Multi-source data fault operation and maintenance decision-making method and system based on point cloud model

Abstract

The invention discloses a multi-source data fault operation and maintenance decision-making method and system based on a point cloud model, and belongs to the technical field of intelligent power grid operation and maintenance management. Performing point cloud reconstruction on missing data in the point cloud model through virtual repair; carrying out physical mapping on the reconstructed point cloud model to identify an invisible fault; hierarchical operation and maintenance are driven by bidirectional sensitivity, and multi-objective optimization modeling is constructed to carry out power transmission maintenance optimization. According to the method, a closed loop of data acquisition, model reconstruction, fault location and decision execution is formed, and triple breakthrough of early screening of invisible faults, elastic allocation of resources and improvement of operation and maintenance economy is realized for the first time in the field of power transmission and transformation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of smart grid operation and maintenance management, and particularly relates to a multi-source data fault operation and maintenance decision method and system based on a point cloud model. BACKGROUND

[0002] With the evolution of smart grid construction towards high reliability, power distribution network fault diagnosis and operation and maintenance decision technology gradually develops from single data dimension to multi-source heterogeneous data fusion. The current mainstream scheme mainly relies on real-time monitoring of electrical quantity parameters by SCADA system, combines with infrared thermal imager to capture local temperature rise anomalies of equipment, or identifies mechanical structure loosening through vibration sensor network. In recent years, the popularity of laser radar technology has promoted the application of three-dimensional point cloud modeling in the digitalization of power equipment, realizing high-precision reconstruction of equipment space topology. In the data processing layer, point cloud segmentation algorithms based on deep learning have been used for component-level semantic labeling, and multi-sensor space-time registration technology has solved the pose synchronization problem in dynamic scanning through GPS / IMU combined navigation. In addition, the fault deduction model based on physical field provides a theoretical basis for hidden defect detection, and the operation and maintenance decision optimization also begins to introduce multi-objective programming method to balance the maintenance cost and power supply reliability.

[0003] Although the above-mentioned technology has made certain progress, there are still significant limitations in actual engineering application: first, the depth of multi-source data fusion is insufficient. The existing method does not effectively solve the point cloud distortion problem caused by equipment vibration, resulting in systematic deviation in the spatial mapping of LiDAR point cloud and vibration, acoustic data; at the same time, the auxiliary parameters such as environmental temperature and humidity and the equipment geometric model lack dynamic coupling mechanism, which restricts the accuracy of physical field calculation. Second, the data missing compensation ability is weak. Traditional point cloud reconstruction algorithm is difficult to adapt to the complex curved surface structure of power equipment, the identification of component damage area depends on artificial experience, and the reflectivity attribute of the reconstructed curved surface is inconsistent with the original point cloud, which affects the reliability of subsequent optical detection. Third, the hidden fault diagnosis model has poor generalization. The existing fault criterion is mostly based on single physical quantity threshold setting, without establishing a cross-domain correlation model of acoustic-thermal-vibration multi-modal signals, especially lacking probabilistic quantitative means for early warning of progressive defects such as micro-cracks. Fourth, the operation and maintenance decision is disconnected with resource constraints. The traditional hierarchical strategy only takes the device fault probability as a single dimension, ignoring the spatial propagation characteristics of the fault influence range; the maintenance optimization model mostly uses static resource allocation strategy, without considering the dynamic changes of spare parts inventory and work time window constraints, leading to the deviation of Pareto frontier solution from the actual operation and maintenance scene. The above defects jointly restrict the timeliness and economy of power distribution network fault disposal, and it is urgent to build a new generation of decision framework integrating high-precision three-dimensional perception, multi-physical field coupled diagnosis and dynamic resource optimization. SUMMARY

[0004] In view of the above-mentioned problems, the present application is proposed.

[0005] Therefore, the present application aims to solve the problem that power distribution network equipment is difficult to accurately and efficiently identify potential faults (especially hidden faults) and scientifically formulate optimal operation and maintenance decisions under complex operating environments (vibration, temperature change, shielding, etc.).

