalized path environmental risk integral multiplied by its weight w R . The weights w L , w E , and w R are preset according to the task priority and the sum of the three is 1. All input components and the comprehensive transition cost are in a positive proportional relationship; the task priority in this embodiment includes “time priority”, “energy consumption priority”, or “safety priority”; The calculation formula of the comprehensive transition cost C trans originally comes from the linear weighting method in multi-objective decision-making; this method fuses multiple indexes of different dimensions into a single evaluation value through normalization and weighting. In this embodiment, the comprehensive transition cost C trans is used to quantify the comprehensive cost of traversing a path segment; The normalized comprehensive transition cost is limited in the interval [0, 1]; When the output of the comprehensive transition cost C trans tends to 0, it indicates that the passing cost of the path segment is lower; the efficiency and safety of the UAV flight are higher; When the comprehensive transition cost C trans tends to 0, each term of the weighted sum tends to 0, that is, the normalized L geom , E pre , and R path are all small. This means that the path segment is short in distance, low in energy consumption, and safe in environment.

[0057] When the comprehensive transition cost C transThe closer the output is to 1, the higher the passing cost of the path segment is, and the lower the efficiency and safety of the UAV flight are; The transition cost C trans The closer to 1, the normalized L geom , E pre , R path The value of at least one or more of them is close to 1. This indicates that the path segment distance is long, or the energy consumption is high, or the environmental risk is large. A The mechanism of the algorithm naturally avoids expanding the path with too high total cost; 3.5) For A For any current node in the search process, the macro-path planning module calculates its efficiency guide heuristic value to the end point; and is recorded as H eff : Let the remaining Euclidean distance be D remain ; D remain is the straight-line Euclidean distance from the current node to the preset end node, with the unit of meters (m); Let the average efficiency of unvisited nodes be I avg,remain ; the average efficiency of unvisited nodes I avg,remain is a dimensionless scalar that reflects the average inspection value of the nodes in the path graph that have not been accessed; I avg,remain The way to obtain it is that the macro-path planning module maintains a set of unvisited nodes. By querying the inspection efficiency index I eff of these nodes that has been calculated in the previous "candidate point evaluation module", the arithmetic mean of these indexes is calculated to obtain the average efficiency of unvisited nodes I avg,remain .

[0058] The calculation method of the efficiency guide heuristic value H eff is: the efficiency guide heuristic value H eff is equal to the remaining Euclidean distance D remain divided by a gain term composed of the average efficiency of unvisited nodes I avg,remain . The expression of the gain term is (1+k I ×I avg,remain ), where k I is a preset "efficiency influence coefficient" used to adjust the guiding strength of the inspection efficiency on the heuristic value; the calculation method of the efficiency guide heuristic value H eff ensures that when there are high-value nodes on the future path, that is, I avg,remain increases, the efficiency guide heuristic value H eff will decrease, so that the A algorithm is more inclined to explore in that direction.

[0059] The efficiency guide heuristic value H effIts range is [0, +∞), and the unit is meters; when the average efficiency of unvisited nodes is I avg,remain Increasing this value indicates a higher overall "value" for the unvisited node region ahead; performance-guided heuristic value H. eff The denominator (1+k) I ×I avg,remain The increase leads to an increase in the performance-guided heuristic value H. eff Decrease; Performance-guided heuristic H eff Decreasing will make A The reduced evaluation value makes the algorithm more inclined to explore this high-value area; this achieves the technical goal of "efficiency guidance", making the path planning "forward-looking" and proactively approaching areas that can bring higher inspection benefits.

[0060] When the average efficiency of unvisited nodes is I avg,remain When it decreases: it indicates that the average value of the area in front is not high; the efficiency-guided heuristic value H eff The denominator decreases, resulting in the performance-guided heuristic value H. eff Relative increase; A relatively large performance-guided heuristic value H eff This reduces the attractiveness of that direction, causing the algorithm to be more inclined to explore other directions or simply rely on the lowest-cost path. This design ensures that paths do not blindly incur huge travel costs to connect to a distant, low-value node.

[0061] Further explanation of the macro-path planning module, with examples of parameter substitution and weight optimization at the application level as follows: The drone needs to start from point S2, pass through three high-value observation points A2, B2, and C2, and finally reach the destination E2. There are two possible macroscopic paths: path 1 (S2→A2→B2→C2→E2) and path 2 (S2→A2→C2→B2→E2). The following weight allocation and scenario optimization are performed: Scenario 1, Time Priority Mode: In this mode, the weight is set to w. L =0.7,w E =0.2,w R =0.1. The system prioritizes the shortest geometric path while also considering energy consumption and safety; Scenario 2: Safety-first mode, severe weather or complex electromagnetic environment: In this mode, the weight is adjusted to w L =0.2,w E =0.2,w R =0.6. The system will avoid path environment risk score R. path High-speed paths, even if it means taking longer detours or consuming more energy.

