Distribution network overhead line detection system

Abstract

The invention relates to the technical field of line detection, and discloses a distribution network overhead line detection system and method.The system comprises an intelligent denoising module, a composite positioning module, a digital twinning construction module, a decision module, an unmanned aerial vehicle inspection module and a data analysis module; performing multi-physics field coupling analysis in combination with physics field data to obtain electric heating power distribution data, further obtaining a fault site, constructing a digital twin model, mapping the electric heating power distribution data to the digital twin model, visualizing the fault site, further generating an optimal decision, and updating the inspection route of the unmanned aerial vehicle based on the optimal decision. And generating a detection report based on a health index algorithm. According to the invention, the detection accuracy of the distribution network overhead line can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of line detection, and particularly relates to a distribution network overhead line detection system and method. BACKGROUND

[0002] As an important part of power transmission, the running state of the distribution network overhead line directly relates to the reliability and safety of the power grid.

[0003] However, the traditional overhead line detection method has serious data noise interference, and the overhead line is easily affected by electromagnetic interference, environmental noise and other factors during operation, resulting in low signal-to-noise ratio of the collected power analysis data, affecting the accuracy of fault diagnosis, and relying on single physical field analysis, lacking coupling modeling of multiple physical fields such as electricity, heat and force, and being difficult to accurately locate the composite fault site. The unmanned aerial vehicle inspection route is fixed and cannot be dynamically optimized according to the real-time environment and fault risk, resulting in incomplete inspection coverage or resource waste and insufficient data integration and visualization, making it difficult to quickly generate a comprehensive health assessment report and affecting the operation and maintenance response speed.

[0004] Therefore, how to integrate intelligent denoising and multi-physical field coupling analysis to improve the accuracy, real-time performance and practicality of fault detection has become a problem to be solved. SUMMARY

[0005] The present application provides a distribution network overhead line detection system and method, which mainly aims to solve the problem of low efficiency during self-service cashing.

[0006] To achieve the above purpose, the present application provides a distribution network overhead line detection system, characterized in that the system comprises an information module, wherein: A distribution network overhead line detection system, the system comprises an intelligent denoising module, a composite positioning module, a digital twin construction module, a decision module, an unmanned aerial vehicle inspection module and a data analysis module, wherein: The intelligent denoising module is used for obtaining power analysis data of the overhead line based on a preset initial route, and performing denoising on the power analysis data through an adaptive filtering algorithm to obtain accurate analysis data of the overhead line. The composite positioning module is used for performing multi-physical field coupling analysis on the accurate analysis data and the physical field data of the overhead line to obtain the electric-thermal-force distribution data of the overhead line, and positioning the fault site in the overhead line based on the electric-thermal-force distribution data. The digital twin construction module is used for constructing a digital twin model of the overhead line, and mapping the fault site and the electric-thermal-force distribution data to the digital twin model. The decision module is configured to construct a knowledge graph of the overhead line based on the perception data of the overhead line, infer the electric-thermal distribution data based on the knowledge graph to obtain a fault type of the overhead line, and generate an optimal decision of the overhead line by using a reinforcement learning algorithm. The unmanned aerial vehicle inspection module is configured to dynamically optimize the initial flight path based on the optimal decision and real-time environmental information to obtain an inspection flight path of the unmanned aerial vehicle, and feed back the inspection flight path to the intelligent denoising module. The data analysis module is configured to visualize the fault site and generate a detection report of the overhead line based on a health index algorithm.

[0007] In a preferred embodiment, the adaptive filtering algorithm is as follows:

[0008] In the formula, is an output value of the adaptive filter, is a transpose of a weight coefficient vector, is an input signal vector of the adaptive filter at the n-th moment, is an i-th element in the weight coefficient vector of the adaptive filter, is a value of the input signal vector of the adaptive filter at the n-th moment, wherein the input signal is power analysis data, and the power analysis data includes three-phase current waveforms, color heat maps, and conductor sag data. The error signal of the adaptive filtering algorithm and the gain vector of the adaptive filtering algorithm are used to update the weight coefficient vector of the adaptive filter.

