Transmission, distribution and transformation unmanned aerial vehicle path planning method based on combination of improved RRT algorithm and Voronoi diagram

Abstract

The invention discloses a transmission and distribution transformer unmanned aerial vehicle path planning method based on combination of an improved RRT algorithm and a Voronoi diagram. The method comprises the steps of defining a corresponding threat degree according to magnetic field distribution of inspection equipment; generating a Voronoi diagram according to the position of the inspection equipment, and weighting the Voronoi diagram according to the threat degree information; a path planning algorithm is improved, a weighted Voronoi graph boundary intersection point is used as an expansion node of an RRT algorithm, a Voronoi graph point closest to a current node is selected as a next node, and a tree structure generated by the RRT algorithm is expanded along the boundary of the Voronoi graph; informed sampling is introduced on the basis of an RRT algorithm, random points are constrained in an ellipse range, and father nodes are reselected and rewiring is performed in combination with an RRT * algorithm; performing redundant node optimization on the planned path by using a triangular optimization method; and smoothing the generated path to further optimize the path of the unmanned aerial vehicle. According to the method, a proper path is planned while the safe distance of the unmanned aerial vehicle is ensured, and collision between the unmanned aerial vehicle and the obstacle is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of inspection drone path planning, and in particular to a distribution transformer drone path planning method based on the combination of an improved RRT algorithm and a Voronoi diagram. Background Art

[0002] The deployment environment of my country's power lines varies greatly, but most of the power lines are directly exposed to the environment for a long time and are affected by many severe climates and human activities. Regular inspections of the power system can timely discover and solve the above problems, ensuring the stability and safety of the power system during operation. Regular inspections, special inspections, and fault inspections are of great significance for prevention and maintenance. With the continuous development of industrial drone technology, the power industry has been able to use drones proficiently. At present, drones are mainly used for transmission line inspections. Drones have outstanding advantages in line inspections. They can detect lines without power outages, reduce labor costs, and have high inspection efficiency. Drone path planning Track planning is to plan a path that meets flight requirements and can avoid threats and complete established tasks given the starting point and target point.

[0003] Drones have the advantages of strong flight flexibility, high inspection efficiency and short observation distance. Therefore, using drones for inspection in substations can effectively solve the visual blind spot problem of intelligent inspection robots, but there are also many safety hazards. For example, the complex strong electromagnetic field environment will affect the normal operation of drones, communication will be disturbed, and the collected information will be distorted; the drone will be hit by discharge when it flies close to the high-voltage electrode, causing damage to the drone; when the drone is in the gap between the high-voltage electrodes, it will cause the space electric field distortion and insulation gap breakdown. The inspection will encounter a large number of equipment, densely arranged and complex scenes, and there will be strong electromagnetic coupling between the live equipment. The complex strong electromagnetic environment puts forward higher requirements for drone inspection safety. If the drone is too close to the live body during the inspection flight, it may cause the metal or exposed parts of the drone body to discharge, thereby damaging the drone or causing the drone to lose control and cause a safety accident. Therefore, the effect of drone path planning will affect the effectiveness of drone intelligent inspection. If the flying robot is to realize autonomous inspection of the tower, it is very necessary to plan the inspection path for it. The goal of path planning is to find an optimal collision-free path from the initial state to the target state based on some performance indicators such as shortest time, shortest path length, and least cost, taking into account complex environments, geometric factors, physical properties, and time constraints.

[0004] At present, many academic achievements have been made in the research on path planning at home and abroad. The research methods of path planning are mainly divided into sampling-based path, search-based (A* algorithm, genetic algorithm) path planning and intelligent optimization algorithm-based path planning. Among them, the most widely used is the rapidly expanding random tree (RRT) algorithm based on sampling. The planning efficiency of the RRT algorithm in high-dimensional space is more obvious. However, due to the randomness of sampling in the algorithm, the path planned by the algorithm each time is inconsistent, the optimal path cannot be obtained, and the path planning speed is slow. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a distribution transformer UAV path planning method based on the combination of an improved RRT algorithm and a Voronoi diagram, which plans a suitable path while ensuring that the UAV maintains a safe distance, avoids collision between the UAV and obstacles, and improves the speed of UAV path planning and improves efficiency.