[0062] Scenario 3, Energy-Saving Priority Mode, Long-Endurance Mission: In this mode, the weight is adjusted to w L =0.2,w E =0.7,w R =0.1. The system will prioritize selecting the predicted flight energy consumption E. pred The lowest path, for example, tends to utilize the path segment of thermally rising airflow to save power consumption; The table below in this embodiment shows that, in "time priority mode", A The algorithm's cost evaluation process for different path segments;

[0063] The total cost of path 1 (S2→A2→B2→C2→E2) is 0.76+0.65+0.88+1.05=3.34; Evaluation of Path 2: (S2→A2→C2→B2→E2): When the algorithm expands from point A2, it compares the paths leading to B2 and C2. Evaluation value. A moving towards B2 With a rating of 250.65, it moves towards C2 (A). The evaluation score is 180.89. This is due to the A in C2. The rating is lower, A The algorithm will prioritize expanding node C2. However, further expansion reveals that the overall transition cost from C2 to B2 is C... trans The value is 0.95, which makes the total cost of path 2 higher than that of path 1. This process demonstrates the intelligence of this macro-path planning module. Even though the direct distance from A2 to C2 is longer, its efficiency guides the heuristic value H. eff Because higher-value nodes exist after point C2, effectively reducing its value, the algorithm will explore this path. However, A In the algorithm, C trans The cumulative mechanism will eventually penalize paths with high actual travel costs; ultimately, the macro-path planning module will use the efficiency-guided heuristic H. eff We weigh the costs against the actual costs and choose the globally optimal path 1 as the globally optimal flight route.

[0064] The navigation mode switching module is used to monitor the RTK positioning quality factor and the environmental electric field strength scalar of the UAV in real time during flight, and calculate a switching confidence index based on these two parameters to characterize the necessity of mode switching; the mode switching includes switching to a vector navigation mode based on local electric field perception. Further description: when the switching confidence index exceeds the preset switching execution threshold, the navigation mode switching module automatically switches the flight control mode of the unmanned aerial vehicle from the path tracking mode based on global coordinates to the vector navigation mode based on local electric field perception; Before performing the switching of the flight control mode, the navigation mode switching module first stores the global coordinates of the unmanned aerial vehicle at the switching time as a safe return point; To realize the automatic switching of the flight control mode, the navigation mode switching module performs the calculation and decision process based on the switching confidence index as follows: 4.1) For the acquisition and processing of the RTK positioning quality factor: the RTK positioning quality factor is a dimensionless score value output by the on-board RTK solving module, which comprehensively reflects the current positioning solving state. The higher the RTK positioning quality factor, the more reliable the global positioning, and the lower the necessity of switching mode. By reading the output of the RTK solving module in real time, the output value is a discrete or continuous value in the interval of 0 to 5, where 5 represents the optimal positioning quality and 0 represents complete failure. The navigation mode switching module converts the RTK positioning quality factor into a positioning failure risk degree; the positioning failure risk degree is inversely proportional to the RTK positioning quality factor; and is normalized to the interval of 0 to 1. Let the RTK positioning quality factor be Q rtk , the positioning failure risk degree is calculated as (5-Q rtk ) / 5; 4.2) For the acquisition and processing of the environmental electric field intensity scalar: the environmental electric field intensity scalar is a physical quantity representing the strength of the electric field at the spatial position of the unmanned aerial vehicle. The higher the environmental electric field intensity scalar, the more reliable and usable the vector navigation mode based on electric field perception, and the higher the necessity of switching mode. The unit is kilovolt per meter (kV / m). The navigation mode switching module obtains the electric field intensity vector from the on-board electric field sensor array and calculates its modulus. The navigation mode switching module converts the environmental electric field intensity scalar into an electric field navigation usability; the electric field navigation usability is proportional to the environmental electric field intensity scalar and is normalized to the interval of 0 to 1. This conversion is realized by a piecewise function: when the environmental electric field intensity scalar is lower than a minimum perception threshold, the electric field navigation usability is 0; when the environmental electric field intensity scalar exceeds a saturated effective threshold, the electric field navigation usability is 1; between the two, the electric field navigation usability increases linearly with the electric field intensity; The minimum perception threshold is initially set to 2 kV / m; the saturated effective threshold is set to 20 kV / m; 4.3) The switching confidence index is equal to the positioning failure risk degree multiplied by its risk weight w risk , plus the electric field navigation usability multiplied by its usability weight w avail ; The risk weight w riskand availability weight w avail : both are dimensionless parameters, and their sum is 1. The two weights are preset according to the emphasis of the inspection task on positioning reliability or near-field operation stability. For example, when flying in an open area, set w risk higher to prioritize positioning accuracy; when approaching high-voltage equipment, set w avail higher to ensure timely switching to more stable electric field navigation.

[0065] 4.4) The navigation mode switching module compares the calculated switching confidence index with a preset switching execution threshold in real time; the switching execution threshold is initially set to 0.7 in this embodiment; When and only when the value of the switching confidence index is greater than the switching execution threshold, the navigation mode switching module immediately triggers a mode switching instruction to transfer flight control from the global path tracking controller to the local electric field perception navigation controller. Switching confidence index: by processing the two input physical quantities (RTK positioning quality factor and environmental electric field intensity scalar) into dimensionless positioning failure risk and electric field navigation availability respectively, and by weighted summation through dimensionless weights, the final switching confidence index is also a dimensionless value.

