Transformer substation positioning and navigation method and system combining positioning and visual auxiliary positioning

Abstract

The invention discloses a transformer substation positioning and navigation method and system integrating positioning and visual auxiliary positioning. The method comprises the following steps: establishing a background electromagnetic field distribution baseline graph associated with a three-dimensional reference map; electromagnetic field data of the transformer substation in an emergency response mode is collected through the unmanned aerial vehicle, and the abnormal field intensity of a space electromagnetic field and a fault source pointing vector pointing to a fault source are calculated; constructing a risk adaptive elliptical forbidden zone; calculating a passage cost index Cto of the unmanned aerial vehicle passing through each three-dimensional grid in the three-dimensional reference map by using the fault source pointing vector and the risk adaptive elliptical forbidden zone; and searching a path which is from the current position of the unmanned aerial vehicle to a preset target point, bypasses the risk adaptive elliptical forbidden zone, faces the fault source and has the lowest total passing cost index Cto by adopting a preset path search algorithm. According to the method, the risk self-adaptive elliptical forbidden zone is constructed and is used as the three-dimensional grid forbidden assignment processing, so that the dynamic global path avoidance of the sudden dangerous zone is realized.

Description

Technical Field

[0001] This invention relates to the field of navigation, measurement, positioning and detection technology, and specifically to a substation positioning and navigation method and system that combines positioning and visual-assisted positioning. Background Technology

[0002] As the core hub of the power system, the safe and stable operation of substations is of paramount importance. Traditional operation and maintenance models heavily rely on manual inspections, which suffer from low efficiency, high risk, and strong data subjectivity. Drones, with their flexible and maneuverable flight capabilities and efficient coverage, can significantly improve inspection efficiency. However, high-precision and highly reliable positioning and navigation remain the key technological foundation for autonomous drone inspections. Utilizing drones to replace manual labor for autonomous inspections has become an important direction for industry development.

[0003] Chinese patent application CN104914865A discloses a positioning and navigation system and method for a substation inspection drone. It describes how information obtained from a 3D LiDAR ranging module and a dead reckoning module is output to a map creation module and a pose calculation module, respectively. The map creation module creates a 3D LiDAR landmark map and a 3D LiDAR grid map representing the drone's operating environment and stores them in a map storage module. The pose calculation module uses the map data in the map storage module and the output data from the dead reckoning module to calculate the drone's current position and attitude in real time and inputs the obtained pose data into a path planning module and a motion control module. The path planning module generates drone flight path data and outputs it to the motion control module. The motion control module generates motion control quantities based on the drone's current position and attitude data and the path planning data, which are then executed by the motor control module. However, in practical applications, the path planning in the above-mentioned prior art heavily relies on a topological map constructed based on static 3D LiDAR landmarks, and generates a preset 2D (or simple 3D) flight path based on this map. When applied to autonomous substation inspections by UAVs in complex 3D space, this pre-planned path planning method suffers from several limitations. In emergency scenarios caused by equipment failures, sudden weather changes (such as line swaying due to gusts of wind or floating debris), or unexpected events (such as small component detachment or smoke), the UAV lacks the ability to dynamically replan its path in real-time and globally in 3D space when faced with dynamic or newly added obstacles (such as temporary construction equipment, flocks of birds, or deformed equipment due to malfunction) that appear on the pre-planned 3D flight path and are not covered by a static map. This results in inherent limitations in flight path adaptability, obstacle avoidance safety, and mission success rate, and may even lead to the risk of collisions between the aircraft and critical substation equipment. Furthermore, the positioning benchmarks relied upon by the aforementioned existing technologies—namely, "3D LiDAR landmarks" or similar environmental features storing historical raw 3D LiDAR scan data—face a serious risk of failure in emergency scenarios. For example, smoke, dust, and water mist generated by fires or explosions inside substations, or suspended particulate matter and environmental anomalies such as strong light and electromagnetic interference caused by faults, will lead to a severe mismatch between the 3D lidar scanning data (or visual image data) collected by the UAV in real time and the pre-stored 3D landmark data. This will cause the positioning calculation to fail, making it impossible for the UAV to maintain accurate self-pose estimation in 3D space. The failure of this positioning reference not only prevents the UAV from completing high-precision inspection operations (such as detecting minute cracks and locating temperature anomalies), but also fundamentally causes it to lose the 3D pose information necessary for any path replanning (including obstacle avoidance or return to base). This constitutes a chain of technical defects from "obstructed flight path" to "positioning failure" and then to "mission interruption or even crash," seriously threatening the safety of the UAV and substation equipment. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a substation positioning and navigation method and system that combines positioning and vision-assisted positioning, in view of the above-mentioned problems in the prior art. The present invention aims to achieve dynamic global path avoidance of sudden dangerous areas by constructing a risk-adaptive elliptical restricted area and treating it as a three-dimensional grid restricted area.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A substation positioning and navigation method combining positioning and vision-assisted positioning includes the following steps: S1. On a preset three-dimensional flight path within the substation, a UAV is used to collect and fuse three-dimensional lidar ranging data, dead reckoning data, and three-dimensional visual data to construct a three-dimensional reference map containing geometric features and visual semantic landmarks; background electromagnetic field data at various locations along the three-dimensional flight path is collected by the UAV to establish a background electromagnetic field distribution baseline map associated with the three-dimensional reference map; S2. Electromagnetic field data of the substation in emergency response mode is collected by the UAV and compared with the background electromagnetic field distribution baseline map to calculate the spatial electromagnetic field anomaly intensity characterizing the fault source; based on the spatial electromagnetic field anomaly... S3. Calculate the spatial electromagnetic field anomaly gradient as the fault source pointing vector; S4. Based on the fault source pointing vector and the spatial electromagnetic field anomaly strength, estimate the location of the fault source on the 3D reference map; On the 3D reference map, dynamically construct a risk-adaptive elliptical restricted area based on the spatial electromagnetic field anomaly strength and the fault source pointing vector at the estimated fault source location; S5. Calculate the passage cost index Cto of the UAV passing through each 3D grid in the 3D reference map using the fault source pointing vector and the risk-adaptive elliptical restricted area, and use a preset path search algorithm to search for the path from the UAV's current position to the preset target point that bypasses the risk-adaptive elliptical restricted area and heads towards the fault source with the lowest total passage cost index Cto.

[0006] Optionally, in step S2, when collecting electromagnetic field data of the substation in emergency response mode using a drone and comparing it with the background electromagnetic field distribution baseline map to calculate the abnormal spatial electromagnetic field strength representing the fault source, the process includes subtracting the background electromagnetic field data in the background electromagnetic field distribution baseline map from the electromagnetic field data of the substation in emergency response mode, thereby obtaining the calculated abnormal spatial electromagnetic field strength representing the fault source.

[0007] Optionally, in step S2, calculating the spatial electromagnetic field anomaly gradient based on the spatial electromagnetic field anomaly strength as the fault source pointing vector to the fault source includes: performing differential calculations on the spatial electromagnetic field anomaly strength at different locations to obtain the spatial electromagnetic field anomaly gradient at each location, and selecting the largest spatial electromagnetic field anomaly gradient as the fault source pointing vector to the fault source.

[0008] Optionally, step S3 includes: S3.1, Based on the direction of the fault source pointing vector, find the location of the maximum value of the abnormal field strength of the spatial electromagnetic field on the three-dimensional reference map as the location of the predicted fault source; S3.2, the location of the predicted fault source is taken as the center of the risk adaptive elliptical restricted area; S3.3 maps the spatial electromagnetic field anomaly strength to the risk radius Rr as the semi-minor axis of the risk adaptive elliptical restricted zone, and maps the validity of the fault source pointing vector to the uncertainty eccentricity Eu as the eccentricity of the risk adaptive elliptical restricted zone, thus obtaining a dynamically constructed risk adaptive elliptical restricted zone.

[0009] Optionally, in step S3.3, mapping the abnormal field strength of the spatial electromagnetic field to the risk radius Rr includes: calculating the product of the abnormal field strength of the spatial electromagnetic field and a preset abnormal field strength gain coefficient Kga, and adding the product to the basic safety radius Rb to obtain the risk radius Rr, wherein the basic safety radius Rb is a preset constant value, representing the minimum safe distance that the UAV needs to maintain; the abnormal field strength gain coefficient Kga is a preset constant coefficient used to convert and scale the physical dimensions of the abnormal field strength of the spatial electromagnetic field to the length dimension.