[0009] In a preferred embodiment, updating the weight coefficient vector of the adaptive filter based on the error signal of the adaptive filtering algorithm and the gain vector of the adaptive filtering algorithm includes: The expression of the error signal of the adaptive filtering algorithm is as follows:

[0010] In the formula, is an error signal, is an expected signal, is an output value of the adaptive filter; The expression of the gain vector of the adaptive filtering algorithm is as follows:

[0011] In the formula, is an inverse correlation matrix, is a forgetting factor, is a gain vector, a transpose of an input signal vector of the adaptive filter, a weight coefficient vector of the adaptive filter, an error signal, an input signal vector of the adaptive filter at an n th moment; updating the weight coefficient vector based on a weight coefficient updating formula, the error signal and the gain vector, wherein an expression of the weight coefficient updating formula is as follows:

[0012] wherein, an updated weight coefficient vector, a weight coefficient vector, a gain vector, an error signal.

[0013] In a preferred embodiment, the composite positioning module performs a multi-physical field coupling analysis on the precision analysis data and the physical field data of the overhead line to obtain the electro-thermal power distribution data of the overhead line, and locates the fault site in the overhead line based on the electro-thermal power distribution data, specifically for: The specific process of the multi-physical field coupling analysis is as follows: substituting the precision analysis data and the physical field data of the overhead line into a multi-physical field coupling equation to solve, to obtain a coupling analysis result of the multi-physical field coupling analysis, wherein an expression of the multi-physical field coupling equation is as follows:

[0014] wherein, a dielectric constant, a charge density, a gradient operator, an electric potential, an electric current intensity, an electric resistance, an electric voltage, heat generated by an electric field, a thermal melting ratio, a conductor temperature change amount, a time change amount, a conductor temperature, a thermal conductivity, a convective heat transfer coefficient, an ambient temperature; generating three-dimensional temperature field distribution data of the overhead line based on the coupling analysis result to obtain the electro-thermal power distribution data of the overhead line; A three-dimensional physical field model of the overhead line is constructed based on the electrothermal distribution data. When a point in the three-dimensional physical field model that exceeds the preset maximum allowable temperature threshold of the overhead line is detected, it is recorded as a fault point.

[0015] In a preferred embodiment, the digital twin construction module, when constructing a digital twin model of the overhead line, is specifically used for: A digital model of the overhead line is obtained, and the digital model is spatially mapped to the overhead line to obtain a digital twin model of the overhead line. The fault location and the electrothermal distribution data are then mapped to the digital twin model.

[0016] In a preferred embodiment, the decision module, when performing reasoning based on the knowledge graph on the electrothermal distribution data to determine the fault type of the overhead line, specifically performs the following: The sensing data of the overhead line is processed for data integrity and data feature construction to obtain the feature data of the overhead line. A knowledge graph is then constructed based on the feature data. The method for constructing the knowledge graph is as follows: The equipment in the overhead line is defined as an entity, and the relationship between the entities is constructed to obtain the equipment association relationship in the overhead line. The entity and the equipment association relationship are stored in the database to obtain a knowledge graph. Based on the knowledge graph, triplet confidence inference is performed on the electrothermal distribution data to obtain the fault type. The expression for the triplet confidence score in the triplet confidence inference is as follows:

[0017] In the formula, The confidence score for the triples of the knowledge graph. This is the head entity of the knowledge graph. For the aforementioned relationship, This is the tail entity of the knowledge graph. This is a low-dimensional vector of the head entity. The head entity vector of the knowledge graph transpose, For the tail entity, (This is a low-dimensional vector.) This is a matrix representing the aforementioned relationships; The formula for converting the confidence score of the triplet is as follows:

[0018] In the formula, This is the probability transformation result of the confidence score of the triplet. For the confidence score of the triplets of the knowledge graph, a fault type reasoning function is constructed based on the probability conversion result, reasoning is performed based on the fault type reasoning function to obtain a fault type:

[0019] In the formula, is the fault type of the final reasoning, is a preset fault type set, is a set of triplets in the knowledge graph associated with the current electric-thermal power distribution data, is a probability conversion result of the confidence score of the triplets, is a fault type weight function, is a fault type weight function, represents an entity in the historical data due to the fault type occurred, wherein the mathematical expression of the fault type weight function is as follows:

[0020] In the formula, is a fault type weight function, is an entity in the historical data due to the fault type occurred, is a preset fault type set.