[0066] Supplementary explanation of the navigation mode switching module: Identify the "positioning failure risk" as ; identify the "environmental electric field intensity scalar" as ; identify the "electric field navigation availability" as ; identify the "switching confidence index" as ; identify the "switching execution threshold" as ; The core innovation of the navigation mode switching module lies in its calculation logic for the switching confidence index , which is jointly constructed by the positioning failure risk and the electric field navigation availability ; The positioning failure risk is derived from the positioning quality assessment provided by the RTK solution system, following the method for judging positioning accuracy reliability in engineering (5-Q rtk ) / 5 ensures that the value of is between [0, 1]; When the positioning failure risk tends to 0, it indicates that the optimal positioning quality represented by Q rtk tends to 5, the global positioning accuracy of the UAV is higher, the positioning function is more stable and reliable, and the necessity of switching to the local navigation mode is lower; When the positioning failure risk The closer Q approaches 1, the better. rtk The closer the positioning quality is to 0, the lower the global positioning accuracy of the UAV, the higher the risk of global path tracking mode, and the higher the need to switch to other navigation modes. Electric field navigation availability This is derived from the relationship between electric field strength and signal perceptibility in physics; a piecewise linear function is used for transformation. A lower limit for electric field strength perception is set. and the upper limit of electric field strength perception The lower limit of electric field strength sensing in this embodiment Values Upper limit of electric field strength sensing Values ; when hour, .

[0067] when hour, .

[0068] when hour, ; Ensure the availability of electric field navigation The range of its value is between [0, 1]. When electric field navigation availability The closer it is to 0, the stronger the scalar value of the ambient electric field intensity. The closer the electric field strength is to 0 or below the lower limit of electric field strength perception, the weaker the ambient electric field strength is, and the less suitable it is as a navigation basis. At this time, the conditions for switching to local electric field navigation mode are less likely to be met. When electric field navigation availability The closer it is to 1, the stronger the scalar value of the ambient electric field strength. The closer the field strength is to or above the upper limit of electric field strength perception, the stronger and more stable the ambient electric field strength is, and the more suitable it is for vector navigation based on electric field perception. At this time, the conditions for switching are more mature.

[0069] Switch Confidence Index The calculation is based on a linear weighted average method, which is a decision fusion algorithm that ensures... The value range is between [0,1], and the location failure risk degree is determined. and electric field navigation availability confidence index for handover All are in a direct proportional relationship.

[0070] When switching confidence index The closer the value is to 0, the more likely the drone is to be in a state where global positioning is reliable but the electric field environment is unsuitable for local navigation; thus, there is a risk of positioning failure. tends to 0 and the electric field navigation availability is high tends to 0, the UAV should maintain the global path tracking mode and no switching is needed; When the switching confidence index tends to 1, it indicates that the UAV is at a high risk of losing global positioning and the electric field environment is more suitable for local navigation; there is a positioning failure risk degree tends to 1 and the electric field navigation availability is high tends to 1, the UAV must urgently switch to the local electric field sensing vector navigation mode to ensure flight safety and continuous execution of the task.

[0071] When Q rtk decreases, the positioning failure risk degree increases, and regardless of how the environmental electric field intensity scalar changes, the switching confidence index will increase. This directly reflects the increased switching demand due to the decline in global positioning capability, ensuring that the system can respond in a timely manner when the positioning risk increases.

[0072] When the environmental electric field intensity scalar increases, the electric field navigation availability increases, and regardless of how Q rtk changes, the switching confidence index will increase. This directly reflects the system's tendency to switch when entering a strong electric field area, thereby taking advantage of the favorable conditions provided by the environment.

[0073] When Q rtk decreases and the environmental electric field intensity scalar increases, both the positioning failure risk degree and the electric field navigation availability increase, which will cause the switching confidence index to rapidly and significantly increase, accelerating mode switching. This design logic can effectively address the challenges of variable substation environments, where positioning signals are easily disturbed and the electromagnetic environment is complex and variable, ensuring that the UAV can decisively and timely complete mode switching when it is most needed, thereby avoiding the risk of flight loss due to positioning failure and improving the safety and success rate of UAV operations in strong electromagnetic environments.

[0074] The values of the risk weight w risk and the availability weight w avail directly affect the system's focus on positioning risk and electric field availability. This allows for optimized configuration based on specific inspection tasks and environmental conditions; example weight settings: Routine inspection mode, emphasizing positioning safety: set wrisk =0.7, w avail =0.3. In this configuration, when the RTK positioning quality drops slightly, even if the environmental electric field intensity scalar is not high, the switching confidence index will quickly rise, prompting the system to prioritize positioning safety. This applies to areas outside the substation, where the positioning signal is relatively stable, and even if the local electric field signal is not strong, the UAV is required to avoid drifting due to positioning quality decline.

[0075] Fine operation mode, emphasizing near-field adaptability: set w risk =0.3, w avail =0.7. In this configuration, only when the electric field intensity reaches a high level, providing stable and reliable local navigation conditions, will the switching confidence index significantly increase, prompting the system to switch. This applies to situations where the UAV needs to pass through complex high-voltage equipment at close range. Even if the RTK signal is completely disabled, as long as the electric field signal can provide navigation, the system can use it for stable control to ensure task completion.

[0076] By dynamically adjusting these two weights, the system can balance between "global positioning reliability" and "local electric field navigation availability" according to the operation requirements, achieving flexible optimization of UAV navigation strategy in complex substation environments.