[0010] Optionally, in step S3.3, mapping the validity of the fault source pointing vector to the uncertainty eccentricity Eu includes: calculating the fault vector validity index Vv according to the following formula: ; Where Wa, Ws, and Wc are weighting coefficients, respectively; the abnormal signal strength factor Fa, the pointing stability factor Fs, and the spatial gradient clarity Fc are all dimensionless normalization factors with a value range of [0,1]; and the fault vector effectiveness index Vv has a value range of [0,1]. The abnormal signal strength factor Fa is obtained by normalizing the abnormal field strength of the spatial electromagnetic field through a preset threshold function. This threshold function outputs 0 when the abnormal field strength of the spatial electromagnetic field is lower than the detection lower limit and outputs 1 when the abnormal field strength of the spatial electromagnetic field is higher than the saturation upper limit, thus reflecting the fault signal. Signal-to-noise ratio; the pointing stability factor Fs is the linear mapping result of the cosine value of the angle between the fault source pointing vector at the current moment and the fault source pointing vector at the previous moment. The cosine value of the angle is in the interval [-1,1] and is transformed to the interval [0,1] through linear mapping to reflect the continuity of directional information and anti-interference ability; the spatial gradient clarity Fc is the normalized result of the variance of the spatial electromagnetic field anomaly field strength measured by all sensors in the UAV's sensor array; the difference between 1 and the fault vector effectiveness index Vv is used as the obtained uncertainty eccentricity Eu.

[0011] Optionally, in step S4, the function expression for calculating the passage cost index Cto of the UAV passing through each 3D grid in the 3D reference map using the fault source pointing vector and the risk adaptive elliptical restricted area is as follows: ; in, The navigation strategy fusion index is Cem, the fault vector alignment is Cst, the 3D grid no-go assignment is Cco, and the target point approach potential is Cco. The calculation of the fault vector alignment Cem involves: first, calculating the vector pointing from the current 3D grid to the adjacent 3D grid; then, calculating the dot product of the vector pointing from the current 3D grid to the adjacent 3D grid and the vector pointing to the fault source; finally, subtracting the dot product result from 1 to obtain the fault vector alignment Cem. If the 3D grid is located within a static obstacle or a risk-adaptive elliptical no-go zone, the 3D grid no-go assignment Cst is infinity; otherwise, the 3D grid no-go assignment Cst is zero. The target point approach potential is Cco, which is the Euclidean distance from the center of the current 3D grid to a preset regular target point. A regular target point refers to the next waypoint to be reached in a series of ordered 3D waypoints planned on a 3D reference map according to the preset inspection task and route. These waypoints represent specific 3D spatial locations that the UAV needs to inspect and fly to in non-emergency situations. The navigation strategy fusion index is Cem. Where Vv is the fault vector validity index, Clc is the conventional position confidence of the current 3D grid, and the value range of the conventional position confidence Clc is [0,1], representing the map quality of the current 3D grid in the 3D reference map. The map quality is the weighted sum of the geometric information completeness Cg, positioning attitude stability Sp, semantic landmark clarity Cl, and electromagnetic background stability Se of the current 3D grid. Wherein, geometric information completeness Cg is the ratio of the number of effective point clouds to the number of ideal point clouds within the laser radar scanning range of the current 3D grid, positioning attitude stability Sp is the value obtained by reciprocal and normalization of the trace of the pose covariance matrix output by the odometry of the current 3D grid, semantic landmark clarity Cl is the recognition confidence score of semantic landmarks within the field of view of the current 3D grid, which is a value in the interval [0,1], and electromagnetic background stability Se is the value obtained by reciprocal and normalization of the standard deviation of multiple background electromagnetic field measurement results of the current 3D grid.

[0012] The present invention also provides a substation positioning and navigation system combining positioning and vision-assisted positioning, including a microprocessor and a memory interconnected thereto, wherein the microprocessor is programmed or configured to execute the substation positioning and navigation method combining positioning and vision-assisted positioning.

[0013] The present invention also provides a computer-readable storage medium storing a computer program or instructions that are programmed or configured to execute the substation positioning and navigation method of combined positioning and vision-assisted positioning by a processor.

[0014] The present invention also provides a computer program product, including a computer program or instructions, which are programmed or configured to execute the substation positioning and navigation method of combined positioning and vision-assisted positioning via a processor.

[0015] Compared with existing technologies, the present invention can mainly achieve the following beneficial effects: 1. The present invention calculates the abnormal field strength of the spatial electromagnetic field and the fault source pointing vector in the emergency response mode; and innovatively constructs a risk adaptive elliptical restricted area on the three-dimensional reference map based on the abnormal field strength of the spatial electromagnetic field and the validity index of the synchronously generated fault vector. Finally, in the path planning, the restricted area is treated as a high-cost three-dimensional grid restricted area assignment, so that the final navigation path can avoid sudden dangerous areas outside the map topology in real time, which solves the limitation of traditional static path planning in lacking global dynamic replanning capability when dealing with large sudden obstacles. 2. This invention also quantifies the failure degree of conventional positioning systems by calculating the reliability of the conventional position in real time; and introduces a navigation strategy fusion index α determined by the reliability of the conventional position and the effectiveness index of the fault vector. This index can dynamically adjust the weights of the target point approach potential value and the alignment degree of the fault vector in the three-dimensional grid passage cost index. When conventional positioning fails due to factors such as smoke, the navigation reference is smoothly switched from the failed SLAM positioning to an independent fault source pointing vector. Thus, even when the UAV loses its own accurate pose information, it can still perform effective obstacle avoidance and source finding tasks, fundamentally solving the chain technical defects of path planning and task execution caused by the failure of the positioning reference. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the basic process of the method in an embodiment of the present invention.

[0017] Figure 2 This is a basic flowchart of step S3 in the method of this embodiment of the invention. Detailed Implementation

[0018] To enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions of the present invention will be further described in detail below with reference to the accompanying drawings in the embodiments of the present invention.

[0019] like Figure 1As shown, the substation positioning and navigation method combining positioning and visual-assisted positioning in this embodiment includes the following steps: S1. On a preset three-dimensional flight path within the substation, a UAV is used to collect and fuse three-dimensional lidar ranging data, dead reckoning data, and three-dimensional visual data to construct a three-dimensional reference map containing geometric features and visual semantic landmarks; background electromagnetic field data at each location point on the three-dimensional flight path is collected by the UAV to establish a background electromagnetic field distribution baseline map associated with the three-dimensional reference map; S2. Electromagnetic field data of the substation in emergency response mode is collected by the UAV and compared with the background electromagnetic field distribution baseline map to calculate the spatial electromagnetic field anomaly intensity characterizing the fault source; based on the spatial electromagnetic field anomaly... S3. Calculate the spatial electromagnetic field anomaly gradient of the constant field strength as the fault source pointing vector; S4. Estimate the location of the fault source on the 3D reference map based on the fault source pointing vector and the spatial electromagnetic field anomaly strength; On the 3D reference map, dynamically construct a risk-adaptive elliptical restricted area based on the spatial electromagnetic field anomaly strength and the fault source pointing vector at the estimated fault source location; S5. Calculate the passage cost index Cto of the UAV passing through each 3D grid in the 3D reference map using the fault source pointing vector and the risk-adaptive elliptical restricted area, and use a preset path search algorithm to search for the path from the UAV's current position to the preset target point that bypasses the risk-adaptive elliptical restricted area and heads towards the fault source with the lowest total passage cost index Cto.

[0020] In step S1 of this embodiment, the method of constructing a 3D reference map containing geometric features and visual semantic landmarks by collecting and fusing 3D lidar ranging data, dead reckoning data, and 3D visual data on a pre-set 3D flight path within the substation using a UAV is a known existing method, and therefore its implementation details will not be described here. Similarly, the method of establishing a background electromagnetic field distribution baseline map associated with the 3D reference map by collecting background electromagnetic field data at various locations along the 3D flight path using a UAV is also a known existing method, and therefore its implementation details will not be described here.