[0021] In a preferred embodiment, the perception data of the overhead line is subjected to data integrity processing and data feature construction processing, and the specific method is as follows: S701, missing value prediction is performed on the perception data to obtain missing values of the perception data, and the missing values are filled to obtain one-stage complete data of the perception data; S702, the Z-score of each data in the one-stage complete data is calculated, the perception data with an absolute value of the Z-score greater than 3 is recorded as an abnormal value, and the abnormal value is deleted to obtain two-stage complete data of the perception data, wherein the Z-score is a deviation degree of data from the mean value; S703, the two-stage complete data is subjected to repeatability screening through a similarity matching algorithm to obtain complete data of the perception data; S704, feature extraction is performed on the complete data through a statistical feature construction method to obtain feature data of the perception data.

[0022] In a preferred embodiment, the decision-making module, in executing and utilizing a reinforcement learning algorithm to generate the optimal decision for the overhead line, specifically performs the following: A reinforcement learning model is constructed based on the probability transformation results; The reinforcement learning algorithm in the reinforcement learning model is as follows:

[0023] In the formula, For instant reward function, To reduce real-time processing costs, The preset cost threshold, For safe operating time, For the monitoring cycle, The number of times the report was missed. Number of false alarms To handle cost weighting factors, As a weighting factor for safety monitoring, To misjudge the weighting factor, The underreporting ratio, For false alarm ratio, Security level; The reward function is substituted into the Q-learning algorithm for training to obtain technical indicators. These technical indicators are then substituted into the optimal decision output algorithm to obtain the optimal decision action sequence. The expression for the Q-learning algorithm is as follows:

[0024] In the formula, This is a technical indicator representing the long-term expected return of action a under system state s. Let be the system state at time t. Let t be the action to be performed. For learning rate, For instant reward function, Discount factor; The expression for the termination condition of the Q-learning algorithm is as follows:

[0025] In the formula, For the first In the next iteration, the long-term expected return of action a under system state s. For the first In the next iteration, the long-term expected return of action a under system state s. For a preset set of system states, A pre-defined set of actions to be executed; The optimal decision output algorithm expression is as follows:

[0026] In the formula, For optimal decision, Indicates the preset action space Select to The action with the highest value is executed. , This is a technical indicator representing the long-term expected return of action a under system state s.

[0027] In a preferred embodiment, the UAV inspection module performs the following actions to dynamically optimize the initial flight path based on the optimal decision and real-time environmental information to obtain the UAV inspection flight path: The method for dynamically optimizing the initial route based on the optimal decision and real-time environmental information is as follows: The route optimization algorithm for the initial route optimizes the initial route to obtain the following route optimization algorithm based on the initial route:

[0028] In the formula, The optimized path planning cost function, Distance weights As energy consumption weight, As time weight, To misjudge the weight, The three-dimensional Euclidean distance of the drone from the starting point to the destination. For the energy consumption of drone flight, The cumulative time from takeoff to landing. The error rate coefficient is the coefficient for misjudgment. The total number of waypoints for the initial route. For waypoint sequence index, For the first The longitude coordinates of each waypoint For the first Latitude coordinates of each waypoint For the first The altitude coordinates of each waypoint Based on power consumption, For speed coefficient, Let t be the flight speed of the drone at time t.

[0029] In a preferred embodiment, the health index algorithm expression is as follows:

[0030] In the formula, Health index For the first Weighting factors for class-aware parameters For the first The normalization function for class-aware data, For the first Sensors in time The measured value, The regional pollution coefficient, To accumulate the duration of high humidity, PM2.5 value, The aging coefficient of the conductor material. For the rate of temperature change, The absolute temperature of the conductor. For the time change, The activation energy of the conductor material, Boltzmann's constant, This represents the minimum value within the sensor's measurement range. This represents the maximum value within the range of sensor measurements.