[0006] The technical solution adopted by the present invention is:

[0007] The transmission and distribution transformer UAV path planning method based on the improved RRT algorithm combined with the Voronoi diagram is characterized by comprising the following steps:

[0008] Step 1: Model the environment of the inspection equipment on the transmission and distribution transformer line, and define the corresponding threat level based on the magnetic field distribution of the inspection equipment;

[0009] Step 2: Generate a Voronoi diagram based on the location of the inspection equipment, and perform weighted processing on the Voronoi diagram based on the threat level information;

[0010] Step 3: Improve the path planning algorithm and use the boundary intersection of the weighted Voronoi diagram as the extension node of the RRT algorithm. In the RRT algorithm extension step, select the Voronoi diagram point closest to the current node as the next node, so that the tree structure generated by the RRT algorithm is extended along the boundary of the Voronoi diagram.

[0011] Step 4: Introduce Informed sampling based on the RRT algorithm, constrain the random points within the ellipse, and then combine the RRT* algorithm to reselect the parent node and rewire;

[0012] Step 5: Use the triangle optimization method to optimize the redundant nodes of the planned path to improve the path quality;

[0013] Step 6: Smooth the generated path to further optimize the drone path.

[0014] In step 1, when inspecting the power system transmission and distribution lines or substations, it is necessary to pay attention to the operating status and potential faults of equipment such as transformers, switchgear, relay protection devices, cables and overhead lines, and mark each inspection equipment, obstacle, channel and important location on the inspection map. In addition to the impact of visible physical obstacles on drone path planning, it is also necessary to consider the impact of invisible magnetic field size on the drone flight path. The magnetic field size of each inspection equipment is collected through sensors and historical data, which is recorded as the threat level of the inspection equipment.

[0015] In step 2, a weighted Voronoi diagram is used to reflect the difference in the impact of various obstacles on the drone. The improved Delaunay triangle side length calculation formula is:

[0016]

[0017] In formula (1), A, B, and C are three obstacles, and their influence on the UAV varies; △ABC is a Delaunay triangle; L AB is the distance between A and B; T A and T B are the magnitudes of the magnetic fields of points A and B respectively; similarly determine the positions of points F and E;

[0018] Connect D, E, and F to obtain the segmented triangle △DEF, and make an inscribed circle. The lines OD, OE, and OF connected to the center of the circle are the edges of the weighted Voronoi diagram. Figure 2 shown.

[0019] In step 3, in the RRT algorithm expansion step, after generating potential nodes, all Voronoi nodes are extracted from the Voronoi diagram, and these Voronoi nodes are used as potential expansion nodes of the RRT tree to guide the growth of the tree.

[0020] For the current node, the Voronoi diagram point closest to the current node is selected as the next node to be expanded, and this node is given priority, so that the tree structure generated by the RRT algorithm will be more inclined to expand along the boundary of the Voronoi diagram, thereby improving the safety of the path.

[0021] In step 3, the current node P c The closest Voronoi node P in the Voronoi v The distance between them is:

[0022]

[0023] In the above formula, d(P c ,P v ) represents the previous node P cThe closest Voronoi node P in the Voronoi v The distance between

[0024] x v Represents the Voronoi node P v The horizontal axis of

[0025] x c Indicates the current node P c The horizontal axis of

[0026] y v Represents the Voronoi node P v The vertical coordinate of

[0027] y c Indicates the current node P c The vertical coordinate of .

[0028] Traverse all nodes in the Voronoi diagram and calculate the value of the node P c The distance between the nodes is 1 and 2, and the node with the smallest distance is selected as the target Voronoi node P. nearest .

[0029]

[0030]

[0031] P new =( x +Δd·v x ,y c +Δd·v y ) (7);

[0032] In the above formula, v is the current node P c To the target Voronoi node P nearest The unit vector of

[0033] x nearest Represents the distance to the current node P c The nearest Voronoi node P nearest The horizontal axis of

[0034] y nearest Represents the distance to the current node P c The nearest Voronoi node P nearest The vertical coordinate of

[0035] x c Indicates the current node P c The horizontal axis of

[0036] y c Indicates the current node P cThe vertical coordinate of

[0037] v x represents the horizontal coordinate of the unit vector;

[0038] v y Represents the ordinate of a unit vector.