[0006] To solve the above technical problems, the present application provides the following technical solutions: a multi-source data fault operation and maintenance decision method based on a point cloud model, comprising,

[0007] Collecting power distribution network point cloud data and constructing a power distribution network point cloud model; missing data in the point cloud model is reconstructed by virtual repair; the reconstructed point cloud model is physically mapped to identify hidden faults; two-way sensitivity is used to drive hierarchical operation and maintenance, and multi-objective optimization modeling is constructed to optimize power transmission maintenance.

[0008] As a preferred scheme of the multi-source data fault operation and maintenance decision method based on the point cloud model, the construction of the power distribution network point cloud model includes collecting sound pressure signals through acoustic sensors and collecting environmental temperature data.

[0009] Deploy a rotating LiDAR to scan power distribution network equipment and obtain an original point cloud set;

[0010] Synchronously start high-precision IMU to collect vibration data including displacement and rotation;

[0011] Add a timestamp t to each point cloud coordinate through a GPS clock synchronizer i Perform spatiotemporal alignment processing to establish a point cloud vibration data mapping table;

[0012] Construct a vibration compensation matrix to compensate for point cloud distortion caused by equipment vibration and obtain compensated point clouds;

[0013] Input the compensated point clouds into a pre-trained PointNet++ network to output a set of device component class labels to which each point belongs, match the spatial position of the point cloud with a power grid topology database, add electrical properties to each point, and output a semantic point cloud model.

[0014] As a preferred scheme of the multi-source data fault operation and maintenance decision method based on the point cloud model, the virtual repair for point cloud reconstruction includes,

[0015] Input the semantic point cloud model, calculate the covariance matrix of the neighborhood points for each point cloud through principal component analysis, and calculate the eigenvalues to calculate the curvature. Set K points in the neighborhood, calculate the local curvature entropy according to the curvature set, and determine that the current point cloud belongs to a damaged area when the curvature entropy is greater than 0.8, and mark it as a hollow point.

[0016] Collect all the hollow points into a hollow region, retrieve the device standard CAD library according to the device ID, call the complete geometric model S of the corresponding component CAD , extract the hollow boundary point set, and use the Laplace equation to edit the surface with the hollow boundary point set as the constraint condition to obtain the reconstructed surface S' CAD And strictly fit the hollow boundary point set, calculate the average reflectivity of the boundary region, and assign the average reflectivity to the newly generated surface S" CAD , and fuse the reconstructed surface with the average reflectivity with the original model to generate a complete point cloud model.

[0017] As a preferred scheme of the multi-source data fault operation and maintenance decision method based on the point cloud model, wherein the physical mapping includes,

[0018] Based on the complete point cloud model, calculate the geodesic distance matrix of the point cloud model, and reduce the dimension to two-dimensional UV plane through multidimensional scaling to obtain the mapping coordinates of each point cloud;

[0019] Based on the temperature data, construct a temperature field T map , calculate the vibration gradient field G v on the UV plane of each mapping coordinate, and formulate a composite fault criterion:

[0020]

[0021] Wherein, η is the fault threshold.

[0022] As a preferred scheme of the multi-source data fault operation and maintenance decision method based on the point cloud model, wherein the identification of the invisible fault includes, based on the fault criterion and the sound pressure signal collected by the acoustic sensor, wavelet packet transform is performed on the sound pressure signal to obtain sub-band features, target band features are extracted, the energy of the target band is calculated, the crack probability is calculated based on the energy of the target band, and the fault mapping report is output according to the result.