[0031] Compared with the prior art, the present invention has the following beneficial effects: 1. By dynamically adjusting the weight coefficients through an adaptive filtering algorithm, electromagnetic interference and sensor noise in power analysis data are effectively eliminated, significantly improving the data signal-to-noise ratio and providing high-confidence basic data for subsequent fault analysis.

[0032] 2. By combining multi-physics field coupling analysis such as electricity, heat, and force, three-dimensional temperature field distribution data is generated, which can accurately identify the composite fault sites such as overheating and deformation of conductors, avoid the limitations of single-physics field analysis, and improve the comprehensiveness and accuracy of fault location.

[0033] 3. The system adjusts decision-making and inspection paths through real-time environmental information and dynamic algorithms to adapt to complex and ever-changing operating environments, reduce false alarms and missed alarms, and improve the robustness and practicality of the system. Attached Figure Description

[0034] Figure 1 This is a system architecture diagram of a power distribution network overhead line detection system provided in an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0035] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the following will be combined with the accompanying drawings to make a clear and complete description of the technical solutions in the embodiments of the present application. Obviously, the described embodiments belong to some of the embodiments of the present application but not all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the scope of protection of the present application.

[0036] The terms used in the embodiments of the present application are only for the purpose of describing particular embodiments and are not intended to limit the present application. The singular forms "the" and "said" used in the embodiments of the present application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. "Plural" generally includes at least two.

[0037] Depending on the context, the word "if" or "if" as used herein can be interpreted as "when" or "when" or "in response to determining" or "in response to detecting". Similarly, depending on the context, the phrase "if determined" or "if detecting (stated condition or event)" can be interpreted as "when determined" or "in response to determining" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)".

[0038] In addition, the step sequence in each of the following method embodiments is only an example and is not strictly limited.

[0039] In fact, the server equipment deployed by a power distribution network overhead line detection system can be composed of one or more devices. The power distribution network overhead line detection system can be implemented as a business instance, a virtual machine, or a hardware device. For example, the power distribution network overhead line detection system can be implemented as a business instance deployed on one or more devices in a cloud node. In short, the power distribution network overhead line detection system can be understood as a software deployed on a cloud node, which provides a power distribution network overhead line detection system for each user terminal. Alternatively, the power distribution network overhead line detection system can also be implemented as a virtual machine deployed on one or more devices in a cloud node. The virtual machine has application software for managing each user terminal. Alternatively, the power distribution network overhead line detection system can also be implemented as a server composed of a plurality of same or different types of hardware devices, and one or more hardware devices are provided to provide a power distribution network overhead line detection system for each user terminal.

[0040] In an implementation form, the overhead line detection system for distribution network and the user end are mutually adaptable. That is, the overhead line detection system for distribution network is installed as an application on a cloud service platform, and the user end is a client that establishes a communication connection with the application; or the overhead line detection system for distribution network is implemented as a website, and the user end is implemented as a webpage; or the overhead line detection system for distribution network is implemented as a cloud service platform, and the user end is implemented as an applet in an instant messaging application.

[0041] As shown in Figure 1 FIG. 1 is a system architecture diagram of an overhead line detection system for distribution network according to an embodiment of the present application.

[0042] The overhead line detection system for distribution network 100 can be set in a cloud server, and in an implementation form, can be one or more service devices, or can be installed as an application on a cloud (for example, a server of a mobile service operator, a server cluster, etc.), or can be developed as a website. According to the functions implemented, the overhead line detection system for distribution network 100 can include an intelligent denoising module 101, a composite positioning module 102, a digital twin construction module 103, a decision module 104, a UAV inspection module 105, and a data analysis module 106. The modules of the present application can also be referred to as units, which refer to a series of computer program segments that can be executed by an electronic device processor and can complete a fixed function, and are stored in the memory of the electronic device.