[0021] In the concepts of spatial electromagnetic field anomaly gradient and spatial electromagnetic field anomaly strength, "space" refers to a local three-dimensional region centered on the UAV and effectively perceived by its sensors. The spatial electromagnetic field anomaly strength is the abnormal field strength value measured by the UAV at a point within its local perception space. The spatial electromagnetic field anomaly gradient is a vector calculated by the UAV within its local perception space, pointing towards the direction of the fastest field strength increase, obtained through multi-point measurements. In step S2 of this embodiment, when collecting electromagnetic field data of the substation in emergency response mode using the UAV and comparing it with the background electromagnetic field distribution baseline map to calculate the spatial electromagnetic field anomaly strength representing the fault source, this includes subtracting the background electromagnetic field data from the background electromagnetic field distribution baseline map from the electromagnetic field data of the substation in emergency response mode, thereby obtaining the calculated spatial electromagnetic field anomaly strength representing the fault source. In step S2 of this embodiment, the calculation of the spatial electromagnetic field anomaly gradient based on the spatial electromagnetic field anomaly strength as the fault source pointing vector to the fault source includes: performing differential calculation on the spatial electromagnetic field anomaly strength at different locations to obtain the spatial electromagnetic field anomaly gradient at each location, and selecting the largest spatial electromagnetic field anomaly gradient as the fault source pointing vector to the fault source.

[0022] Specifically, in step S2 of this embodiment, an abnormal signal strength factor is obtained based on the abnormal field strength of the spatial electromagnetic field; a pointing stability factor and spatial gradient clarity are obtained based on the fault source pointing vector; and a fault vector validity index is generated by fusing the abnormal signal strength factor, pointing stability factor, and spatial gradient clarity, which serves as the reliability of the navigation reference. The specific implementation calculation process for generating the above fault vector and its validity index is broken down as follows: S2.1: At the current location of the UAV, obtain the real-time field strength measurement value and query the background field strength baseline value of the corresponding location from the background electromagnetic field distribution baseline map constructed in S1; the abnormal field strength of the spatial electromagnetic field is equal to the real-time field strength measurement value minus the background field strength baseline value; the essence of this calculation is background subtraction, the initial source of which is the differential amplification principle in signal processing, which aims to extract the incremental signal caused by the fault; S2.2: Using multiple spatial electromagnetic field strength values ​​measured at different points in the local coordinate system of the UAV by the sensor array, the direction of the fastest increase in field strength value is calculated based on the idea of ​​finite difference method in numerical analysis. This direction vector is determined as the fault source pointing vector. S2.3: To calculate the validity index of the fault vector, multiple physical phenomena affecting its reliability must first be quantified to obtain three dimensionless numerical quantization factors, all of which are mapped to the closed interval [0,1]. The abnormal signal strength factor is calculated by normalizing the abnormal field strength of the spatial electromagnetic field calculated in S2.1 through a preset threshold function. This function outputs 0 when the field strength is below the detection lower limit and 1 when the field strength is above the saturation upper limit, reflecting the signal-to-noise ratio of the fault signal. The pointing stability factor is calculated by calculating the cosine of the angle between the pointing vector of the fault source at the current moment and the pointing vector at the previous moment. This cosine value is itself in the interval [-1, 1], and is transformed to the interval [0,1] through linear mapping, reflecting the continuity of directional information and anti-interference ability. The spatial gradient clarity is calculated by calculating the variance of the abnormal field strength of the spatial electromagnetic field measured by all sensors in the sensor array and normalizing the variance. The larger the variance, the more uneven the spatial distribution of the field strength and the clearer the gradient direction. This factor reflects the steepness of the gradient peak, used to distinguish the true gradient from flat noise; S2.4: The determination of contribution weights and the fusion calculation of the effectiveness index are based on the above three factors, and the weights are determined using the idea of ​​principal component analysis. In a large amount of offline simulation or actual fault data, the contribution of the three factors to the result of "whether the fault source can be successfully found" is analyzed. Generally, signal strength is fundamental and has the highest contribution weight; directional stability is key to navigation and has the next highest weight; spatial gradient clarity is used as an auxiliary criterion and has a relatively low weight. For example, an initial set of weights can be set: the contribution weight of the abnormal signal strength factor is 0.5, the contribution weight of the directional stability factor is 0.3, and the contribution weight of the spatial gradient clarity is 0.2; the sum of all contribution weights is always 1. The calculation formula for the fault vector effectiveness index is as follows: the fault vector effectiveness index equals the product of the "abnormal signal strength factor" and its contribution weight, plus the product of the "directional stability factor" and its contribution weight, plus the product of the "spatial gradient clarity" and its contribution weight.

[0023] In this embodiment, the operability and logical deduction of step S2 are as follows: Set a set of instantaneous data collected by the UAV as follows: Parameter substitution: After the quantization calculations in S2.2 and S2.3, the abnormal signal strength factor Fa=0.8, the pointing stability factor Fs=0.9, and the spatial gradient sharpness Fc=0.7 are obtained; The weights set in S2.4 are adopted, where Wa is the weight of the abnormal signal strength factor Fa and Wa=0.5; Ws is the weight of the pointing stability factor Fs and Ws=0.3; Wc is the weight of the spatial gradient sharpness Fc and Wc=0.2.

[0024] Data Calculation: Calculate the fault vector effectiveness index Vv according to the formula in S2.4. ; The results show that the calculated effectiveness index of the fault source pointing vector, Vv, is 0.81. This Vv value will be output to S3 and S4 along with the pointing vector as an important basis for decision-making.

[0025] In this embodiment, the physical dimensions and range of step S2 are explained as follows: Dimensional consistency: The abnormal signal intensity factor Fa, the direction stability factor Fs, and the spatial gradient clarity Fc are all dimensionless normalization factors with a range of [0,1]; the weights Wa, Ws, and Wc are also dimensionless; therefore, the fault vector effectiveness index Vv, as their weighted sum, is also a dimensionless pure number, and its calculation is both physically and mathematically self-consistent; Output range: The output range of the fault vector effectiveness index Vv is limited to the closed interval [0,1]. Boundary value analysis: When the fault vector validity index Vv is 0, it means that the current abnormal signal strength factor Fa, the pointing stability factor Fs, and the spatial gradient clarity Fc are all 0; this means that the signal is weak, the direction is random, and the spatial field is flat, so the calculated pointing vector is completely unreliable; when the fault vector validity index Vv is 1, it means that the current abnormal signal strength factor Fa, the pointing stability factor Fs, and the spatial gradient clarity Fc are all 1, which means that there is a theoretically perfect pointing signal, indicating that the signal strength is saturated, the direction is absolutely stable, and the gradient is extremely clear; The technical effects of step S2 are related to the invention's objectives as follows: When the fault vector validity index Vv output is closer to 1, the reliability of the calculated fault source pointing vector as a navigation reference is higher; when the signal strength factor Fa is close to 1, it means the fault signal characteristics are more obvious; when the pointing stability factor Fs is close to 1, it means that random interference is gradually eliminated and the direction is increasingly stable in pointing to the same source; when the spatial gradient clarity Fc is close to 1, it means that multipath effect interference is less and the gradient direction is more unique and clear; Application status: The system determines that the currently acquired navigation direction is of high quality and adopts this pointing vector with high weight, decisively planning a path in that direction; When the fault vector validity index Vv output is closer to 0, it indicates that the reliability of the calculated fault source pointing vector is lower, and it is noise or artifact; This situation is due to anomalies. This is caused by one or more of the following factors: signal strength factor Fa approaches 0, pointing stability factor Fs approaches 0, and spatial gradient clarity Fc approaches 0. Specifically, an abnormal signal strength factor Fa approaching 0 indicates the signal is gradually submerged in background noise; pointing stability factor Fs approaching 0 indicates the direction swings more violently in a short period; and spatial gradient clarity Fc approaching 0 indicates the electromagnetic field is more uniformly distributed in local space, and the gradient is less obvious. Application status: The system determines that the currently acquired navigation direction is of low quality and unreliable; the weight of this vector is reduced, or even completely ignored when the fault vector validity index Vv is below a preset threshold, to avoid the UAV "chasing noise" and may trigger the UAV to execute strategies such as "stationary detection" or "small-area scanning" to wait for the fault vector validity index Vv to rise to a reliable level.