[0023] As a preferred scheme of the multi-source data fault operation and maintenance decision method based on the point cloud model, wherein the two-way sensitivity driven hierarchical operation and maintenance includes, based on the fault report, calculating the fault probability for any device, assigning weights to the crack probability and the composite fault according to the 6:4 weight distribution standard respectively, obtaining the summation result as the fault probability of the current device, calculating the fault influence range of the current device according to the fault probability, taking the fault probability and the fault influence range of the current device as the sensitivity index matrix, and driving the hierarchical operation and maintenance;

[0024] Two-way clustering is performed through the improved DBSCAN algorithm, the resource constraint factor γ is defined by the proportion of currently available operation and maintenance personnel, and the resource constraint radius is calculated;

[0025] Based on the sensitivity index of each device point, the number of points in the corresponding neighborhood is calculated, if the number is greater than a preset threshold, it is marked as a core point, otherwise it is marked as a noise point, and the region connected with the core point density is divided into clusters;

[0026] For each device point s i = (P f (i), I o (i)), where P f (i) is the failure probability of the current device, I o (i) is the failure influence range of the current device, and the neighborhood density of the resource constraint radius ∈ is calculated Where Σ is the number of all devices, |s e -s f | 2 represents the Euclidean distance between device e and device f in the two-dimensional sensitivity space;

[0027] According to the neighborhood density, the output is classified: when ρ i >ρ min and P f (i) > the first failure probability threshold, it is marked as red zone level; when ρ i >ρ min and the second failure probability threshold < P f (i) ≤ the first failure probability threshold, it is marked as yellow zone level; when ρ i ≤ρ min , it is marked as green zone level; power transmission maintenance optimization is carried out for different operation and maintenance partitions; ρ min is the neighborhood density threshold.

[0028] As a preferred scheme of the multi-source data fault operation and maintenance decision method based on the point cloud model, wherein: the power transmission maintenance optimization based on the multi-objective optimization modeling includes: for the green zone level, it is determined that the risk is low, and the distribution network is not affected, and no maintenance is performed, and daily operation and maintenance is ensured;

[0029] For the yellow zone level, it is determined that there are multiple high failure probabilities, which seriously affect the distribution network, and a fault warning of immediately performing maintenance is issued;

[0030] For the red zone level, it is determined that there is a single high-risk device, and a fault warning of maintenance within 24 hours is issued;

[0031] For the yellow zone level and the red zone level, the maintenance cost C main minimization and power supply reliability R reliability maximization are the optimization objectives:

[0032]

[0033] wherein c labor is the unit labor cost, t i is the maintenance labor hours of the ith device, c part is the unit spare part cost, q i is the number of spare parts required for the ith device, T outage is the estimated power outage duration of the jth region, ω j is the power outage influence weight of the jth region, TotalLoad is the total load of the system, N is the total number of devices under maintenance, i is the index of the device under maintenance, J is the number of affected user groups, and j is the index of the affected user group.

[0034] The constraint conditions are set as follows:

[0035] Constraint condition one is that all power devices classified as red zones must be included in the maintenance scope, and no omission is allowed.

[0036] Constraint condition two is that the total time consumption of all maintenance tasks cannot exceed the maximum time window allowed for safe operation of the power grid.

[0037] Constraint condition three is that the types and quantities of spare parts consumed by the maintenance scheme cannot exceed the real-time inventory.

[0038] The branch and bound method is used to solve the integer programming to generate a Pareto optimal solution set, and the optimal maintenance strategy is selected as the decision for fault operation and maintenance.

[0039] Another object of the present application is to provide a multi-source data fault operation and maintenance decision system based on a point cloud model.

[0040] To solve the above technical problems, the present application provides the following technical solutions: a multi-source data fault operation and maintenance decision system based on a point cloud model, comprising: a data acquisition module, a point cloud reconstruction module, a fault identification module, and a hierarchical operation and maintenance module.

[0041] The data acquisition module acquires point cloud data of the distribution network and constructs a point cloud model of the distribution network.

[0042] The point cloud reconstruction module reconstructs the point cloud by virtual repair for missing data.

[0043] The fault identification module performs physical mapping on the reconstructed point cloud model to identify hidden faults.

[0044] The hierarchical operation and maintenance module drives hierarchical operation and maintenance with bidirectional sensitivity, and constructs a multi-objective optimization model for transmission maintenance optimization.

[0045] The application provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements steps of the multi-source data fault operation and maintenance decision-making method based on a point cloud model when executing the computer program.