[0039] Get the new node P according to the step size Δd new . The new node P new Add to the RRT tree and add P current Set as the parent node.

[0040] The new node P new Set as the current node for the next expansion, and repeat the above operation until the target node is reached. Step 4 includes improving the path planning Informed-RRT* algorithm, as follows:

[0041] For the sampling area, a local ellipse sampling method is used instead of global uniform sampling. By shrinking the sampling area into an ellipse area, a large number of meaningless sampling points can be avoided. As the path length shortens, the ellipse sampling area gradually shrinks until the path optimization is completed:

[0042] The ellipse area of ​​the Informed-RRT* algorithm takes the path starting point and the path target point as the focus, and the initial sampling ellipse equation is:

[0043]

[0044] In the above formula, a represents the length of the major semi-axis of the ellipse; b represents the length of the minor semi-axis of the ellipse.

[0045] According to the rotation and translation matrix, the sampling point ellipse coordinates are converted to actual coordinates:

[0046] [x'y'] = R · [xy] + T (3);

[0047] Among them, x represents the actual X-axis coordinate after coordinate transformation; y' represents the actual Y-axis coordinate after coordinate transformation. Rotation matrix Translation matrix T = [x c y c ], θ represents the angle between the line connecting the two foci of the ellipse and the actual coordinate system X; x c Indicates the amount of translation in the horizontal direction; y c Indicates the amount of translation in the vertical direction.

[0048] In step 4, the RRT* algorithm is combined to reselect the parent node and rewire, as follows:

[0049] Reselect the parent node mainly with x newWith x as the center, all adjacent nodes are found within a fixed radius. All nodes within the range will be used as backup parent nodes of the new node. The value of the path after each backup parent node is connected is calculated respectively. After comparing the path costs, the backup parent node with the smallest path cost is selected as the final parent node. The parent node is replaced by the original x nearest Become the new parent node x min .like Figure 3 As shown, the original path is A→C→F→J, with J as the origin, the fixed path as the radius, the backup nodes as E, F, I, and G. After comparing the path costs, E is selected as the new parent node.

[0050] After completing the reselection of the parent node, perform the rewiring operation and change x new As the parent nodes of the remaining nodes in the circle, and calculate each node with x new As parent node to x start Is the path cost of the new path less than the original path cost? If so, choose x new The parent node of this node, otherwise the child nodes are not updated. Figure 3 As shown, for node F, comparing A→C→F and A→E→J→F, the original path is shorter and is not updated; for node I, comparing paths A→E→J→I and A→C→F→I, the updated path is shorter, so node I is taken as x new The child nodes of .

[0051] In step 5, the triangle optimization method is used to optimize the redundant nodes of the planned path to improve the path quality. Specifically, the specific process of the triangle optimization method is as follows:

[0052] Name the nodes in the path [X start , X1, X2...X k-1 , X k , X goal ], and change the starting point X start The length of the edge between node X1 and node X2 is recorded as b, and the length of the edge between node X2 and the starting point X is recorded as start The side length is recorded as c, forming a triangle;

[0053] Find the first triangle node X in the initial path with side length c that collides with the obstacle k ;

[0054] Node X k-1 With node X k The connection between them is discretely split into n segments;

[0055] From the starting point X start Perform collision detection with discrete points P in turn and find the child node X of the first collision node in the collection P. stemp;

[0056] X stemp Update to the new starting node;

[0057] Follow the above steps, such as Figure 4 As shown in the figure, the three black circles represent obstacles, the black path is the initial path, and there are 9 path nodes from the starting point to the target point; the first triangle from the starting point is X start X1X2, such as Figure 4 As shown by the green dashed line in the middle, the long side of the triangle does not collide with the obstacle, so we continue to detect the second triangle X starting from the starting point. start X2X3, such as Figure 4 As shown by the red dotted line, the long side of the triangle collides with the obstacle, so X3 is the X in the above step. k , X2 is X k Child node X k-1 ; Discrete splitting is performed on the line connecting X2-X3 to obtain a series of discrete points P, such as Figure 4 As shown by the red dot on the X2-X3 line;

[0058] The discrete split expression is as follows:

[0059] P=(x k-1 ,y k-1 )+[(x k ,y k )-(x k-1 ,y k-1 )]×θ

[0060] =(x k-1 ,y k-1 )+(x k ,y k )·θ-(x k-1 ,y k-1 )·θ

[0061] =(1-θ)·(x k-1 ,y k-1 )+θ·(x k ,y k ) (8)

[0062] Among them, the value of θ is θ is the discrete step length, which is used to control the discrete splitting of the line segment into n segments;

[0063] (x k-1 ,y k-1 )、(x k ,y k ) are the coordinates of the path nodes;

[0064] P is the node set after the line segment is discretely split, which is used for collision detection with the starting point.

[0065] Set the starting point X start Perform collision detection with discrete points in turn to find the child node of the first collision node in the collection P, which is X in the figure. stemp1 ; At this time, the path is X start -X1-X2-X3 updated to X start -X stemp1 , that is, the redundant nodes X1 and X2 are deleted; then X stemp1 Update to the new starting point and repeat the above steps until the target point X goal , you can delete the redundant nodes in the initial path, optimize the tortuosity of the path, and reduce the length of the path; the path after triangle optimization is as follows Figure 4 As shown in the yellow path, X start -X stemp1 -X stemp2 -X goal .

[0066] In step 6, a cubic B-spline curve is used to smooth the path; the details are as follows:

[0067] The B-spline curve expression is

[0068]

[0069] Where: P i Indicates the curve control point; F i,k (t) represents the k-order B-spline basis function, with k=3.

[0070] The cubic B-spline curve basis function expression is:

[0071]

[0072] Substituting equation (10) into equation (9), we finally get the cubic B-spline curve expression:

[0073] P(t)=P0*F 0,3 (t)+P1*F 1,3 (t)+P2*F 2,3 (t)+P3*F 3,3 (t) (11)

[0074] After the curve control points are given, a smooth cubic B-spline curve can be obtained, such as Figure 7 shown.

[0075] The present invention provides a transmission and distribution transformer UAV path planning method based on the combination of improved RRT algorithm and Voronoi diagram, and the technical effects are as follows:

[0076] 1) The method of the present invention simplifies the search nodes in path planning, reduces the calculation amount of the RRT algorithm, and improves planning efficiency.

[0077] 2) The present invention takes into account that during power inspections, due to the concentration of equipment and strong electromagnetic coupling, drones and live equipment need to maintain a safe distance. The map is pre-processed through a weighted Voronoi diagram to avoid the impact of live equipment on the drone, effectively avoid obstacles, and ensure the safety of the inspection path.

[0078] 3) The RRT algorithm of the present invention gives priority to the intersection of the Voronoi diagram boundary for new nodes, and then improves the RRT global uniform sampling strategy to Informed ellipse sampling, which reduces the sampling range of the RRT algorithm and improves the path planning speed.

[0079] 4) The present invention uses a B-spline curve to smooth the sudden change of the curvature of the obtained path, reduce the violent turning of the UAV during flight, and ensure the safety and stability of the UAV.

[0080] 5) Based on the RRT algorithm, the present invention optimizes the selection and sampling methods of extended nodes, improves the path planning efficiency, shortens the path planning length, realizes rapid UAV path planning, and meets the UAV flight dynamics requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0082] Figure 1 This is a flow chart of the UAV path planning of the present invention.

[0083] Figure 2 Schematic diagram of the improved weighted Voronoi diagram.

[0084] Figure 3 Schematic diagram of reselecting parent nodes and rewiring.

[0085] Figure 4 Schematic diagram of the triangle optimization process.

[0086] Figure 5 This is the traditional Voronoi diagram after preprocessing the 100×100km inspection map.

[0087] Figure 6 is the weighted Voronoi diagram.

[0088] Figure 7 is a cubic B-spline curve.