[0026] On a three-dimensional baseline map, based on the fault source pointing vector, the abnormal field strength of the spatial electromagnetic field, and the fault vector effectiveness index, the size of the dynamically constructed risk-adaptive elliptical restricted area is determined by a risk radius, while the shape of the risk-adaptive elliptical restricted area is determined by an uncertainty eccentricity. The risk radius is positively correlated with the amplitude of the abnormal field strength of the spatial electromagnetic field; the uncertainty eccentricity is negatively correlated with the fault vector effectiveness index. Specifically, the risk radius is calculated by adding a preset basic safety radius representing the minimum safe distance to the product of the abnormal field strength of the spatial electromagnetic field and a preset abnormal field strength gain coefficient. The uncertainty eccentricity is calculated by subtracting the fault vector effectiveness index from the value of 1. Specifically, as follows... Figure 2 As shown, step S3 in this embodiment includes: S3.1, Based on the direction of the fault source pointing vector, find the location of the maximum value of the abnormal field strength of the spatial electromagnetic field on the three-dimensional reference map as the location of the predicted fault source; S3.2, the location of the predicted fault source is taken as the center of the risk adaptive elliptical restricted area; S3.3 maps the spatial electromagnetic field anomaly strength to the risk radius Rr as the semi-minor axis of the risk adaptive elliptical restricted zone, and maps the validity of the fault source pointing vector to the uncertainty eccentricity Eu as the eccentricity of the risk adaptive elliptical restricted zone, thus obtaining a dynamically constructed risk adaptive elliptical restricted zone.

[0027] In step S3.3 of this embodiment, mapping the abnormal field strength of the spatial electromagnetic field to the risk radius Rr includes: calculating the product of the abnormal field strength of the spatial electromagnetic field and a preset abnormal field strength gain coefficient Kga, and adding the product to the basic safety radius Rb to obtain the risk radius Rr, wherein the basic safety radius Rb is a preset constant value, representing the minimum safe distance that the UAV needs to maintain; the abnormal field strength gain coefficient Kga is a preset constant coefficient used to convert and scale the physical dimensions of the abnormal field strength of the spatial electromagnetic field to the length dimension.

[0028] In step S3.3 of this embodiment, mapping the validity of the fault source pointing vector to the uncertainty eccentricity Eu includes: calculating the fault vector validity index Vv according to the following formula: ; Where Wa, Ws, and Wc are weighting coefficients, respectively; the abnormal signal strength factor Fa, the pointing stability factor Fs, and the spatial gradient clarity Fc are all dimensionless normalization factors with a value range of [0,1]; and the fault vector effectiveness index Vv has a value range of [0,1]. The abnormal signal strength factor Fa is obtained by normalizing the abnormal field strength of the spatial electromagnetic field through a preset threshold function. This threshold function outputs 0 when the abnormal field strength of the spatial electromagnetic field is lower than the detection lower limit and outputs 1 when the abnormal field strength of the spatial electromagnetic field is higher than the saturation upper limit, thus reflecting the fault signal. Signal-to-noise ratio; the pointing stability factor Fs is the linear mapping result of the cosine value of the angle between the fault source pointing vector at the current moment and the fault source pointing vector at the previous moment. The cosine value of the angle is in the interval [-1,1] and is transformed to the interval [0,1] through linear mapping to reflect the continuity of directional information and anti-interference ability; the spatial gradient clarity Fc is the normalized result of the variance of the spatial electromagnetic field anomaly field strength measured by all sensors in the UAV's sensor array; the difference between 1 and the fault vector effectiveness index Vv is used as the obtained uncertainty eccentricity Eu. Among them, the basic safety radius Rb is a preset constant value, representing the minimum safe distance that the UAV needs to maintain, and its value is determined by the substation safety regulations; the abnormal field strength gain coefficient Kga is a preset constant coefficient used to convert and scale the physical dimensions of the abnormal field strength of the spatial electromagnetic field to the length dimension; the abnormal field strength Ea of the spatial electromagnetic field is directly taken from the calculation result of S2, and its initial source is a linear mapping based on the physical safety model, that is, it is assumed that the risk level is linearly positively correlated with the measurable abnormal field strength; the calculation of uncertainty eccentricity is a key parameter that determines the shape of the restricted area. The value reflects the system's confidence in locating the fault source direction. When calculating the uncertainty eccentricity, the fault vector effectiveness index is directly taken from the calculation result of step S2, and its value range is [0,1]. The initial source of this calculation is the combination of the definition of elliptical eccentricity in geometry and the innovation point of this application. When the uncertainty eccentricity is 0, it is a circle, and when it approaches 1, it is an extremely elongated ellipse. Through the function of "1 minus effectiveness index", the uncertainty eccentricity approaching 1 is mapped to low eccentricity, and the eccentricity approaching 0 is mapped to high eccentricity. The logic is clear and the calculation is simple.

[0029] Furthermore, the technical effect and logical rationality of step S3 are explained as follows: Taking the detection of a fault signal by a UAV as an example, the algorithmic operability and logical deduction of S3 are used to prove that the restricted area construction method proposed in this application has complete operability. The specific process is as follows: Parameter substitution: In this embodiment, the abnormal field strength of the output spatial electromagnetic field of S2 is set to Ea=100, and the fault vector effectiveness index is set to Vv=0.9; the preset constant basic safety radius is Rb=2 meters, and the abnormal field strength gain coefficient is Kga=0.05 meters / unit; Data calculation: The risk radius is calculated as Rr=2+(0.05×100)=7 meters; the uncertainty eccentricity is calculated as Eu=1-0.9=0.1; The result shows that the system will generate an elliptical restricted area at the estimated fault source location, with a semi-minor axis of 7 meters and an eccentricity of only A value of 0.1 indicates a shape close to a circle; its major axis is along the direction of the fault source vector; this restricted area is large in size but regular in shape, further illustrating that this restricted area has high risk but high directional certainty; in step S3, the physical dimensions and range of the parameters are explained as follows: Dimensional consistency: in the calculation of the risk radius Rr, the basic safety radius Rb is added to the result of multiplying the abnormal field strength gain coefficient Kga by the abnormal field strength Ea of the spatial electromagnetic field, and the final unit is meters, which conforms to the dimension of length; in the calculation of the uncertainty eccentricity Eu, 1 is subtracted from the dimensionless fault vector effectiveness index Vv, and the result is a dimensionless pure number, which conforms to the definition of eccentricity and is logically consistent; since the value range of the fault vector effectiveness index Vv is [0,1], the output value range of the uncertainty eccentricity Eu is also limited to the closed interval [0,1]. The case where the uncertainty eccentricity Eu is 0 occurs when the fault vector validity index Vv is 1, representing a theoretically perfect directional signal. The case where the uncertainty eccentricity Eu is 1 occurs when the fault vector validity index Vv is 0, representing an unreliable directional signal. In step S3, the change in the range of uncertainty eccentricity Eu has a clear and intelligent safety strategy significance, as follows: When the uncertainty eccentricity Eu output approaches 0, the shape of the risk area approaches a standard circle. This occurs when the fault vector validity index Vv approaches 1, that is, the abnormal signal strength factor Fa, the directional stability factor Fs, and the spatial gradient clarity Fc in S2 all approach 1. This indicates that the system's confidence in the directional positioning of the fault source meets the current requirements. At this time, the risk is considered isotropic, spreading uniformly from the estimated source point to all sides. Therefore, generating a circular restricted area with the risk radius Rr as the radius is the most reasonable safety strategy. It maximizes the passable area outside the restricted area while ensuring safety, avoiding unnecessary path detours.As the uncertainty eccentricity Eu output approaches 1, the shape of the risk area tends to be an extremely elongated ellipse along the pointing vector direction. This occurs when the fault vector effectiveness index Vv approaches 0, meaning that at least one of the abnormal signal strength factor Fa, pointing stability factor Fs, or spatial gradient clarity Fc in S2 approaches 0. This indicates that the system has extremely low confidence in directional localization of the fault source. At this point, although there is a "guessed" direction, the system "knows" that this guess is inaccurate, and the fault source is located further or closer in that direction, or even deviates in the direction. Therefore, generating a narrow elliptical restricted area becomes a conservative and intelligent strategy. It maintains a small restricted area width in the direction perpendicular to the pointing vector, which is determined by the risk radius Rr, but extends significantly in the pointing vector direction to cover all areas that may be at risk due to directional uncertainty, ensuring that the safety of the UAV is still guaranteed with the highest priority even in the most uncertain situation.