[0046] The application provides a computer readable storage medium, which stores a computer program, and wherein the computer program implements steps of the multi-source data fault operation and maintenance decision-making method based on a point cloud model when executed by a processor.

[0047] The application has the following beneficial effects: the application realizes millimeter-level spatial calibration of acoustic / temperature / vibration / point cloud data and integrated expression of device physical-electrical properties through multi-source space-time alignment and semantic modeling, solves the problem that a traditional static point cloud cannot reflect dynamic deformation of a device, and achieves the core effect of constructing a computable fault analysis base.

[0048] The application realizes accurate positioning of a damaged area and reconstruction of a physically real surface through curvature entropy driven adaptive reconstruction, solves the problem of feature smoothing distortion caused by a traditional interpolation method, and achieves the key effect of ensuring the accuracy of subsequent physical field calculation.

[0049] The application realizes high-dimensional fault feature decoupling and multi-sensor evidence chain fusion through geodesic mapping multi-physical field coupling diagnosis, solves the industry pain point of missing hidden faults caused by a single detection method, and achieves the breakthrough effect of improving the microscopic crack detection rate to 92%.

[0050] The application realizes dynamic adjustment of operation and maintenance partition according to manpower and Pareto optimization of maintenance strategies through bidirectional sensitivity dynamic hierarchical optimization, solves the problems of resource mismatch and response lag caused by a traditional fixed threshold, and achieves the dual optimization effects of 24-hour rapid response of high-risk equipment and 30% reduction of inventory cost. BRIEF DESCRIPTION OF DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0052] Figure 1 A multi-source data fault operation and maintenance decision-making method based on a point cloud model is provided for an embodiment of the application. DETAILED DESCRIPTION

[0053] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0054] Example 1, referring to Figure 1 This is one embodiment of the present invention, which provides a multi-source data fault operation and maintenance decision-making method based on a point cloud model, including:

[0055] S1. Collect point cloud data of the power distribution network and construct a point cloud model of the power distribution network.

[0056] It should be noted that the sound pressure signal {p} collected by the acoustic sensor m (t)};

[0057] Collect ambient temperature data T s ;

[0058] Deploy rotating LiDAR scanning equipment to obtain raw point cloud datasets.

[0059] Simultaneously start a high-precision IMU to acquire vibration data, including displacement (Δx, Δy, Δz) and rotation (θ). x ,θ y ,θ z );

[0060] A timestamp t is added to each point cloud coordinate using a GPS clock synchronizer. i Spatiotemporal alignment processing is performed to establish a point cloud vibration data mapping table Φ:(x i ,y i ,z i ,t i )→(Δx i ,Δy i ,Δz i ,θ xi ,θ yi ,θ zi );

[0061] Construct the vibration compensation matrix M comp :

[0062]

[0063] Compensation for point cloud distortion caused by equipment vibration: P corrected =M comp -1 ·P raw, the compensated point cloud P c = (x' i , y' i , z' i ) is obtained.

[0064] The compensated point cloud P c is input into a pre-trained PointNet++ network, and a device component class label set {l i} (such as an insulator, a conductor, and a clamp) to which each point belongs is output. According to the spatial position of the point cloud, the point cloud is matched with a power grid topology database, and electrical properties, including a device ID (DeviceID i ) and a rated voltage (Voltage i ), are attached to each point. A semantic point cloud model M sem = (x' i , y' i , z' i , l i , DeviceID i , Voltage i ) is output.

[0065] S2, the missing data in the point cloud model is virtually repaired to reconstruct the point cloud.

[0066] It should be noted that the input semantic point cloud model M sem , for each point cloud, the curvature feature of the local neighborhood is calculated. Specifically, the covariance matrix of the neighborhood points is calculated by principal component analysis, and the eigenvalues λ0≤λ1≤λ2 are calculated. The curvature c i = λ0 / (λ0+λ1+λ2) is calculated.

[0067] Suppose there are K points in the neighborhood, and the local curvature entropy H is calculated according to the curvature set When H i >0.8, it is determined that the current point cloud belongs to a damaged area, and is marked as a hollow point.