[0077] The following table shows the values, calculation process, and final switching confidence index under different working conditions, with the regular inspection mode (w risk =0.7, w avail =0.3) as an example for calculation; the switching execution threshold is 0.5.

[0078] Case One (Ideal): Q rtk is high, and the environmental electric field intensity scalar is low, indicating that global positioning is reliable and there is no strong electric field interference, the positioning failure risk degree and electric field navigation availability both tend to 0, the switching confidence index is low, and the system maintains the global mode

[0079] Case Two (RTK Decline): Q rtk drops from 4.5 to 2.5, causing the positioning failure risk degree to rise to 0.5, even if the environmental electric field intensity scalar does not change much, the switching confidence index also rises.

[0080] Case 3 (Electric field enhancement): environmental electric field intensity scalar from 1.0 to 15.0, resulting in electric field navigation availability to 0.72, even if Q rtk changes little, switching confidence index also increases.

[0081] Case 4 (RTK difference / electric field): Q rtk further decreases to 1.0, environmental electric field intensity scalar increases to 8.0, positioning failure risk degree and electric field navigation availability are at medium-high levels, switching confidence index significantly increases and exceeds the threshold value of 0.5, triggering mode switching. This demonstrates that the system can switch in time to ensure safety when global positioning is impaired but local conditions are acceptable.

[0082] Case 5 (RTK failure / electric field strength): Q rtk decreases to 0.5 (close to complete failure), environmental electric field intensity scalar reaches 22.0 (strong electric field area), positioning failure risk degree and electric field navigation availability tend to 1, switching confidence index reaches 0.930, far exceeding the switching execution threshold value. In the scenario where the UAV penetrates the core area of the high-voltage equipment, it is necessary to force the switch to electric field navigation at this time, which reflects the strong adaptability and safety guarantee capability of the invention under extreme conditions.

[0083] In specific embodiments of the invention, in order to adapt to the limited precision of digital processing systems and improve computational efficiency, the system further includes a processing step: comparing the calculated positioning failure risk degree and electric field navigation availability with a preset minimum threshold value (preset to 10 -6 ); if the calculated value of the parameter is less than the minimum threshold value, the value of the parameter is forcibly set to 0 in the subsequent calculation of the switching confidence index. Similarly, if the calculated value is greater than 1-10 -6 , it is forcibly set to 1. This processing method is a common technical means for those skilled in the art when implementing numerical algorithms, aiming to effectively process parameters that theoretically approach the target value infinitely in engineering, thereby ignoring the minimum value or saturation value that has no actual impact on the final result.

[0084] The scheme fuses the two influencing factors into a continuous switching confidence index, making the switching decision "soft switching" and avoiding unstable behavior in the critical region. When the parameters fluctuate around the threshold value, the change of the switching confidence index is continuous, thereby providing a more stable and robust basis for decision-making.

[0085] The switching confidence index not only indicates whether to switch, but also quantifies the "urgency" or "necessity" of switching. High values indicate high urgency, and low values indicate low urgency. This quantification allows the system to make more fine-grained control according to the actual situation, for example, different switching execution thresholds can be set to adjust the sensitivity of switching, or warnings can be given based on the rate of change of the switching confidence index .

[0086] By adjusting the risk weight w risk and the availability weight w avail , the system can flexibly adapt to different task priorities and operating environments. In tasks with extremely high positioning accuracy requirements, the risk weight w risk can be increased to make the system more sensitive to RTK quality degradation; while in high-pressure equipment, areas with severe electromagnetic interference, the availability weight w avail can be increased to make the system more actively use electric field signals for navigation. This configurability greatly enhances the flexibility and practicality of the system.

[0087] Near-field navigation control module: for processing signals from the on-board electric field sensor array after the flight control mode is switched to the vector navigation mode, real-time solving the local space electric field gradient vector of the position of the UAV, and generating control instructions for each actuator of the UAV based on the local space electric field gradient vector.