[0030] In step S4 of this embodiment, the function expression for calculating the passage cost index Cto of the UAV passing through each three-dimensional grid in the three-dimensional reference map using the fault source pointing vector and the risk adaptive elliptical restricted area is as follows: ; in, Cem is the navigation strategy fusion index, Cst is the fault vector alignment, Cco is the 3D grid no-go assignment, and Cco is the target point approach potential. The calculation of fault vector alignment Cem includes: first, calculating the vector pointing from the current 3D grid to the adjacent 3D grid; then, calculating the dot product of the vector pointing from the current 3D grid to the adjacent 3D grid and the vector pointing from the fault source; finally, subtracting the dot product result from 1 to obtain the fault vector alignment Cem. If the 3D grid is located within a static obstacle or a risk adaptive elliptical no-go zone, the 3D grid no-go assignment Cst is infinity; otherwise, the 3D grid no-go assignment Cst is zero. The target point approach potential Cco is the Euclidean distance from the center of the current 3D grid to the preset regular target point. The regular target point refers to the next waypoint to be reached in a series of ordered 3D waypoints planned on the 3D reference map according to the preset inspection task and route. These waypoints to be reached represent specific 3D spatial locations that the UAV needs to check and fly to in sequence in non-emergency situations. Through the above steps, a hybrid navigation cost function is constructed and solved to plan the final navigation path. The hybrid navigation cost function introduces a navigation strategy fusion index to dynamically balance the emergency navigation target guided by the fault source pointing vector with the regular navigation target guided by the preset target point. Furthermore, the value of the navigation strategy fusion index is correlated with the reliability of the conventional location and the effectiveness index of the fault vector, thereby enabling the planning of the final navigation path to achieve a smooth transition from regular navigation to emergency navigation at different reliability levels and to avoid the dynamic three-dimensional virtual restricted area.

[0031] The specific steps of constructing and solving a hybrid navigation cost function include: applying the hybrid navigation cost function to a three-dimensional rasterized cost map; using a path planning algorithm to search for a path with the lowest total cost on the cost map to generate the final navigation path; setting a conventional positioning failure threshold and comparing it with the real-time acquired conventional position reliability; and the planning and output of the final navigation path are performed under the following evaluation conditions: outputting the final navigation path as a series of three-dimensional waypoints for execution by the underlying motion control module; setting a conventional positioning failure threshold and comparing it with the real-time acquired conventional position reliability; and the planning and output of the final navigation path are performed under the following evaluation conditions: when the conventional position reliability is lower than the conventional positioning failure threshold, it is determined as a conventional positioning failure; in this case, by adjusting the navigation strategy fusion index, the weights of the fault vector alignment and the target point approach potential in the hybrid navigation cost function are changed, thereby driving the path planning algorithm to search for and output the final navigation path that bypasses the dynamic three-dimensional virtual restricted area and heads towards the fault source on the hybrid navigation cost map. Furthermore, by combining a hybrid navigation cost function and a dynamically adjusted navigation strategy fusion index, safe and efficient fault source search is achieved. Specifically, the risk-adaptive ellipsoidal restricted area is assigned an extremely high cost in the 3D cost map, serving as a dynamic danger zone that the UAV must avoid. Its size and shape are adjusted in real time according to the severity of the fault and the uncertainty of positioning, ensuring flight safety. Simultaneously, the fault source pointing vector is incorporated into the cost function as a core direction preference, guiding the path planning algorithm to prioritize directions towards the fault source. When conventional positioning fails, the navigation strategy fusion index dynamically increases the weight of the fault vector alignment, reducing dependence on conventional target points, thereby driving the UAV to avoid restricted areas and resolutely move towards the fault source. In this process, the destination point is a dynamic fault source detection target with the fault source as its adaptive center. From initial directional guidance, it gradually converges to the estimated approximate location of the fault source, ultimately guiding the UAV to the precise fault point or the nearest point within the safe detection range to complete fault location and diagnosis. To achieve the above-mentioned path planning based on the hybrid navigation cost function, the specific implementation calculation process is broken down as follows: S4.1 Construct and solve a hybrid navigation cost function to plan the final navigation path on the global cost map; label the navigation strategy fusion index as α, and by introducing the navigation strategy fusion index α, dynamically weight and fuse the "target point gravitational field" representing the regular task and the "fault source vector field" representing the emergency response, thereby generating a unified, instantaneous hybrid navigation potential field; wherein, the value of the navigation strategy fusion index α is jointly determined by the conventional fixed position confidence and the fault vector validity index output by S2, which enables the path planning to achieve a seamless and smooth transition from regular inspection to emergency source tracing under different system states, and ensures that the path avoids the dynamic three-dimensional virtual restricted area generated by S3.The path planning described in S4.2 employs the A-Star search algorithm based on a 3D grid map. On the cost map formed by the hybrid navigation potential field, it searches for the path with the lowest total cost from the UAV's current position to the preset target point. The final navigation path is output as a series of 3D waypoints, which are then tracked and executed by the underlying motion control module.

[0032] The conventional position confidence (Clc) is calculated in real time by acquiring the UAV pose covariance matrix output by the conventional SLAM positioning module. This matrix represents the uncertainty of pose estimation. The conventional position confidence is calculated as follows: the conventional position confidence equals 1 minus the trace of the normalized pose covariance matrix. The smaller the trace of the pose covariance matrix, the smaller the positioning uncertainty of the UAV in the direction, and the more accurate the UAV's coordinate position and orientation positioning. By setting a historical maximum trace value or a theoretical maximum trace value, the current trace value is normalized to the [0,1] interval. The initial source of this calculation is Kalman filter theory, where the pose covariance matrix is ​​the core measure of the state estimation quality. The navigation strategy fusion index α is the core factor determining the navigation strategy bias. The value can be selected according to actual needs. As an optional implementation, this embodiment further incorporates information from the current 3D grid in the 3D reference map to calculate the navigation strategy fusion index. To improve the accuracy, reliability, and interpretability of navigation strategy fusion. In this embodiment, the navigation strategy fusion index α is equal to the fault vector effectiveness index Vv multiplied by (1 minus the conventionally defined position confidence level Clc), that is: , Wherein, Vv is the fault vector validity index, and Clc is the conventional position reliability of the current 3D grid. The conventional position reliability Clc is calculated from the data collected in step S1, with a value range of [0,1]. It represents the map quality of the current 3D grid in the 3D reference map. Its design logic is: emergency navigation is strongly favored only when conventional navigation fails and emergency signals are reliable. Wherein, conventional navigation failure indicates low conventional position reliability, and reliable emergency signals indicate high fault vector validity index.

[0033] The hybrid navigation cost map is constructed on a 3D rasterized 3D baseline map. A 3D raster passage cost index Cto is calculated for each 3D raster. The Cto index for each 3D raster is composed of three superimposed parts: 3D raster no-entry assignment Cst: if the 3D raster is located within a static obstacle or a dynamically generated 3D virtual no-entry zone by S3, its cost is infinite; otherwise, it is zero; Target point approach potential value Cco: equal to the Euclidean distance from the center of the 3D raster to a preset conventional target point. In the context of routine substation inspections by UAVs, the conventional target point refers to the target point determined based on the preset inspection task and route... A line is the next waypoint to be reached in an ordered sequence of 3D waypoints planned on a 3D baseline map. These waypoints represent specific 3D spatial locations that the UAV needs to inspect and fly to in non-emergency situations. They are task-driven, pre-set, and dynamically updated to ensure that the UAV can complete daily inspection tasks as planned, including flying from checkpoint A to checkpoint B, and then to checkpoint C. Points B and C are current or future regular target points. Fault vector alignment Cem: equal to 1 minus the dot product of the vector pointing from the current 3D grid to the adjacent 3D grid and the fault source pointing vector. This cost encourages the alignment of the path direction with the fault source pointing vector. The formula for calculating the 3D grid passage cost index Cto for each 3D grid is as follows: the 3D grid passage cost index Cto equals the navigation strategy fusion index α multiplied by the fault vector alignment Cem plus the 3D grid no-entry assignment Cst, plus 1 minus the navigation strategy fusion index α, and finally multiplied by the target point approach potential value Cco. The initial source for the calculation of the 3D grid passage cost index Cto is the cost function design in optimal control theory. By weighted summation, the multi-objective optimization problem is transformed into a single-objective optimization problem. Finally, the A-Star search algorithm is called to search for the path with the lowest total cost from the current position to the regular target point on this cost map.