[0089] Figure 8(a) shows the layout of a 500kV substation Figure 1 ;

[0090] Figure 8(b) is the layout diagram of the 500kV substation. DETAILED DESCRIPTION

[0091] A transmission and distribution transformer UAV path planning method based on the combination of an improved RRT algorithm and a Voronoi diagram. The method first models the environment of the equipment that needs to be inspected on the transmission and distribution transformer line, and defines the corresponding threat level based on the magnetic field distribution of the equipment. The map is processed by a weighted Voronoi diagram to extract the guide points for path planning, and the global path guide points of the RRT algorithm are generated accordingly. The present invention innovates the path planning algorithm, and uses the boundary intersection of the Voronoi diagram as the node of the RRT algorithm. In the RRT expansion step, the Voronoi diagram point closest to the current node is selected as the next node, so that the tree structure generated by the RRT algorithm will be more inclined to expand along the boundary of the Voronoi diagram. On the basis of the RRT algorithm, Informed sampling is introduced to constrain random points within the ellipse range, and then combined with RRT* to reselect parent nodes and rewire, and finally the generated path is smoothed to further optimize the UAV path.

[0092] The transmission and distribution transformer UAV path planning method based on the improved RRT algorithm combined with the Voronoi diagram is as follows: Figure 1 As shown, the following steps are included:

[0093] Step 1: Environment construction and map preprocessing:

[0094] In step 1, when inspecting the power system transmission and distribution lines or substations, it is necessary to pay attention to the operating status and potential faults of equipment such as transformers, switchgear, relay protection devices, cables and overhead lines, and mark each inspection equipment, obstacle, channel and important location on the inspection map. As shown in Figure 8(a) and Figure 8(b), a 500kV substation layout diagram. In addition to visible physical obstacles that drones cannot pass through and need to detour, it is also necessary to consider the impact of invisible magnetic fields generated by equipment such as transformers on the flight path of drones, as shown in Table 1. The magnetic field size of each inspection device is collected through sensors and historical data. It is recorded as the threat level of the device.

[0095] Table 1 Effect of magnetic field on the flight path of UAV

[0096] Magnetic field strength Impact on drones ≥50μT Inaccurate magnetometer readings, affecting drone stability and flight control ≥100μT May affect the signal quality of the GPS receiver, resulting in navigation errors ≥200μT Loss of control due to magnetic field interference, leading to a crash

[0097] Step 2: Generate a Voronoi diagram based on the device location and perform weighted processing on the Voronoi diagram based on the threat level information:

[0098] Generate a Voronoi diagram based on the device location, such as Figure 5As shown in the figure, taking a 100×100km map as an example, the blue dots represent 30 energized devices, and a conventional Voronoi diagram is generated. The conventional Voronoi diagram assumes that the intensity values ​​of the threat bodies are the same, and each Voronoi edge is the perpendicular bisector of two adjacent threat points. Therefore, when the drone flies along the Voronoi edge, it does not lean to any side. However, during the actual power system inspection, the drone may be disturbed in the strong electromagnetic field, affecting the normal operation of the navigation and control system.

[0099] Since the magnetic field strength of all devices does not have the same impact on the drone, the weighted Voronoi diagram used in the present invention can better reflect the difference in the impact of various obstacles on the drone. The improved Delaunay triangle side length calculation formula is:

[0100]

[0101] In the formula, A, B, and C are three obstacles, each posing a different degree of threat to the drone, and △ABC is a Delaunay triangle. AB is the distance between A and B; T A and T B is the magnitude of the magnetic field of A and B. Similarly, determine the positions of points F and E. Connect D, E, and F to obtain the segmented triangle △DEF, make an inscribed circle, and the lines OD, OE, and OF with the center of the circle are the edges of the improved weighted Voronoi diagram.

[0102] like Figure 6 As shown in the figure, the pink line is the weighted Voronoi diagram, and the solid pink line is more inclined to the threat source that poses less threat to drones. It can better reflect the differences in the threats of various threat sources to drones than the traditional Voronoi diagram.

[0103] Step 3: Improve the path planning Informed-RRT* algorithm:

[0104] The present invention adopts a local elliptical sampling method for the sampling area instead of global uniform sampling, and avoids generating a large number of meaningless sampling points by shrinking the sampling area into an elliptical area. As the path length shortens, the elliptical sampling area gradually shrinks until the path optimization is completed.