[0088] Further, the near-field navigation control module is further configured to: construct a two-channel control law, which includes a radial control channel and a tangential control channel; The radial control channel is configured to generate a radial velocity instruction along the direction of the local space electric field gradient vector based on the difference between the currently measured environmental electric field intensity scalar and a preset target electric field intensity, to control the distance of the UAV from the surface of the charged device; The tangential control channel is configured to generate a tangential velocity instruction perpendicular to the direction of the local space electric field gradient vector based on the dot product of the current velocity vector of the UAV and the local space electric field gradient vector, to control the UAV to fly along the equipotential surface; The technical effects are that: complex motion in three-dimensional space is decomposed into two independent channels of "radial" (control distance) and "tangential" (control trajectory); mutual interference and coupling between different control targets are avoided, and the stability and precision of control are significantly improved. Further, the local space electric field gradient vector is obtained by finite difference calculation on the environmental electric field intensity vectors measured by the airborne electric field sensor array at different positions. To realize the vector navigation based on local electric field perception, the near-field navigation control module performs the following calculation and instruction generation process based on the double-channel control law: 5.1) Let the local space electric field gradient vector be E, local space electric field gradient vector E is a three-dimensional vector, whose direction points to the direction of the fastest growth of electric field intensity, i.e. perpendicular to the surface of the charged device, and its magnitude represents the rate of change of electric field intensity in that direction, with the unit of kilovolt per square meter (kV / m²); The local space electric field gradient vector is obtained by real-time calculation using spatial finite difference method through the electric field intensity vectors measured by at least four non-coplanar three-axis electric field sensors installed on the unmanned aerial vehicle; Let the current speed vector of the unmanned aerial vehicle be V, current speed vector of the unmanned aerial vehicle V is the three-dimensional speed of the unmanned aerial vehicle in the body coordinate system, with the unit of meters per second (m / s); the near-field navigation control module obtains the current speed vector from the inertial measurement unit (IMU) and flight controller of the unmanned aerial vehicle in real time; 5.2) Calculation of speed instruction of radial control channel: aims to drive and maintain the unmanned aerial vehicle on an equipotential surface with a constant equivalent distance from the charged device; The radial error is equal to the scalar of environmental electric field intensity minus a target electric field intensity; the target electric field intensity is a constant value preset according to the safety distance required by the inspection task; 5.3) Calculation of radial speed instruction: adopts proportional-integral-derivative control law. The radial speed instruction is equal to the current radial error multiplied by the radial proportional gain, plus the cumulative integral of the radial error with respect to time multiplied by the radial integral gain, plus the rate of change of the radial error with respect to time multiplied by the radial derivative gain; 5.4) Calculation of speed instruction of tangential control channel: aims to eliminate the component of the unmanned aerial vehicle speed perpendicular to the equipotential surface, so that it strictly flies along the equipotential surface; The tangential coupling degree is equal to the dot product of the current speed vector of the unmanned aerial vehicle and the normalized local space electric field gradient vector; 5.5) Calculate Tangential Correction Velocity Command: Adopt PID control law; Tangential Correction Velocity equals Tangential Proportional Gain times current Tangential Coupling Degree, plus Tangential Integral Gain times cumulative integral of Tangential Coupling Degree over time, plus Tangential Differential Gain times rate of change of Tangential Coupling Degree over time; Radial Control Component: Multiply the calculated Radial Velocity Command by the normalized Local Space Electric Field Gradient Vector to obtain a velocity vector that controls the UAV to approach or move away from the charged device; Tangential Control Component: Multiply the calculated Tangential Correction Velocity by a direction vector that is orthogonal to the Local Space Electric Field Gradient Vector to obtain a velocity vector that corrects the UAV's flight trajectory to keep it on the equipotential surface; The final target velocity command sent to the UAV's underlying flight controller is the vector sum of the above-mentioned Radial Control Component, Tangential Control Component, and a pre-set base orbit flight velocity vector.

[0089] In a specific embodiment, the magnitude of the pre-set base orbit flight velocity vector is set equal to the pre-set cruise speed used in the Macroscopic Path Planning module.

[0090] Supplementary explanation of the Near-Field Navigation Control Module: Let the Local Space Electric Field Gradient Vector be denoted as ; let the current UAV velocity vector be denoted as ; let the target electric field intensity be denoted as ; let the radial error be denoted as ; let the radial velocity command be denoted as ; let the tangential coupling degree be denoted as ; let the tangential correction velocity command be denoted as ; let the final resultant velocity command be denoted as ; let the radial proportional gain be denoted as ; let the tangential proportional gain be denoted as ; The design of the radial control channel in this embodiment is derived from the feedback control idea in classical control theory, which indirectly represents and controls the "distance" in the physical world through the measurable electric field intensity; the calculation logic is to calculate the radial error , which is calculated as follows: ; this radial error directly quantifies the "electric field distance" between the current position and the target equipotential surface. The radial error serves as the input of the PID controller to generate the radial velocity command .

[0091] In this embodiment, is a static set value that defines the control target. The environmental electric field intensity scalar is the real-time dynamic input. Radial error is the difference between the two, whose sign and magnitude directly drive the subsequent control behavior, ensuring that the control system moves in the direction of reducing the error; When , it represents the environmental electric field intensity scalar is less than the target electric field intensity , the UAV is too far away from the charged device. The controller will generate a positive radial velocity command , driving the UAV to fly in the direction of the local space electric field gradient vector to increase the environmental electric field intensity scalar , until the radial error tends to zero; When , it represents the environmental electric field intensity scalar is greater than the target electric field intensity , the UAV is too close to the charged device. The controller will generate a negative radial velocity command , driving the UAV to fly in the direction of the local space electric field gradient vector to reduce the environmental electric field intensity scalar , until the radial error tends to zero. Through this closed-loop feedback, the UAV can automatically and accurately lock itself on the equipotential surface defined by the target electric field intensity , achieving constant distance maintenance from the surface of the charged device.