[0068] All hollow point sets are set as a hollow region D, and the DeviceID i is used to search a device standard CAD library to retrieve a complete geometric model S CAD of the corresponding component. The hollow boundary point set B is extracted, and Laplace equation is used for surface editing with the hollow boundary point set B as a constraint condition to obtain a reconstructed surface S' CAD which strictly fits the hollow boundary point set. The average reflectivity of the boundary region is calculated, and the average reflectivity is assigned to the newly generated surface S" CAD . The reconstructed surface with the average reflectivity is fused with the original model to generate a complete point cloud model M complete =Msem ∪S″ CAD .

[0069] S3, performing physical mapping on the reconstructed point cloud model to identify hidden faults.

[0070] Specifically, based on the complete point cloud model M complete , the geodesic distance matrix D g of the point cloud model is calculated g (p i ,p j ) and is reduced to a two-dimensional UV plane by multidimensional scaling (MDS) to obtain the mapping coordinates (u i ,v i ) of each point cloud;

[0071] Based on the temperature data, a temperature field T map (u,v) is constructed:

[0072]

[0073] Where d g is the geodesic distance, s s is the sensor projection position, sigma is the Gaussian kernel width, N s is the total number of sensors, s is the variable index of the s-th sensor, p i and p j are the coordinates of the i-th and j-th spatial points in the point cloud model, i.e. p i =(x' i ,y' i ,z' i );

[0074] On the UV plane of each mapping coordinate, the vibration gradient field is calculated

[0075]

[0076] Where eta is the fault threshold.

[0077] It should be noted that based on the fault criterion and the sound pressure signal collected by the acoustic sensor, the sound pressure signal is subjected to wavelet packet transform to obtain sub-band features, extract target frequency band features W(t), and calculate the energy of the target frequency band Where H0 and Hz are the upper and lower limits of the target frequency band, for example: 5 Hz to 8 Hz;

[0078] Based on the energy of the target frequency band, the crack probability P crack is calculated:

[0079]

[0080] wherein μ E and σ E are the historical energy mean and standard deviation of the normal equipment target frequency band, respectively, and a is a fault sensitivity coefficient calibrated by historical fault data of the equipment;

[0081] outputting a fault mapping report according to the result

[0082] S4, driving hierarchical operation and maintenance with bidirectional sensitivity, and constructing a multi-objective optimization model for transmission maintenance optimization.

[0083] It should be noted that, based on the fault report For any equipment, the fault probability is calculated, the crack probability and the compound fault are allocated weights respectively according to the weight allocation standard of 6:4, the sum of the two is obtained as the fault probability of the current equipment, the total downstream load caused by the fault of the current equipment is calculated as the fault impact range of the current equipment according to the fault probability, the fault probability and the fault impact range of the current equipment are taken as the sensitivity index matrix, and the hierarchical operation and maintenance is driven;

[0084] Two-way clustering is performed through the improved DBSCAN algorithm, the resource constraint factor γ is defined according to the proportion of currently available operation and maintenance personnel, the resource constraint radius ∈ = ∈0·(0.3+0.7γ) is calculated, and ∈0 is the initial constraint radius;

[0085] Based on the sensitivity index of each equipment point, the number of points in the corresponding neighborhood is calculated, if the number is greater than a preset threshold, the point is marked as a core point, otherwise it is marked as a noise point, and the region connected with the core point density is divided into a cluster;

[0086] For each equipment point s i = (P f (i), I o (i)), wherein P f (i) is the fault probability of the current equipment, I o (i) is the fault impact range of the current equipment, and the neighborhood density of the resource constraint radius ∈ is calculated wherein Σ is the number of all equipment, |s e -s f | 2 represents the Euclidean distance between equipment e and equipment f in the two-dimensional sensitivity space;

[0087] Hierarchical output is performed according to the neighborhood density: when p i > p min and P f (i) > the first fault probability threshold, it is marked as the red zone level; when p i > p min and the second fault probability threshold < Pf (i) ≤ first failure probability threshold, marked as yellow zone level; when p i ≤ p min , marked as green zone level; power transmission maintenance optimization is performed for different operation and maintenance partitions; p min is a neighborhood density threshold.