[0043] In an embodiment of the present application, each of the modules in the overhead line detection system for distribution network can be independently implemented and called by other modules. Here, calling can be understood as that a module can connect multiple modules of another type and provide corresponding services for the connected multiple modules. In the overhead line detection system for distribution network provided by the embodiment of the present application, without modifying the program code, the application range of the overhead line detection system for distribution network architecture can be adjusted by adding modules and directly calling, realizing cluster horizontal expansion, so as to achieve the purpose of quickly and flexibly expanding the overhead line detection system for distribution network. In actual application, the modules can be set in the same device or different devices, or can be set in a virtual device, such as a service instance in a cloud server.

[0044] The following will describe the components and specific work flow of the overhead line detection system for distribution network with reference to specific embodiments: The intelligent denoising module is used to obtain power analysis data of the overhead line based on a preset initial flight route, and to denoise the power analysis data by using a self-adaptive filtering algorithm to obtain accurate analysis data of the overhead line. In the embodiment of the present application, the self-adaptive filtering algorithm is as follows:

[0045] wherein, is an output value of the adaptive filter, is a transpose of a weight coefficient vector, is an input signal vector of the adaptive filter at time n, is an i-th element in the weight coefficient vector of the adaptive filter, is a value of the input signal vector of the adaptive filter at time n, wherein the input signal is power analysis data, and the power analysis data includes three-phase current waveform, color heat map and conductor sag data; The weight coefficient vector of the adaptive filter is updated based on an error signal of the adaptive filter algorithm and a gain vector of the adaptive filter algorithm.

[0046] It should be noted that the adaptive filter algorithm is used to realize real-time denoising of power analysis data by dynamically adjusting the weight coefficient vector. The three-phase current waveform, color heat map and conductor sag data in the power analysis data are taken as the input signal vector. The error signal between the filter output value and the expected signal is calculated. The gain vector is generated by combining the inverse correlation matrix and the forgetting factor, and then the weight coefficient is iteratively updated.

[0047] Further, the weight coefficient is dynamically corrected according to the current input signal characteristics, error feedback and historical state, so as to optimize the denoising effect. In the updating process, the algorithm minimizes the error signal to ensure that the error between the accurate analysis data after denoising and the real signal is close to 0.

[0048] Further, the algorithm has environmental adaptability and can suppress complex noise characteristics such as electromagnetic interference and sensor drift in overhead lines, thereby providing high-confidence electric-thermal analysis basis for subsequent composite positioning modules and avoiding fault misjudgment caused by noise.

[0049] It should be noted that the output value of the adaptive filter algorithm refers to the denoising signal after the adaptive filter processing. The denoising signal is generated by linear combination of the input signal vector and the dynamically adjusted weight coefficient vector. This value is obtained by real-time filtering of the power analysis data, eliminating electromagnetic interference and sensor noise, and obtaining a characteristic signal close to the real power state, thereby providing a high signal-to-noise ratio data basis for subsequent fault location.

[0050] It should be noted that the three-phase current waveform in the power analysis data is captured in real time by a current transformer or a Hall sensor installed on each phase conductor of the overhead line through a phase-locked loop synchronization sampling technology, the color heat map in the power analysis data is obtained by a UAV carrying an infrared thermal imager flying along the conductor to scan the conductor surface temperature distribution image, and the spatial heat map is generated by synchronously fusing GPS coordinate information, the conductor sag data in the power analysis data is measured by a laser radar or a binocular vision sensor to measure the spatial offset of the conductor relative to the reference position, the attitude error of the UAV is compensated by combining an inclination sensor, and the sag calculation formula is used for calculation.

[0051] Further, the expression of the sag calculation formula is as follows:

[0052] In the formula, is the displacement of the conductor in the direction of the line layout, is the displacement of the conductor in the horizontal transverse direction perpendicular to the line layout, is the displacement of the conductor in the vertical direction, is the sag value of the conductor in the time .

[0053] In the embodiment of the application, the error signal of the adaptive filtering algorithm and the gain vector of the adaptive filtering algorithm are used to update the weight coefficient vector of the adaptive filter, and the method comprises the following steps: The expression of the error signal of the adaptive filtering algorithm is as follows:

[0054] In the formula, is the error signal, is the expected signal, is the output value of the adaptive filter; The expression of the gain vector of the adaptive filtering algorithm is as follows:

[0055] In the formula, is the inverse correlation matrix, is the forgetting factor, is the gain vector, is the transpose of the input signal vector of the adaptive filter, is the weight coefficient vector of the adaptive filter, is the error signal, is the input signal vector of the adaptive filter at the n time; updating the weight coefficient vector based on a weight coefficient updating formula, the error signal and the gain vector, wherein an expression of the weight coefficient updating formula is as follows:

[0056] wherein, is the updated weight coefficient vector, is the weight coefficient vector, is the gain vector, is the error signal.