[0034] In this embodiment, map quality is the weighted sum of the geometric information completeness Cg, positioning attitude stability Sp, semantic landmark clarity Cl, and electromagnetic background stability Se of the current 3D grid. Geometric information completeness Cg is the ratio of the number of effective point clouds to the number of ideal point clouds within the lidar scanning range of the current 3D grid. Positioning attitude stability Sp is the value obtained by reciprocal and normalization of the trace of the pose covariance matrix output by the odometry of the current 3D grid. Semantic landmark clarity Cl is the recognition confidence score of semantic landmarks within the field of view of the current 3D grid, which is a value in the interval [0,1]. Electromagnetic background stability Se is the value obtained by reciprocal and normalization of the standard deviation of multiple background electromagnetic field measurements of the current 3D grid. This embodiment generates a map quality index for each location point (3D grid) on a 3D reference map. The map quality index quantifies the overall confidence and information completeness of the map data at that location point. The map quality index fusion is obtained by fusing multiple single-source map quality indices. Specifically, the single-source map quality indices include geometric information completeness determined by 3D LiDAR data, positioning attitude stability determined by inertial measurement data, semantic landmark clarity determined by visual sensor data, and electromagnetic background stability determined by electromagnetic field sensor data. The generation of the map quality index specifically includes the following steps: quantizing each single-source map quality index into a dimensionless value within a preset value range; pre-setting a contribution weight based on the dimensionless value within the preset value range; multiplying each dimensionless value by its corresponding contribution weight, and then summing all products to obtain the final map quality index. To achieve the generation of the above map quality indicators, the specific calculation process is broken down as follows: S1.1: Data Acquisition and Preprocessing. While the UAV moves at a constant speed along the preset path, four types of basic data are collected simultaneously, including 3D LiDAR point cloud data, odometry pose data, visual landmark recognition data, and background electromagnetic field data; the four types of basic data correspond to geometric information completeness Cg, positioning attitude stability Sp, semantic landmark clarity Cl, and electromagnetic background stability Se, respectively; S1.2: Single-source data quality quantification: For each location point, the four types of basic data collected are calculated into dimensionless numerical quantification indicators, and their value ranges are all mapped to the closed interval [0,1]. The specific process is as follows: Geometric information completeness Cg: By calculating the number of effective point clouds within the scanning range of the current location's 3D LiDAR, and comparing it with the preset ideal number of point clouds, a normalized value is obtained; it reflects the clarity of the perceived geometric structure of the local environment; Positioning attitude stability Sp: Extract the trace of the pose covariance matrix output by the odometry. The smaller the trace, the more stable the pose estimation; the trace is inversely divided and normalized to obtain a value; it reflects the unmanned The stability of the machine's positioning at this point; Semantic landmark clarity Cl: calling the visual recognition algorithm to output the confidence score of the recognition of semantic landmarks in the field of view, which is itself a value in the interval [0,1]; if there are no landmarks in the field of view, the value is 0; reflecting the recognizability of environmental semantic information; Electromagnetic background stability Se: taking multiple short-term background electromagnetic field measurements at the current location point and calculating the standard deviation of this set of data; the smaller the standard deviation, the more stable the background field; similarly, the standard deviation is inversely normalized to obtain a value; and used to reflect the reliability of the electromagnetic environment as "background"; S1.3: The contribution weight is determined by combining the four in S1.2 above. Numerical quantification indicators are used to determine their respective contribution weights. These weights are not fixed but determined based on expert knowledge and application scenario priorities, using the Analytic Hierarchy Process (AHP). Specifically, in a substation environment, geometric information completeness (Cg) has the highest weight; positioning attitude stability (Sp) has the second highest weight; semantic landmark clarity (Cl) provides auxiliary positioning and its weight is lower than positioning attitude stability (Sp); electromagnetic background stability (Se), as the basis for special applications, has the lowest weight. Based on the above information, this implementation sets a set of initial weights: geometric information completeness contribution weight is 0.4, positioning attitude stability contribution weight is 0.3, and semantic landmark clarity contribution weight is 0.4. The contribution weight is 0.2, the contribution weight of electromagnetic background stability is 0.1, and the sum of all contribution weights is always 1; S1.4: The calculation formula of map quality index is as follows: map quality index equals the product of "geometric information completeness Cg" and its contribution weight, plus the product of "positioning attitude stability Sp" and its contribution weight, plus the product of "semantic landmark clarity Cl" and its contribution weight, plus the product of "electromagnetic background stability Se" and its contribution weight. The essence of the calculation process of map quality index is a weighted summation. Its initial source is the weighted average method in statistics, which is used to fuse multi-source heterogeneous data to obtain a comprehensive evaluation value. Further, the technical effect and logical rationality explanation: (1) The map quality index Mq calculation method proposed in this embodiment has complete operability. Taking any point P on the map as an example, its specific algorithm operability and logical deduction process are as follows: In this embodiment, point P is set. After the second step of quantitative calculation, the geometric information completeness of the index Cg=0 is obtained.9. Positioning attitude stability Sp=0.8, Cl=0.7, electromagnetic background stability Se=0.95; using the set weights; where Wg is the weight of the geometric information completeness Cg, and Wg=0.4; Wp is the weight of the positioning attitude stability Sp, and Wp=0.3; Wl is the weight of the semantic landmark clarity Cl, and Wl=0.2; We is the weight of the electromagnetic background stability Se, and We=0.1; data calculation: the formula described above is used for calculation, as follows: ; The results show that the map quality index Mq at point P is 0.835 and is recorded on the three-dimensional reference map, stored together with the coordinates of point P, visual landmarks and background electromagnetic field baseline values; (2) The physical dimensions and value ranges are explained as follows: after normalization, geometric information completeness Cg, positioning attitude stability Sp, semantic landmark clarity Cl and electromagnetic background stability Se are all dimensionless pure numbers with a value range of [0,1]; Wg, Wp, Wl and We are also dimensionless weight coefficients; therefore, Mq calculated by weighted summation is also a dimensionless pure number, which is logically consistent; since the value range of all input indicators is [0,1] and the sum of all weights is 1, the output value range of map quality index Mq is limited to the closed interval of [0,1]. When the map quality index Mq is 0, it only occurs when the completeness of the input index geometric information Cg, the stability of the positioning attitude Sp, the clarity of the semantic landmark Cl, and the stability of the electromagnetic background Se are all 0; this means that none of the sensors at that point have collected any valid data; when the map quality index Mq is 1, it only occurs when the completeness of the input index geometric information Cg, the stability of the positioning attitude Sp, the clarity of the semantic landmark Cl, and the stability of the electromagnetic background Se are all 1; this represents a theoretical "perfect data point" where all sensors have reached an ideal working state; (3) The technical effect is related to the invention objective as follows: when the map quality index The closer the Mq output is to 1, the higher the reliability of the 3D reference map at that point. A geometric information completeness Cg approaching 1 indicates a wider field of view for the 3D LiDAR scan; positioning attitude stability Sp approaching 1 indicates smoother UAV movement and smaller odometer error; semantic landmark clarity Cl approaching 1 indicates clearer visual landmarks; and electromagnetic background stability Se approaching 1 indicates less surrounding electromagnetic interference and a more stable background field. In subsequent positioning and navigation, the system will highly rely on the map information at this point, and the path planning algorithm will tend to choose to traverse the area with the highest map quality index Mq value to ensure positioning accuracy and safety. When the map quality index Mq output is closer to 0, it indicates that the reliability of the 3D reference map at that point is lower, and there is a risk of data loss or conflict. This situation is caused by the deterioration of one or more parameters, specifically including being in a narrow corner or with obstruction causing the geometric information completeness Cg to approach 0, slippery ground or a drone's sharp turn causing severe odometer drift causing the positioning attitude stability Sp to approach 0, the absence of effective semantic landmarks in the field of view causing Cl to approach 0, and being close to a normally operating device that generates strong electromagnetic fluctuations causing the electromagnetic background stability Se to approach 0. The system will mark this area as a "low confidence area". In emergency navigation, even if the area is geometrically passable, the path planner may increase its passage cost due to its low map quality index Mq value, or even actively avoid it during planning, thereby improving the robustness of the entire navigation scheme.

[0035] Furthermore, the technical effects and logical rationality of step S4 are explained in detail below: The specific algorithmic operability and logical deduction of the navigation strategy fusion index α and the three-dimensional grid passage cost index Cto are as follows: Scenario 1 is set as: routine inspection, good positioning; The parameters were set as follows: Clc=0.98, fault vector effectiveness index Vv=0.9; Data calculation: ; ; The results show that: the navigation strategy fusion index α is low, the total cost is determined by the target point approach potential Cco, the path planner will generate a path that is almost entirely oriented toward the preset target point, only guided by a very weak emergency target, and the behavior is stable. Scenario 2 is set as follows: Location service fails, but emergency signals are reliable; The parameters were set as follows: conventional location confidence Clc = 0.2, and fault vector validity index Vv = 0.9; Data calculation: ; ; The results show that: the navigation strategy fusion index α is high, and the total cost is mainly determined by the fault vector alignment Cem; the path planner will ignore the regular target points in the distance and give priority to generating a path that strictly avoids restricted areas and moves along the direction of the fault source pointing vector, thus realizing emergency navigation.