[0088] It should be noted that the green zone level is identified as low risk, which does not affect the distribution network and does not perform maintenance to ensure daily operation and maintenance;

[0089] The yellow zone level is identified as a high failure probability, which seriously affects the distribution network, and a fault warning is issued to immediately perform maintenance;

[0090] The red zone level is identified as a single high-risk device, and a fault warning is issued to perform maintenance within 24 hours;

[0091] For the yellow zone level and the red zone level, the maintenance cost C main minimization and power supply reliability R reliability maximization are optimization objectives:

[0092]

[0093] Where c labor is the unit labor cost, t i is the maintenance time of the i-th device, c part is the unit spare part cost, q i is the number of spare parts required for the i-th device, T outage is the estimated power outage time of the j-th region, ω j is the power outage influence weight of the j-th region, TotalLoad is the system total load, N is the total number of maintenance devices, i is the maintenance device index, J is the number of affected user groups, and j is the affected user group index.

[0094] Constraint conditions are set as follows:

[0095] ① All power devices classified as red zone must be included in the maintenance range, and any omission is not allowed.

[0096] ② Time window constraint: the total time consumption of all maintenance tasks must not exceed the maximum time window allowed for safe operation of the power grid;

[0097] Specifically, the daily maintainable period is obtained, the expected time consumption of each maintenance device is accumulated, when the total time consumption approaches the upper limit of the window, the addition of yellow zone devices is automatically suspended, and the maintenance sequence is rearranged.

[0098] ③ Spare parts matching: the type and quantity of spare parts consumed by the maintenance scheme shall not exceed the real-time inventory, when the inventory is insufficient, whether the spare parts can be replaced is found, if the replaceable spare parts exist, the spare parts model is automatically replaced, if the replaceable spare parts do not exist, the maintenance scheme is reduced, and the maintenance is replaced by replacement.

[0099] The branch and bound method is used to solve the integer programming, a Pareto optimal solution set is generated, an optimal maintenance strategy is selected, and the optimal maintenance strategy is used as a decision for fault operation and maintenance.

[0100] Embodiment 2 is an embodiment of the present application, and the above is a schematic scheme of a multi-source data fault operation and maintenance decision method based on a point cloud model. It should be noted that the technical scheme of a multi-source data fault operation and maintenance decision system based on a point cloud model and the technical scheme of the above multi-source data fault operation and maintenance decision method based on a point cloud model belong to the same concept. The technical scheme of the multi-source data fault operation and maintenance decision system based on a point cloud model in this embodiment is not described in detail, and the description of the technical scheme of the multi-source data fault operation and maintenance decision method based on a point cloud model is referred to.

[0101] The embodiment provides a multi-source data fault operation and maintenance decision system based on a point cloud model, which comprises a data acquisition module, a point cloud reconstruction module, a fault identification module and a hierarchical operation and maintenance module.

[0102] The data acquisition module acquires point cloud data of a power distribution network and constructs a point cloud model of the power distribution network.

[0103] The point cloud reconstruction module reconstructs the point cloud model by virtual repair for missing data.

[0104] The fault identification module performs physical mapping on the reconstructed point cloud model to identify hidden faults.

[0105] The hierarchical operation and maintenance module drives hierarchical operation and maintenance by bidirectional sensitivity, constructs a multi-objective optimization model, and optimizes power transmission maintenance.

[0106] The embodiment also provides an electronic device suitable for a multi-source data fault operation and maintenance decision method based on a point cloud model, which comprises a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the multi-source data fault operation and maintenance decision method based on a point cloud model as described in the above embodiment.

[0107] The embodiment also provides a storage medium having a computer program stored thereon, and the program is executed by a processor to realize the multi-source data fault operation and maintenance decision method based on a point cloud model as described in the above embodiment.