[0057] It should be noted that the error signal is a deviation of the expected signal from the output value of the adaptive filter, and is essentially a quantitative index of the denoising effect, which is used for dynamically feeding back the performance defects of the filter.

[0058] It should be noted that the gain vector is data used for controlling the direction and step length of the weight coefficient updating, and is determined by an inverse correlation matrix and a forgetting factor which balances the weight of the historical data and the new input signal and determines the response speed and stability of the filter to the dynamic noise environment.

[0059] Further, the larger the gain vector is, the more significant the adjustment amplitude of the weight coefficient is, and the adaptive filter is suitable for a noise mutation scene, and vice versa, which focuses on the smoothing of the historical state and prevents overfitting.

[0060] It should be noted that the weight coefficient updating formula is used for dynamically adjusting the weight of the adaptive filter to make it close to the optimal filtering state.

[0061] Further, the current weight coefficient represents a weight vector of the filter at time , which determines the weighting proportion of each component of the input signal, and the initial weight is usually set as a random value or a zero vector.

[0062] The composite positioning module is configured to perform multi-physical field coupling analysis on the precise analysis data and the physical field data of the overhead line to obtain electric-thermal power distribution data of the overhead line, and locate a fault site in the overhead line based on the electric-thermal power distribution data. In the embodiment of the present application, the composite positioning module performs the multi-physical field coupling analysis on the precise analysis data and the physical field data of the overhead line to obtain the electric-thermal power distribution data of the overhead line, and locates the fault site in the overhead line based on the electric-thermal power distribution data, and is specifically configured to: A specific process of the multi-physical field coupling analysis is as follows: the precise analysis data and the physical field data of the overhead line are substituted into a multi-physical field coupling equation for solving to obtain a coupling analysis result of the multi-physical field coupling analysis, wherein an expression of the multi-physical field coupling equation is as follows:

[0063] in the formula, is a dielectric constant, is a charge density, is a gradient operator, is an electric potential, is an electric current intensity, is an electric resistance, is a voltage, is heat generated by an electric field, is a thermal melting ratio, is a conductor temperature change amount, is a time change amount, is a conductor temperature, is a thermal conductivity, is a convective heat transfer coefficient, is an ambient temperature; three-dimensional temperature field distribution data of the overhead line is generated based on the coupling analysis result to obtain electric-thermal force distribution data of the overhead line; a three-dimensional physical field model of the overhead line is constructed based on the electric-thermal force distribution data, and when a site exceeding a preset maximum temperature threshold of the overhead line is detected in the three-dimensional physical field model, the site is recorded as a fault site.

[0064] It should be noted that the multi-physical field coupling analysis is performed by spatiotemporally aligning the precise analysis data and the physical field data of the overhead line, constructing a multi-physical field coupling matrix, constructing a multi-physical field coupling model based on the multi-physical field coupling matrix and electromagnetic, thermal and mechanical control equations in an algorithm of the multi-physical field coupling analysis, solving the electromagnetic, thermal and mechanical control equations in the algorithm of the multi-physical field coupling analysis to obtain electric-thermal force distribution data of the overhead line, mapping the electric-thermal force distribution data to the multi-physical field coupling model, extracting thermal analysis data therefrom to obtain three-dimensional temperature field distribution data, obtaining a continuous overheating region in the overhead line according to a temperature tolerance threshold in a conductor tolerance threshold of the overhead line and the three-dimensional temperature field distribution data, recording a gravity center coordinate of the continuous overheating region as a candidate fault site, and performing a compound fault analysis on the candidate fault site in combination with a solution of an electromagnetic control equation in the algorithm of the multi-physical field coupling analysis and a deformation amount of a conductor in the overhead line, and recording a site with an analysis result of fault as a fault site of the overhead line.