[0092] The design of the "tangential control channel" comes from the orthogonality principle in vector analysis; physically, flying along the equipotential surface means that the flight velocity vector must always be orthogonal to the normal direction of the point (i.e. the electric field gradient direction); "the normal direction represents the electric field gradient direction"; the calculation logic is to calculate the tangential coupling degree , which accurately quantifies the size of the component of the current UAV speed in the direction of "penetrating" the equipotential surface. The target value of the tangential coupling degree is zero. As the input of the PID controller, it generates the tangential correction velocity command . Among them, the current UAV velocity vector and the local space electric field gradient vector are two real-time dynamic vector inputs. Through the dot product operation, the complex vector relationship is projected into a single scalar error representation of the tangential coupling degree to simplify the control problem; When When the UAV is not on the equipotential surface, there is a velocity component deviating from the equipotential surface, i.e. there is a drift. The controller generates a correction velocity command in the tangential direction opposite to the drift direction, which actively offsets the drift component and drives the tangential coupling degree to zero. By continuously maintaining the tangential coupling degree around zero, the "tangential control channel" ensures that the flight trajectory of the UAV is strictly constrained on the equipotential surface, achieving precise orbit flight or super-stable hovering without drift.

[0093] The final synthesized velocity command is the vector sum of the radial velocity component, the tangential correction velocity component, and the preset orbit velocity component. This decoupling and recombination makes the complex spatial motion of the UAV decomposed into three controllable parts: "maintaining distance" by the radial channel, "eliminating drift" by the tangential channel, and "orbit flight" in sequence; ensuring that the system can meet multiple control objectives at the same time at any time, such as maintaining a constant distance from the guide wire while smoothly flying along the guide wire direction.

[0094] The following table shows the parameter calculation and control effect of the UAV in different flight states, taking the target electric field intensity as an example. This example here only shows the P control effect in proportional control, i.e. and ; the negative sign indicates that the correction direction is opposite to the coupling direction. Set , .

[0095]

[0096] States one, two, and three show the independent action of the radial control channel. When the UAV has a distance deviation from the target equipotential surface, the controller generates a radial velocity command pointing to or away from the device until the distance is appropriate , and the command velocity becomes zero; States four and five show the independent action of the tangential control channel. When the UAV is on the correct equipotential surface but has a drift velocity perpendicular to the surface, the controller generates a tangential correction velocity to actively offset the drift.

[0097] The composite error of state six is a scenario that embodies the advantages of the invention; the UAV is both too far away, and has an outward drift, . The dual-channel control law can simultaneously calculate the radial velocity command that needs to fly inward and the tangential velocity command that needs to be corrected inwardAfter the two instruction vectors are synthesized, the UAV is driven to correct the distance and drift simultaneously along an optimal path and finally stabilize on the target orbit.

[0098] In the specific embodiment of the present application, in order to adapt to the limited precision of the digital processing system and improve the calculation efficiency, the system further comprises a processing step: comparing the calculated radial error and tangential coupling degree with a preset zero threshold value; if the absolute value of the parameter is less than the zero threshold value, the value of the parameter is forcibly set to 0 in the subsequent PID controller input. This processing method is a common technical means for those skilled in the art when implementing numerical algorithms, which aims to effectively process the parameter that theoretically approaches the target value infinitely in engineering, so as to ignore the extremely small value or saturation value that has no actual impact on the final result, and avoid unnecessary response of the controller to small noise; the zero threshold value is set to 10 -3 in the embodiment. The beneficial effects of the embodiment are as follows: 1. The traditional method attempts to directly control the UAV in three-dimensional space, which faces complex dynamic coupling and nonlinear problems. The present scheme decomposes the problem into two sub-problems with clear physical meaning and mutual orthogonality, namely distance control and trajectory control, which reduces the complexity and difficulty of the algorithm.

[0099] 2. The vague inspection requirements of "maintaining distance" and "flying along the track" are converted into two precise numerical targets of tracking and driving to zero . This makes the system performance can be quantitatively evaluated and optimized, and also makes the mature PID control algorithm can be directly applied, which guarantees the precision and convergence of the control.

[0100] 3. The control law completely depends on local and robust electric field signals and does not depend on any external positioning source. The control target is to maintain a balanced state of a physical field, rather than to track an abstract coordinate point. This control method based on physical field balance has inherent robustness. Even if the UAV is disturbed by external gusts, the controller will immediately perceive the change of the physical field and quickly compensate to pull the UAV back to a stable state.

[0101] It should be noted that all the calculation formulas in the present application file use regression analysis including but not limited to machine learning algorithms to deeply analyze the collected relevant parameters, identify their natural trends and mutual relationships. Professional software such as Python's Scikit-learn library or R language is used to automatically generate mathematical models matching the data. Then, the performance of the model is objectively evaluated by methods such as cross-validation, and combined with continuous feedback and optimization to ensure that the created formula truly reflects the inherent law of the data, thereby ensuring its effectiveness and accuracy. In all the calculation formulas in the present application, the parameters in each formula are processed by consistent range of dimensionless to ensure that different physical quantities are compared on the same scale; the dimensionless technique includes but is not limited to Min-Max-Normalization, Z-Score standardization; The technical solutions of the present application can be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as a floppy disk, a read-only memory (ROM), a random-access memory (RAM), a FLASH, a hard disk or an optical disk of a computer, etc., including a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the method of each embodiment of the present application.

[0102] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered a list of executable instructions for implementing logic functions, and can be specifically embodied in any computer-readable medium for use by an instruction execution system, apparatus or device, such as a computer-based system, a system including a processor or other system that can fetch and execute instructions from the instruction execution system, apparatus or device. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate or transport a program for use by or in connection with an instruction execution system, apparatus or device, or in conjunction with such an instruction execution system, apparatus or device.