[0108] The storage medium proposed in the embodiment belongs to the same inventive concept as the multi-source data fault operation and maintenance decision method based on a point cloud model proposed in the above embodiment. Technical details not described in the embodiment can be seen from the above embodiment, and the embodiment has the same beneficial effects as the above embodiment.

[0109] Through the above description of the embodiments, those skilled in the art can clearly understand that the present application can be realized by means of software and necessary general hardware, and of course can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer readable storage medium, such as a floppy disk, a read-only memory (ROM), a random access memory (RAM), a FLASH, a hard disk or an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the method of each embodiment of the present application.

[0110] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. A method for multi-source data fault operation and maintenance decision based on a point cloud model, characterized in that: The application relates to a power distribution network fault diagnosis method and device. Point cloud data of a power distribution network is collected and a point cloud model of the power distribution network is constructed; Missing data in the point cloud model is reconstructed through virtual repair; Physical mapping is performed on the reconstructed point cloud model to identify hidden faults; Two-way sensitivity is used to drive hierarchical operation and maintenance, and multi-objective optimization modeling is constructed to optimize power transmission maintenance.

2. The multi-source data fault operation and maintenance decision method based on a point cloud model according to claim 1, characterized in that: The construction of the point cloud model of the power distribution network comprises the following steps: collecting sound pressure signals through an acoustic sensor, collecting environmental temperature data, deploying a rotating LiDAR to scan power distribution network equipment, obtaining an original point cloud set, synchronously starting high-precision IMU to collect vibration data including displacement and rotation, constructing a vibration compensation matrix to compensate for point cloud distortion caused by equipment vibration, and obtaining compensated point cloud. The compensated point cloud is input into a pre-trained PointNet++ network, and a device component category label set to which each point belongs is output; the spatial position of the point cloud is matched with a power grid topology database, and electrical properties are added to each point, and a semantic point cloud model is output. The point cloud reconstruction through virtual repair comprises the following steps: Each point cloud coordinate is time-stamped t by a GPS clock synchronizer i A spatio-temporal alignment process is performed to establish a point cloud vibration data mapping table; The semantic point cloud model is input, the covariance matrix of the neighborhood points is calculated through principal component analysis for each point cloud, the eigenvalues are calculated, the curvature is calculated, K points in the neighborhood are set, the local curvature entropy is calculated according to the curvature set, and when the curvature entropy is greater than 0.8, it is determined that the current point cloud belongs to a missing area, and the missing area is marked as a hollow point. The physical mapping comprises the following steps:

3. The multi-source data fault operation and maintenance decision method based on a point cloud model according to claim 2, characterized in that: Based on the complete point cloud model, the geodesic distance matrix of the point cloud model is calculated, and multi-dimensional scaling is used to reduce the dimension to a two-dimensional UV plane to obtain the mapping coordinates of each point cloud. Wherein, eta is a fault threshold. All the hollow points are set as a hollow region, the device standard CAD library is retrieved according to the device ID, and the complete geometric model S of the corresponding component is called CAD The hollow boundary point set is extracted, and the Laplace equation is used for surface editing with the hollow boundary point set as the constraint condition to obtain the reconstructed surface S ' CAD The average reflectivity of the boundary region is calculated, and the average reflectivity is assigned to the newly generated surface S" CAD The reconstructed surface with the average reflectivity is fused with the original model to generate a complete point cloud model.

4. The multi-source data fault operation and maintenance decision method based on a point cloud model according to claim 3, characterized in that: The identification of hidden faults comprises the following steps: based on the fault criterion and the sound pressure signals collected by the acoustic sensor, wavelet packet transformation is performed on the sound pressure signals to obtain sub-band features, target frequency band features are extracted, The energy of the target frequency band is calculated, the crack probability is calculated based on the energy of the target frequency band, and a fault mapping report is output according to the result. constructing a temperature field T based on the temperature data map on the UV plane of each mapping coordinate, a vibration gradient field G is calculated v formulating a compound fault criterion: The two-way sensitivity is used to drive hierarchical operation and maintenance, and multi-objective optimization modeling is constructed to optimize power transmission maintenance.