[0065] Further, the preset conductor resistance threshold of the overhead line is a limit physical parameter value that the conductor can withstand under safe operation conditions, including temperature, current, mechanical stress and other key indicators.

[0066] The digital twin construction module is configured to construct a digital twin model of the overhead line, and map the fault site and the electric-thermal power distribution data to the digital twin model. In the embodiment of the present application, the digital twin construction module is configured to, in the process of constructing the digital twin model of the overhead line, specifically: obtain a digital model of the overhead line, spatially map the digital model to the overhead line to obtain a digital twin model of the overhead line, and map the fault site and the electric-thermal power distribution data to the digital twin model.

[0067] It should be noted that the specific method of constructing the digital twin model is as follows: obtain the geometric point cloud data of the overhead line by using a UAV carrying a laser radar, generate a three-dimensional network model based on the geometric point cloud data, spatially map the three-dimensional network model to a three-dimensional modeling model of the overhead line, and embed the physical property data of the overhead line into the three-dimensional network model to obtain the digital twin model.

[0068] The decision module is configured to construct a knowledge graph of the overhead line based on the perception data of the overhead line, infer the electric-thermal power distribution data based on the knowledge graph to obtain the fault type of the overhead line, and generate an optimal decision of the overhead line by using a reinforcement learning algorithm. In the embodiment of the present application, the decision module is configured to, in the process of inferring the electric-thermal power distribution data based on the knowledge graph to obtain the fault type of the overhead line, specifically: perform data integrity processing and data feature construction processing on the perception data of the overhead line to obtain feature data of the overhead line, and construct a knowledge graph based on the feature data, wherein the construction method of the knowledge graph is as follows: define the devices in the overhead line as entities, construct the relationships between the entities to obtain the association relationships of the devices in the overhead line, and store the entities and the association relationships of the devices to a database to obtain the knowledge graph. perform triple confidence reasoning on the electric-thermal power distribution data based on the knowledge graph to obtain the fault type, wherein the expression of the triple confidence score in the triple confidence reasoning is as follows:

[0069] In the formula, is the triple confidence score of the knowledge graph, This is the head entity of the knowledge graph. The aforementioned relationship, This is the tail entity of the knowledge graph. This is a low-dimensional vector of the head entity. The head entity vector of the knowledge graph transpose, For the tail entity, (This is a low-dimensional vector.) This is a matrix representing the aforementioned relationships; The formula for converting the confidence score of the triplet is as follows:

[0070] In the formula, This is the probability transformation result of the confidence score of the triplet. Based on the triple confidence scores of the knowledge graph, a fault type inference function is constructed according to the probability transformation results. Inference is then performed based on the fault type inference function to obtain the fault type. The expression of the fault type inference function is as follows:

[0071] In the formula, For the failure type of the final reasoning, For a preset set of fault types, This refers to the set of triples in the knowledge graph that are associated with the current electrothermal distribution data. This is the probability transformation result of the confidence score of the triplet. For fault type weighting function, The value of t is determined by taking the number of faults of the most frequent fault type among the fault types. Representing entities Historical data based on fault type The number of occurrences, wherein the mathematical expression of the fault type weighting function is as follows:

[0072] In the formula, For fault type weighting function, For entities Historical data based on fault type Number of times it happens This is a preset set of fault types.

[0073] It should be noted that the entities include: conductors, towers, insulators, and sensors, and the equipment relationships include: conductor-tower, insulator-tower, sensor-conductor, and tower-tower; Further, in the entity class, the conductor attribute includes material, cross-sectional area, rated current and service life; The tower pole attribute includes tower pole type, height, foundation type and GPS coordinate; The insulator attribute includes material type, number of sheds, creepage distance and surface contamination level; The sensor attribute includes sensor type, installation position and sampling frequency.