[0105] The ellipse area of ​​the Informed-RRT* algorithm takes the path starting point and the path target point as the focus, and the initial sampling ellipse equation is:

[0106]

[0107] According to the rotation and translation matrix, the sampling point ellipse coordinates are converted to actual coordinates:

[0108] [x' y'] = R · [xy] + T (3);

[0109] The rotation matrix Translation matrix T = [x c y c ], θ represents the angle between the line connecting the two foci of the ellipse and the actual coordinate system X.

[0110] Step 4: Use the boundary intersection of the weighted Voronoi diagram as a potential expansion node of the RRT algorithm:

[0111] In the RRT expansion step, after generating potential nodes, all Voronoi nodes are extracted from the Voronoi diagram. These nodes are used as potential expansion nodes of the RRT tree to guide the growth of the tree. Figure 3 As shown in the figure. For the current node, the Voronoi diagram node closest to the current node is selected as the next node to be expanded, and this node is given priority, so that the tree structure generated by the RRT algorithm will be more inclined to expand along the boundary of the Voronoi diagram, thereby improving the safety of the path.

[0112] Current node P c The closest Voronoi node P in the Voronoi v The distance between them is:

[0113]

[0114] Traverse all nodes in the Voronoi diagram and calculate the c The distance between the nodes is 1 and 2, and the node with the smallest distance is selected as the target Voronoi node P. nearest .

[0115]

[0116]

[0117] P new =(x c +Δd·v x ,y c +Δd·v y ) (7)

[0118] v is the current node P c To the target Voronoi node P nearest The unit vector of the node P is obtained by step length Δd. new . The new node P new Add to the RRT tree and add P current Set as the parent node.

[0119] The new node Pnew Set it as the current node for the next expansion, and repeat the above steps until the target node is reached.

[0120] This method guides the expansion direction of RRT nodes, solves the problem of a large number of redundant nodes in RRT, and enables a path to be obtained faster and more stably when performing path planning in any environment.

[0121] Step 5: Reselect parent nodes and rewire the new nodes:

[0122] On the random tree planned by the RRT algorithm, the RRT* algorithm is used to reselect parent nodes and rewire when adding new nodes, so as to select nodes with higher quality and achieve the purpose of reducing path costs.

[0123] Step 5: Use the triangle optimization method to optimize the redundant nodes of the planned path to improve the path quality.

[0124] Step 6: Use cubic B-spline curve to smooth the path:

[0125] Since the paths finally planned by the improved RRT algorithm in the present invention are composed of random nodes, and due to the limitation of the expansion step size, the generated paths are tortuous, which does not conform to the actual flight trajectory of the drone, and will reduce the operating efficiency and stability of the drone. Therefore, a cubic B-spline curve is used to smooth the sudden change of the curvature of the obtained path. The characteristics of each generated path are recorded, and the comparison is shown in Table 2.

[0126] Table 2 Performance comparison of RRT, RRT* and improved RRT of the present invention under the same environment

[0127] algorithm Time consumed / s Path cost / km Number of nodes RRT 0.456 221.8 28.8 RRT* 0.961 201.3 32.3 The present invention improves the RRT algorithm 0.378 168.5 15.8

[0128] It can be seen from Table 2 that, compared with the traditional RRT algorithm and the RRT* algorithm, the improved RRT algorithm of the present invention not only takes the shortest time, but also generates the path with the lowest cost and requires the least number of nodes, which proves the superiority and effectiveness of the algorithm.

Claims

1. A transmission and distribution UAV path planning method based on the improved RRT algorithm combined with the Voronoi diagram is characterized by The following steps are involved: Step 1: Conduct environmental modeling for the inspection equipment on the transmission and distribution transformer line, and define the corresponding threat level based on the magnetic field distribution of the inspection equipment; Step 2: Generate a Voronoi diagram based on the location of the inspection equipment, and perform weighted processing on the Voronoi diagram based on the threat level information; Step 3: Improve the path planning algorithm and use the boundary intersection of the weighted Voronoi diagram as the extension node of the RRT algorithm. In the RRT algorithm extension step, select the Voronoi diagram point closest to the current node as the next node, so that the tree structure generated by the RRT algorithm is extended along the boundary of the Voronoi diagram. Step 4: Introduce Informed sampling based on the RRT algorithm, constrain the random points within the ellipse, and then combine the RRT* algorithm to reselect the parent node and rewire; Step 5: Use the triangle optimization method to optimize the redundant nodes of the planned path; Step 6: Smooth the generated path to further optimize the drone path.