[0103] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it, although the present application 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 application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and all should be covered in the scope of the claims of the present application.

Claims

1. A substation unmanned aerial vehicle three-dimensional route real-time planning and autonomous obstacle avoidance system, characterized in that, Specifically comprising: A model construction module: for obtaining a three-dimensional geometric model, a three-dimensional electromagnetic field intensity distribution map and a three-dimensional thermal field distribution map in a patrol space containing a plurality of patrol targets before generating a three-dimensional route, and then performing spatial registration and data fusion to generate a fusion environment map containing a plurality of physical field attributes; A candidate point evaluation module: for generating a plurality of candidate observation points for each patrol target based on the fusion environment map, and calculating a flight point patrol efficiency index of each candidate observation point; A macroscopic path planning module: for selecting a plurality of nodes with the highest flight point patrol efficiency index from a plurality of candidate observation points subordinate to each patrol target as a path graph, and generating a global optimal flight route connecting the nodes; A navigation mode switching module for monitoring the RTK positioning quality factor and the environmental electric field intensity scalar of the unmanned aerial vehicle in the flight process in real time, and calculating a switching confidence index representing the necessity of mode switching based on the two parameters; the mode switching includes switching to a vector navigation mode based on local electric field perception; A near-field navigation control module: for calculating the local space electric field gradient vector of the position of the unmanned aerial vehicle in real time by processing the signals from the on-board electric field sensor array after the flight control mode is switched to the vector navigation mode, and generating control instructions for each motor of the unmanned aerial vehicle based on the local space electric field gradient vector. 2.The substation unmanned aerial vehicle three-dimensional route real-time planning and autonomous obstacle avoidance system according to claim 1, characterized in that: The fusion environment map is in units of three-dimensional voxels, and each three-dimensional voxel contains geometric, electromagnetic and thermal field attribute information; The three-dimensional geometric model is used to represent the geometric shape of the space obstacles in the patrol space; The three-dimensional electromagnetic field intensity distribution map is used to represent the electromagnetic interference intensity of the patrol space; and the three-dimensional thermal field distribution map is used to represent the device heat dissipation influence of the patrol space; The patrol space is discretized into a plurality of three-dimensional voxels, and a voxel comprehensive risk index representing the comprehensive flight risk of each three-dimensional voxel is calculated; The voxel comprehensive risk index is calculated based on the electromagnetic interference potential and the thermal disturbance intensity in the three-dimensional voxel through a preset risk assessment model. 3.The substation UAV three-dimensional route real-time planning and autonomous obstacle avoidance system according to claim 2, characterized in that: When performing spatial registration, the model construction module takes the three-dimensional geometric model as the reference coordinate system, extracts common feature points in the three-dimensional electromagnetic field intensity distribution map and the three-dimensional thermal field distribution map, and transforms them to the reference coordinate system; For the voxel comprehensive risk index of each three-dimensional voxel, the model construction module performs the following calculation process: The electromagnetic interference potential is used to quantify the potential interference of the electromagnetic environment in the three-dimensional voxel on the navigation and communication system of the unmanned aerial vehicle, and the larger the value, the stronger the interference; The thermal disturbance intensity is used to comprehensively evaluate the comprehensive disturbance effect of the rising air current and micro-scale turbulent flow formed by device heat dissipation in the three-dimensional voxel on the attitude stability of the unmanned aerial vehicle, and the larger the value, the stronger the disturbance; The micro-scale turbulent flow is represented by the micro-scale wind shear coefficient table extracted from the pre-calculated flow field database.

4. The substation UAV three-dimensional route real-time planning and autonomous obstacle avoidance system according to claim 3, characterized in that: The model construction module constructs the risk assessment model by principal component analysis method; and the output voxel comprehensive risk index represents the comprehensive threat degree of the three-dimensional voxel to the safe flight of the unmanned aerial vehicle; The smaller the voxel comprehensive risk index is, the smaller the comprehensive flight risk of the three-dimensional voxel is, and the easier the three-dimensional voxel is to be selected by the planned route; The larger the voxel comprehensive risk index is, the larger the comprehensive flight risk of the three-dimensional voxel is, and the more difficult the three-dimensional voxel is to be selected by the planned route.