[0036] The physical dimensions and range of step S4 are explained as follows: Dimensional consistency: Since both the conventionally defined position confidence index Clc and the fault vector effectiveness index Vv are dimensionless, the navigation strategy fusion index α is also dimensionless; the 3D grid no-pass assignment Cst, the target point approach potential value Cco, and the fault vector alignment Cem can all be considered as dimensionless "cost" units, therefore the calculation of the 3D grid passage cost index Cto is dimensionally consistent; Output range of the navigation strategy fusion index α: Since the ranges of the fault vector effectiveness index Vv and the conventionally defined position confidence index Clc are both [0,1], the calculation result of α is the fault vector effectiveness index... The range of values ​​is also limited to the closed interval [0,1]. When the fault vector validity index Vv=0 or the conventional position confidence index Clc=1, the navigation strategy fusion index α is 0; when the fault vector validity index Vv=1 and the conventional position confidence index Clc=0, the navigation strategy fusion index α is 1. The technical effects of step S4 are related to the invention objectives as follows: When the navigation strategy fusion index α output approaches 0, the UAV navigation strategy tends to perform routine inspection tasks. At this time, in the three-dimensional grid passage cost index function, the weight of the target point approach potential value Cco approaches 1, and the UAV will firmly move towards the preset inspection target point. The path planning prioritizes "efficiency," conforming to the normal working mode. When the navigation strategy fusion index α output approaches 1, the UAV navigation strategy tends to perform emergency source tracing tasks. At this time, in the three-dimensional grid passage cost index function, the weight of the fault vector alignment Cem approaches 1, and the UAV will abandon the original inspection target and instead plan a path along the fault vector direction with "approaching the fault source" as the highest criterion. This ensures that when the main positioning system fails, the UAV can still perform the most critical emergency tasks with the help of an independent emergency navigation reference.

[0037] To verify the actual effect of the fault source pointing vector and its effectiveness index generation method in this embodiment, the following experiment was designed and executed: The test site was a simulated 110kV outdoor gas-insulated switchgear (GIS) substation environment. This environment had been pre-constructed using step S1, which included a three-dimensional reference map containing geometric and visual features and a background electromagnetic field distribution baseline map. The test equipment was a "Rayx-II" wheeled inspection drone equipped with the system described in this invention. A square-shaped electromagnetic sensor array consisting of four triaxial gaussmeters was symmetrically deployed at the center of its top platform, with a sensor spacing of 20 cm. The test objective was to simulate a progressive partial discharge fault inside the GIS equipment and evaluate the drone's ability to generate the fault source pointing vector and its effectiveness index under different locations and electromagnetic interference environments. The fault simulation was achieved by placing a programmable microsecond-level pulse current source near the target GIS basin insulator. This current source could simulate the leakage magnetic field generated by partial discharges of different intensities. The test process was as follows: First, the "Rayx-II" drone was started and placed in normal inspection mode. At a certain moment, the pulse current source was remotely activated to simulate the occurrence of a fault. Upon receiving an anomaly alarm from the Supervisory Control and Data Acquisition (SCADA) system, the UAV immediately switches to emergency response mode and begins executing step S2. As the UAV moves from its initial position (test point A) towards the simulated fault source (test point F), the system continuously records and calculates all core parameters related to S2 at a frequency of 1Hz. The test points involved include: Test point A (far from the fault source), Test point B (initial approach), Test point C (medium distance), Test point D (close-range clear path), Test point E (close-range interfered path), and Test point F (immediately adjacent to the fault source). Specifically, in a single calculation cycle, the UAV's execution details are as follows: At the current location, the sensor array simultaneously acquires real-time three-dimensional magnetic field strength data and calculates its vector sum to obtain the real-time field strength measurement.The system then retrieves the background field strength baseline value corresponding to the current coordinate from the background electromagnetic field distribution baseline map loaded from memory. Based on the principle of differential analysis, the system accurately calculates the abnormal field strength of the spatial electromagnetic field generated by the simulated fault source by subtracting the baseline value from the real-time measured value. Next, based on the multi-point gradient estimation method, the system uses the difference in abnormal field strength values ​​measured by four sensors to calculate the direction of the fastest field strength increase in the local space using the finite difference method, generating a three-dimensional fault source pointing vector. To evaluate the reliability of this vector in real time, the system simultaneously performs quantization calculations. First, the calculated abnormal field strength of the spatial electromagnetic field is substituted into a preset normalization function. In this embodiment, the detection threshold is set to 0.3 μT and the saturation threshold to 8.0 μT, yielding the abnormal signal intensity factor Fa. Second, the system caches the pointing vector of the current cycle and compares it with the pointing vector of the previous cycle. After dot product operation and normalization, the cosine value of the angle between the pointing vectors is obtained, and then the pointing stability factor Fs is calculated through linear mapping. At the same time, the system calculates the variance of the abnormal field strength values ​​measured by the four sensors, and normalizes the variance value to obtain the spatial gradient clarity Fc. Finally, based on the preset contribution weights, the final fault vector effectiveness index Vv is calculated by weighted summation. To verify the robustness of the invention in complex electromagnetic environments, a large metal plate was placed between the UAV and the fault source at test point E to simulate multipath reflection interference caused by other equipment or metal structures. By comparing the data of test points D and E, i.e., the clear path and the interfered path, the key role of the fault vector effectiveness index Vv in identifying the quality of navigation information in the invention can be clearly demonstrated. The final results are shown in Table 1.

[0038] Table 1 is the experimental data record table in this embodiment.

[0039] As shown in Table 1, from test point A to test point D, as the UAV approaches the fault source, all original data and factors (Fa, Fs, Fc) show a monotonically increasing trend. The final fault vector effectiveness index Vv steadily improves from an almost zero unreliable state to a highly reliable 0.91. This proves that the method can accurately reflect the objective law that pointing vector quality improves with the improvement of signal-to-noise ratio and geometric relationship. Comparing test points D and E is the core of this embodiment. Although the abnormal field strength at test point E is similar to that at test point D, resulting in a still high abnormal signal strength factor Fa, the interference from the metal plate causes severe electromagnetic wave multipath reflection. This is directly reflected in the sharp increase of the "pointing vector angle" to 45.5 degrees, causing the pointing stability factor Fs to plummet to 0.35. Simultaneously, the "field strength measurement variance" also decreases significantly, causing the spatial gradient clarity Fc to drop to 0.18. Ultimately, the fault vector effectiveness index Vv calculated by this invention is only 0.48 at test point E, far lower than 0.91 at test point D. This perfectly demonstrates the advantages of this invention: the system does not... Instead of being "deceived" by the high field strength on the surface, the system accurately identifies the low reliability of the current pointing vector due to directional instability and gradient ambiguity through comprehensive evaluation. In the subsequent S4 decision, the system will reduce the adoption weight of the pointing vector accordingly to prevent the drone from heading in the wrong direction, thus demonstrating excellent robustness and intelligence. Data from test point F shows that at the ideal location close to the fault source, all indicators reach or approach the saturation value of 1, and the obtained fault vector effectiveness index Vv is also 1.00, indicating that the pointing vector quality has reached the optimal level, providing the most reliable basis for the final accurate positioning.

[0040] Furthermore, this embodiment also provides a substation positioning and navigation system combining positioning and vision-assisted positioning, including a microprocessor and a memory interconnected, wherein the microprocessor is programmed or configured to execute the substation positioning and navigation method combining positioning and vision-assisted positioning. This embodiment also provides a computer-readable storage medium storing a computer program or instructions programmed or configured to execute the substation positioning and navigation method combining positioning and vision-assisted positioning via a processor. This embodiment also provides a computer program product including a computer program or instructions programmed or configured to execute the substation positioning and navigation method combining positioning and vision-assisted positioning via a processor.