2. According to claim 1, the method for UAV path planning for power transmission and distribution transformer based on the improved RRT algorithm combined with the Voronoi diagram is characterized by: In step 1, when inspecting the power system transmission and distribution lines or substations, it is necessary to pay attention to the operating status and potential faults of equipment such as transformers, switchgear, relay protection devices, cables and overhead lines, and mark each inspection equipment, obstacle, channel and important location on the inspection map; including the impact of visible physical obstacles on drone path planning, and the impact of invisible magnetic field size on drone flight path; collect the magnetic field size of each inspection equipment through sensors and historical data, and record it as the threat level of the inspection equipment.

3. According to claim 1, the method for UAV path planning based on the improved RRT algorithm combined with the Voronoi diagram is characterized by: In step 2, a weighted Voronoi diagram is used to reflect the difference in the impact of various obstacles on the drone. The improved Delaunay triangle side length calculation formula is: In formula (1), A, B, and C are three obstacles with different degrees of influence on the UAV; △ABC is a Delaunay triangle; L is the distance between A and B; T and T are the magnetic field magnitudes of A and B respectively; the positions of points F and E are determined in the same way; Connect D, E, and F respectively to obtain the segmented triangle △DEF, make an inscribed circle, and the lines OD, OE, and OF connecting the center of the circle are the edges of the weighted Voronoi diagram.

4. According to claim 1, the UAV path planning method for power transmission and distribution transformer based on the improved RRT algorithm combined with the Voronoi diagram is characterized by: In step 3, in the RRT algorithm expansion step, after generating potential nodes, all Voronoi nodes are extracted from the Voronoi diagram, and these Voronoi nodes are used as potential expansion nodes of the RRT tree to guide the growth of the tree; For the current node, the Voronoi diagram point closest to the current node is selected as the next node to be expanded, and this node is given priority, so that the tree structure generated by the RRT algorithm will be more inclined to expand along the boundary of the Voronoi diagram, thereby improving the safety of the path.

5. According to claim 4, the method for UAV path planning based on the improved RRT algorithm combined with the Voronoi diagram is characterized in that: Current node P c The closest Voronoi node P in the Voronoi v The distance between them is: In the above formula, d(P c , P v ) represents the previous node P c The closest Voronoi node P in the Voronoi v The distance between x v Represents the Voronoi node P v The horizontal axis of x c Represents the current node P c The horizontal axis of y v Represents the Voronoi node P v The vertical coordinate of y c Represents the current node P c The vertical coordinate of Traverse all nodes in the Voronoi diagram and calculate the value of the node P c The distance between the nodes is 1 and 2, and the node with the smallest distance is selected as the target Voronoi node P. nearest ; P new =(x c +Δd·v x ,y c +Δd·v y ) (7); In the above formula, v is the current node P c To the target Voronoi node P nearest The unit vector of x nearest Represents the distance to the current node P c The nearest Voronoi node P nearest The horizontal axis of y nearest Represents the distance to the current node P c The nearest Voronoi node P nearest The vertical coordinate of x c Represents the current node P c The horizontal axis of y c Represents the current node P c The vertical coordinate of v x represents the horizontal coordinate of the unit vector; v y represents the ordinate of the unit vector; Get the new node P according to the step size Δd new ; The new node P new Add to the RRT tree and add P current Set as the parent node; set the new node P new Set as the current node for the next expansion; Repeat the above steps until the target node is reached.

6. According to claim 1, the method for UAV path planning for power transmission and distribution transformer based on the combination of improved RRT algorithm and Voronoi diagram is characterized by: The step 4 includes improving the path planning Informed-RRT* algorithm, which is as follows: For the sampling area, a local ellipse sampling method is used instead of global uniform sampling. By shrinking the sampling area into an ellipse area, a large number of meaningless sampling points can be avoided. As the path length shortens, the ellipse sampling area gradually shrinks until the path optimization is completed: The ellipse area of ​​the Informed-RRT* algorithm takes the path starting point and the path target point as the focus, and the initial sampling ellipse equation is: In the above formula, a represents the length of the major semi-axis of the ellipse; b represents the length of the minor semi-axis of the ellipse; According to the rotation and translation matrix, the sampling point ellipse coordinates are converted to actual coordinates: [x' y'] = R · [xy] + T (3); Among them, x represents the actual X-axis coordinate after coordinate transformation; y' represents the actual Y-axis coordinate after coordinate transformation; rotation matrix Translation matrix T = [x c y c ], θ represents the angle between the line connecting the two foci of the ellipse and the actual coordinate system X; x c Indicates the amount of translation in the horizontal direction; y c Indicates the amount of translation in the vertical direction.