5. The substation UAV three-dimensional route real-time planning and autonomous obstacle avoidance system according to claim 4, characterized in that: The candidate point evaluation module includes performance components: for each candidate observation point, the expected image sharpness representing the imaging quality thereof, the key feature visibility representing the diagnostic value thereof, and the attitude control margin representing the flight stability thereof are calculated respectively; the expected image sharpness, the key feature visibility, and the attitude control margin are represented as performance components; The candidate point evaluation module determines the objective weight of each performance component by calculating the dispersion degree of each performance component based on an entropy weight method, and performs weighted summation on the expected image sharpness, the key feature visibility, and the attitude control margin to obtain the inspection point inspection performance index; The candidate point evaluation module generates the plurality of candidate observation points by taking the three-dimensional model of the inspection target as the center, generating a set of initial point sets uniformly distributed in space outside the safe distance range according to a preset spherical coordinate system rule, and taking the initial point sets as the candidate observation points; The expected image sharpness is used to quantify the contrast and detail performance capability when an image is taken at the candidate observation point due to lighting and camera angle, and the larger the value is, the clearer the expected image is; The key feature visibility is used to quantify the number and integrity of the preset key equipment features that can be observed simultaneously at the candidate observation point, and the larger the value is, the richer the diagnostic information is; The attitude control margin is used to quantify the margin of the percentage of the control output consumed by the unmanned aerial vehicle to resist environmental disturbance to the total output capacity when hovering at the candidate observation point, and the larger the value is, the more stable and safer the flight is; When the inspection point inspection performance index output is smaller, the comprehensive performance of the candidate observation point is lower, and the expected benefit of the inspection task is lower and the risk is higher; When the inspection point inspection performance index output is larger, the comprehensive performance of the candidate observation point is higher, and the expected benefit of the inspection task is higher and the risk is lower. 6.The substation UAV three-dimensional route real-time planning and autonomous obstacle avoidance system according to claim 5, characterized in that: The macro-path planning module is further configured to: adopt an improved A A search algorithm, wherein the passage cost of the path segment connecting any two nodes is defined as a comprehensive transition cost obtained by weighted summation of a geometric path length, a predicted flight energy consumption and a path environment risk integral. The A The heuristic function of the search algorithm is defined as a performance-guided heuristic value, which is calculated according to the remaining distance from the current node to the end point and the average waypoint inspection performance index of all unvisited nodes. The macro-path planning module performs the A Before the search algorithm, a fully connected graph is constructed with several selected candidate observation points as intermediate nodes and a preset UAV takeoff point and task end point as the starting node and the terminal node respectively; the geometric path length represents the straight-line Euclidean distance between any two nodes in three-dimensional space; The predicted flight energy consumption represents the energy consumed by the unmanned aerial vehicle to travel from any two nodes at a preset cruising speed; The path environment risk integral comprehensively reflects the environmental danger degree of the path traversal space between any two nodes. 7.The substation UAV 3D route real-time planning and autonomous obstacle avoidance system of claim 6, wherein: When the comprehensive transition cost output is smaller, the passing cost of the path segment is lower, and the efficiency and safety of the unmanned aerial vehicle flight are higher; When the comprehensive transition cost output is larger, the passing cost of the path segment is higher, and the efficiency and safety of the unmanned aerial vehicle flight are lower; For A For any current node in the search process, the macro-path planning module calculates its efficiency guide heuristic value to the end point. 8.The substation UAV 3D route real-time planning and autonomous obstacle avoidance system of claim 7, wherein: When the switching confidence index exceeds a preset switching execution threshold, the navigation mode switching module automatically switches the flight control mode of the unmanned aerial vehicle from the path tracking mode based on the global coordinates to the vector navigation mode based on the local electric field perception; Before performing the switching of the flight control mode, the navigation mode switching module first stores the global coordinates of the unmanned aerial vehicle at the switching time as a safe return point; The RTK positioning quality factor is a dimensionless score value output by the airborne RTK solving module, which comprehensively reflects the current positioning solving state. The higher the RTK positioning quality factor, the more reliable the global positioning, and the lower the necessity of switching modes. The navigation mode switching module converts the RTK positioning quality factor into a positioning failure risk degree; the positioning failure risk degree is inversely proportional to the RTK positioning quality factor. The environmental electric field intensity scalar is a physical quantity representing the strength of the electric field at the spatial position of the UAV. The higher the environmental electric field intensity scalar, the more reliable and usable the vector navigation mode based on electric field perception, and the higher the necessity of switching modes. The navigation mode switching module converts the environmental electric field intensity scalar into an electric field navigation usability; the electric field navigation usability is proportional to the environmental electric field intensity scalar; the switching confidence index is equal to the positioning failure risk degree multiplied by its risk weight, plus the electric field navigation usability multiplied by its usability weight. The navigation mode switching module compares the calculated switching confidence index with a preset switching execution threshold in real time. When and only when the value of the switching confidence index is greater than the switching execution threshold, the navigation mode switching module immediately triggers a mode switching instruction to transfer the flight control right from the global path tracking controller to the local electric field perception navigation controller. The smaller the switching confidence index, the more likely the UAV is in a state where global positioning is reliable and the electric field environment is not suitable for local navigation. The larger the switching confidence index, the more likely the UAV is in a state where global positioning is at high risk and the electric field environment is suitable for local navigation. 9.The substation UAV 3D route real-time planning and autonomous obstacle avoidance system of claim 8, wherein: The near-field navigation control module is further configured to construct a dual-channel control law, which includes a radial control channel and a tangential control channel. The radial control channel is configured to generate a radial velocity instruction along the direction of the local spatial electric field gradient vector based on the difference between the currently measured environmental electric field intensity scalar and a preset target electric field intensity, to control the distance of the UAV from the surface of the charged device. The tangential control channel is configured to generate a tangential velocity instruction perpendicular to the direction of the local spatial electric field gradient vector based on the dot product of the current velocity vector of the UAV and the local spatial electric field gradient vector, to control the UAV to fly along the isopotential surface. 10.The substation UAV 3D route real-time planning and autonomous obstacle avoidance system of claim 9, wherein: The local spatial electric field gradient vector is calculated by finite difference calculation on the environmental electric field intensity vectors measured by the airborne electric field sensor array at different positions.

Images