5. The multi-source data fault operation and maintenance decision method based on a point cloud model according to claim 4, characterized in that: Through the improved DBSCAN algorithm, two-way clustering is performed, the current available operation and maintenance personnel ratio is used to define the resource constraint factor gamma, and the resource constraint radius is calculated; Based on the sensitivity index of each device point, the number of points in the corresponding neighborhood is calculated, if the number is greater than a preset threshold, the point is marked as a core point, otherwise, the point is marked as a noise point, and the regions with the core point density connected are divided into clusters; 6. The multi-source data fault operation and maintenance decision method based on a point cloud model according to claim 5, characterized in that: The multi-objective optimization modeling for power transmission maintenance optimization comprises the following steps: for the green area, it is identified as low risk, does not cause impact on the power distribution network, and does not perform maintenance to ensure daily operation and maintenance; For the yellow area, it is identified as a high fault probability, causes serious impact on the power distribution network, and issues a fault warning for immediate maintenance; For the red area, it is identified as a single high-risk device, and a fault warning for maintenance within 24 hours is issued; for each device point s i = (P f (i), I o (i)), where P f (i) is the failure probability of the current device, I o (i) is the failure impact range of the current device, and the neighborhood density of the computational resource constraint radius ∈ where Σ is the number of all devices, and |s e -s f | 2 denotes the Euclidean distance between device e and device f in the two-dimensional sensitivity space. According to the neighborhood density, the output is classified: when ρ i >ρ min and P f (i) > the first failure probability threshold, it is marked as the red zone level; when ρ i >ρ min and P f (i) ≤ the first failure probability threshold, it is marked as the yellow zone level; when ρ i ≤ρ min , it is marked as the green zone level; power transmission maintenance optimization is performed for different operation and maintenance partitions; and ρ min is a neighborhood density threshold.

7. The multi-source data fault operation and maintenance decision method based on a point cloud model according to claim 6, characterized in that: The constraint condition is set as follows: ​ ​ For the yellow and red zones, the maintenance cost C main minimization and power supply reliability R reliability maximization is the optimization objective: where c labor is the unit labor cost, t i is the maintenance labor hours for the ith device, c part is the unit spare part cost, q i is the number of spare parts required for the ith device, T outage is the estimated outage duration for the jth region, ω j is the outage impact weight for the jth region, TotalLoad is the system total load, N is the total number of devices under maintenance, i is the index of the device under maintenance, J is the number of affected user groups, and j is the index of the affected user group. ​ The first constraint is that all power equipment classified as red zone must be included in the maintenance scope, and no omission is allowed; The second constraint is that the total time consumption of all maintenance tasks must not exceed the maximum time window allowed for safe operation of the power grid; The third constraint is that the types and quantities of spare parts consumed by the maintenance scheme must not exceed the real-time inventory; The branch and bound method is used to solve the integer programming, generate a Pareto optimal solution set, select an optimal maintenance strategy, and implement the optimal maintenance strategy as the decision of fault operation and maintenance.

8. A multi-source data fault operation and maintenance decision system based on a point cloud model, applying a multi-source data fault operation and maintenance decision method based on a point cloud model according to any one of claims 1-7, characterized in that, It includes: a data acquisition module, a point cloud reconstruction module, a fault identification module, and a hierarchical operation and maintenance module; The data acquisition module acquires point cloud data of the distribution network and constructs a point cloud model of the distribution network; The point cloud reconstruction module reconstructs the point cloud by virtual repair for missing data based on the point cloud model; The fault identification module identifies hidden faults by physically mapping the reconstructed point cloud model; The hierarchical operation and maintenance module drives hierarchical operation and maintenance based on bidirectional sensitivity, constructs a multi-objective optimization model, and optimizes power transmission maintenance. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor executes the computer program to realize the steps of the point cloud model-based multi-source data fault operation and maintenance decision method of any one of claims 1-7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the point cloud model-based multi-source data fault operation and maintenance decision method of any one of claims 1-7.

Images