7. The method for UAV path planning for power transmission and distribution transformer based on the improved RRT algorithm combined with the Voronoi diagram according to claim 6 is characterized by: In step 4, the RRT* algorithm is combined to reselect the parent node and rewire, as follows: Reselect the parent node mainly with x new With x as the center, all adjacent nodes are found within a fixed radius. All nodes within the range will be used as backup parent nodes of the new node. The value of the path after each backup parent node is connected is calculated respectively. After comparing the path costs, the backup parent node with the smallest path cost is selected as the final parent node. The parent node is replaced by the original x nearest Become the new parent node x min ; After completing the reselection of the parent node, perform the rewiring operation and change x new As the parent nodes of the remaining nodes in the circle, and calculate each node with x new As parent node to x start Is the path cost of the new path less than the original path cost? If so, choose x new The parent node of this node, otherwise the child nodes are not updated.

8. According to claim 1, the method for UAV path planning for power transmission and distribution transformer based on the combination of improved RRT algorithm and Voronoi diagram is characterized by: In step 5, the triangle optimization method is used to optimize the redundant nodes of the planned path. The specific process is as follows: Name the nodes in the path [X star t, X1, X2...X k-1 , X k , X goal ], and change the starting point X start The length of the edge between node X1 and node X2 is recorded as b, and the length of the edge between node X2 and the starting point X is recorded as start The side length is recorded as c, forming a triangle; Find the first triangle node X in the initial path with side length c that collides with the obstacle k ; Node X k-1 With node X k The connection between them is discretely split into n segments; From the starting point X start Perform collision detection with discrete points P in turn and find the child node X of the first collision node in the collection P. stemp ; X stemp Update to the new starting node.

9. The method for UAV path planning for power transmission and distribution transformer based on the improved RRT algorithm combined with the Voronoi diagram according to claim 8 is characterized in that: The discrete split expression is as follows: P=(x k-1 ,and k-1 )+[(x k ,and k )-(x k-1 ,and k-1 )]×θ (x k-1 y k-1 )+(x k y k )·θ−(x k-1 y k-1 )·θ =(1-θ)·(x k-1 ,y k-1 )+θ·(x k ,y k ) (8) Among them, the value of θ is θ is the discrete step length, which is used to control the discrete splitting of the line segment into n segments; (x k-1 ,y k-1 )、(x k ,y k ) are the coordinates of the path nodes; P is the node set after the line segment is discretely split, which is used for collision detection with the starting point; Set the starting point X start Perform collision detection with discrete points in turn to find the child node of the first collision node in the collection P, which is X in the figure. stemp1 ; At this time, the path is X start -X1-X2-X3 updated to X start -X stemp1 , that is, the redundant nodes X1 and X2 are deleted; then X stemp1 Update to the new starting point and repeat the above steps until the target point X goal , the redundant nodes in the initial path can be deleted, the tortuosity of the path can be optimized, and the length of the path can be reduced; the path after triangle optimization is X start -X stemp1 -X stemp2 -X goal .

10. The method for UAV path planning for power transmission and distribution transformer based on the improved RRT algorithm combined with the Voronoi diagram according to claim 1 is characterized in that: In step 6, a cubic B-spline curve is used to smooth the path; specifically, the B-spline curve expression is: Where: P i Indicates the curve control point; F i，k (t) represents the k-order B-spline basis function; The cubic B-spline curve basis function expression is: Substituting equation (10) into equation (9), we finally get the cubic B-spline curve expression: P(t)=P0*F 0，3 (t)+P1*F 1，3 (t)+P2*F 2，3 (t)+P3*F 3，3 (t) (11) After the curve control points are given, a smooth cubic B-spline curve is obtained.

Images