[0041] Those skilled in the art will understand that the technical solutions provided by this invention may take the form of a method, system, or computer program product. Therefore, this invention may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this invention may take the form of a computer program product embodied on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, produce an implementation of the flowchart... Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0042] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A substation positioning navigation method combining positioning and visual aided positioning, characterized in that, The method comprises the following steps: S1, collecting and fusing three-dimensional laser ranging data, dead reckoning data and three-dimensional vision data by an unmanned aerial vehicle (UAV) on a preset three-dimensional flight path in a substation to construct a three-dimensional reference map containing geometric features and visual semantic landmarks; collecting background electromagnetic field data of each position point on the three-dimensional flight path by the UAV to establish a background electromagnetic field distribution baseline map associated with the three-dimensional reference map; S2, collecting electromagnetic field data of the substation in an emergency response mode by the UAV and comparing the electromagnetic field data with the background electromagnetic field distribution baseline map to calculate a spatial electromagnetic field anomaly field strength representing a fault source; calculating a spatial electromagnetic field anomaly gradient based on the spatial electromagnetic field anomaly field strength to serve as a fault source pointing vector pointing to the fault source; S3, estimating a position of the fault source on the three-dimensional reference map according to the fault source pointing vector and the spatial electromagnetic field anomaly field strength; dynamically constructing a risk adaptive elliptical forbidden zone at the estimated position of the fault source on the three-dimensional reference map according to the spatial electromagnetic field anomaly field strength and the fault source pointing vector; S4, calculating a passing cost index Cto of the UAV passing through each three-dimensional grid in the three-dimensional reference map by using the fault source pointing vector and the risk adaptive elliptical forbidden zone, and searching for a path from a current position of the UAV to a preset target point, which bypasses the risk adaptive elliptical forbidden zone, points to the fault source and has the lowest total passing cost index Cto by using a preset path searching algorithm. In step S2, the electromagnetic field data of the substation in the emergency response mode is subtracted from the background electromagnetic field data in the background electromagnetic field distribution baseline map to obtain the calculated spatial electromagnetic field anomaly field strength representing the fault source.

2. The combined location and visual aid located substation positioning navigation method of claim 1, wherein, In step S2, the spatial electromagnetic field anomaly gradient is calculated based on the spatial electromagnetic field anomaly field strength to serve as the fault source pointing vector pointing to the fault source, which comprises: differentiating the spatial electromagnetic field anomaly field strength of different position points to obtain the spatial electromagnetic field anomaly gradient of each position point, and selecting the maximum spatial electromagnetic field anomaly gradient as the fault source pointing vector pointing to the fault source.

3. The combined location and visual aid located substation positioning navigation method of claim 1, wherein, Step S3 comprises:

4. The combined location and visual aid located substation positioning navigation method of claim 1, wherein, S3.1, finding a position where the maximum value of the spatial electromagnetic field anomaly field strength is located on the three-dimensional reference map as the estimated position of the fault source according to the direction of the fault source pointing vector; S3.2, taking the estimated position of the fault source as the center of the risk adaptive elliptical forbidden zone; S3.3, mapping the spatial electromagnetic field anomaly field strength as a risk radius Rr to serve as a semi-minor axis of the risk adaptive elliptical forbidden zone, and mapping the effectiveness of the fault source pointing vector as an uncertainty eccentricity Eu to serve as an eccentricity of the risk adaptive elliptical forbidden zone, thereby obtaining a dynamically constructed risk adaptive elliptical forbidden zone. ​ 5. The combined location and visual aid located substation positioning navigation method of claim 4, wherein, In step S3.3, the mapping of the spatial electromagnetic field anomaly field strength to the risk radius Rr includes: calculating a product of the spatial electromagnetic field anomaly field strength and a preset anomaly field strength gain coefficient Kga, and adding a basic safety radius Rb to the product to obtain the risk radius Rr, wherein the basic safety radius Rb is a preset constant value representing a minimum safety distance that the UAV needs to maintain; the anomaly field strength gain coefficient Kga is a preset constant coefficient for converting and scaling the physical dimension of the spatial electromagnetic field anomaly field strength to a length dimension.

6. The combined location and visual aid located substation positioning navigation method of claim 4, wherein, In step S3.3, the mapping of the validity of the fault source pointing vector to the uncertainty eccentricity Eu includes: calculating a fault vector validity index Vv according to the following formula: ； Wherein, Wa, Ws and Wc are weight coefficients, the anomaly signal strength factor Fa, the pointing stability factor Fs and the spatial gradient clarity factor Fc are all dimensionless normalized factors with a value range of [0, 1], and the value range of the fault vector validity index Vv is [0, 1]; the anomaly signal strength factor Fa is obtained by normalizing the spatial electromagnetic field anomaly field strength through a preset threshold function, which outputs 0 when the spatial electromagnetic field anomaly field strength is lower than a detection lower limit, and outputs 1 when the spatial electromagnetic field anomaly field strength is higher than a saturation upper limit to reflect the signal-to-noise ratio of the fault signal; the pointing stability factor Fs is a linear mapping result of the cosine value of the angle between the fault source pointing vector at the current moment and the fault source pointing vector at the previous moment, which is in the interval [-1, 1] and is linearly mapped to the interval [0, 1] to reflect the continuity and anti-interference ability of the direction information; the spatial gradient clarity factor Fc is a normalized processing result of the variance of the spatial electromagnetic field anomaly field strength measured by all sensors in the sensor array of the UAV; the difference between 1 and the fault vector validity index Vv is taken as the obtained uncertainty eccentricity Eu.

7. The combined location and visual aid located substation positioning navigation method of claim 1, wherein, The function expression of the passing cost index Cto of the UAV through each three-dimensional grid in the three-dimensional reference map calculated in step S4 is: ； wherein, is the navigation strategy fusion index, Cem is the fault vector alignment degree, Cst is the three-dimensional grid forbidden assignment, and Cco is the target point approach potential value. The calculation of the fault vector alignment degree Cem includes: first, calculating the vector of the current three-dimensional grid pointing to the adjacent three-dimensional grid, calculating the dot product of the vector of the current three-dimensional grid pointing to the adjacent three-dimensional grid and the fault source pointing vector, and then subtracting 1 from the dot product calculation result to obtain the fault vector alignment degree Cem; if the three-dimensional grid is located in the static obstacle or the risk adaptive elliptical forbidden area, the three-dimensional grid forbidden assignment Cst is infinite; otherwise, the three-dimensional grid forbidden assignment Cst is zero; the target point approach potential value Cco is the Euclidean distance from the center of the current three-dimensional grid to the preset regular target point. The regular target point refers to the next to-be-arrived waypoint in a series of ordered three-dimensional waypoint sequences planned on the three-dimensional reference map according to the preset inspection task and route. These to-be-arrived waypoints represent specific three-dimensional spatial positions that need to be checked and flown to by the unmanned aerial vehicle in a non-emergency state; the navigation strategy fusion index wherein Vv is the fault vector validity index, and Clc is the regular position reliability of the current three-dimensional grid. The value range of the regular position reliability Clc is [0, 1], which represents the map quality of the current three-dimensional grid in the three-dimensional reference map. The map quality is the weighted sum result of the geometric information completeness Cg, the positioning attitude stability Sp, the semantic landmark clarity Cl, and the electromagnetic background smoothness Se of the current three-dimensional grid. The geometric information completeness Cg is the ratio of the number of effective point clouds in the laser radar scanning range of the current three-dimensional grid to the ideal point cloud number. The positioning attitude stability Sp is the value obtained by taking the inverse and normalizing the trace of the pose covariance matrix output by the odometer of the current three-dimensional grid. The semantic landmark clarity Cl is the recognition confidence score of the semantic landmark in the field of view of the current three-dimensional grid, which is a value in the interval [0, 1]. The electromagnetic background smoothness Se is the value obtained by taking the inverse and normalizing the standard deviation of the multiple background electromagnetic field measurement results of the current three-dimensional grid.

8. A substation positioning navigation system combining positioning and visual aided positioning, comprising a microprocessor and a memory connected to each other, characterized in that, The microprocessor is programmed or configured to perform the substation positioning and navigation method of combined positioning and visual aided positioning according to any one of claims 1-7.

9. A computer-readable storage medium having stored therein a computer program or instructions, characterized in that, The computer program or instructions are programmed or configured to perform the substation positioning and navigation method of combined positioning and visual aided positioning according to any one of claims 1-7 by the processor.

10. A computer program product comprising computer programs or instructions, characterized in that, The computer program or instructions are programmed or configured to perform the substation positioning and navigation method of combined positioning and visual aided positioning according to any one of claims 1-7 by the processor.

Images