[0074] It should be noted that the triple confidence inference is firstly based on the equipment association relationship to perform triple mapping on the electric power distribution data to obtain a temporary triple set associated with the equipment state, generate a matrix of the equipment association relationship based on the temporary triple set, calculate the triple confidence score by substituting the temporary triple set and the matrix of the equipment association relationship into an expression of the triple confidence score, obtain a quantitative result of the triple confidence score, substitute the quantitative result into a triple confidence score conversion formula to obtain a probability conversion result of the triple confidence score, construct a fault type inference function based on the probability conversion result to perform fault type inference, and obtain the fault type.

[0075] Further, the fault type inference based on the probability conversion result of the triple confidence score is to perform weighted linear combination on a historical fault type weight function to obtain a comprehensive confidence score of the fault type, perform fault type decision on the comprehensive confidence score to obtain a fault type label, and generate the fault type based on the fault type label, wherein an expression of the weighted linear combination is as follows:

[0076] In the formula, is the comprehensive confidence score of the fault type, is the probability conversion result of the triple confidence score corresponding to the i-th fault type, is the historical fault type.

[0077] In the embodiment of the present application, the perception data of the overhead line is subjected to data integrity processing and data feature construction processing, and the specific method is as follows: S701, missing value prediction is performed on the perception data to obtain missing values of the perception data, and the missing values are filled to obtain one-stage complete data of the perception data; S702, Z-score of each data in the one-stage complete data is calculated, the perception data with an absolute value of the Z-score greater than 3 is recorded as an abnormal value, and the abnormal value is deleted to obtain two-stage complete data of the perception data, wherein the Z-score is a deviation degree of data from the mean value. S703, performing repetitive screening on the two-stage complete data by a similarity matching algorithm to obtain complete data of the perception data; S704, performing feature extraction on the complete data by a statistical feature construction method to obtain feature data of the perception data.

[0078] It should be noted that the prediction and filling of missing values is based on a regression model to model the spatio-temporal features of the perception data containing missing values to obtain the prediction and filling results of the missing values. The adjacent data of the missing point is input to the model, the hyperparameters of the model are optimized by grid search, the predicted value of the missing value is output, the predicted value is filled into the original data according to the timestamp, and one-stage complete data without missing values is generated to ensure the continuity of the data to meet the subsequent analysis requirements.

[0079] It should be noted that Z-score is a statistical measure for quantifying the deviation of a single data point from the mean of the data set, in units of standard deviation, wherein the calculation formula of the Z-score is as follows:

[0080] In the formula, Z-score is the standard score, is the data point to be evaluated, is the mean of the data set, is the standard deviation of the data set.

[0081] Further, the greater the absolute value of the Z-score, the farther the data is.

[0082] It should be noted that the two-stage complete data is screened by a similarity matching algorithm. First, the two-stage complete data is segmented into time series based on a sliding window segmentation method to obtain a set of equal-length or variable-length data segments. The similarity between the data segment set is calculated based on a similarity matching algorithm to obtain a similarity cost matrix of all data segment pairs. Based on a pre-set similarity threshold and a time overlap rule, the cost matrix is determined for repeated segments to obtain a set of redundant data segments marked as redundant. Based on the principle of earliest timestamp priority, the redundant data segment set is processed to remove duplicates to obtain three-stage complete data without duplicates, wherein the expression of the similarity matching algorithm is as follows:

[0083] In the formula, is the similarity cost matrix, are two data segments to be compared in the data segment set, is the total number of alignment steps, is the step index of the alignment path, For the corresponding data of the step, For the corresponding data of the step.

[0084] In the embodiment of the present application, the decision module executes and utilizes the reinforcement learning algorithm to generate the optimal decision of the overhead line, and is specifically used for: constructing a reinforcement learning model based on the probability conversion result; The reinforcement learning algorithm in the reinforcement learning model is as follows:

[0085] In the formula, is an instant reward function, is a real-time processing cost, is a preset cost threshold, is a safe operation time, is a monitoring period, is a number of missed reports, is a number of false reports, is a processing cost weight factor, is a safety monitoring weight factor, is a false judgment weight factor, is a missed report ratio, is a false report ratio, is a safety level; substitute the reward function into the Q-learning algorithm for training to obtain a technical index, and substitute the technical index into an optimal decision output algorithm to obtain an optimal decision action sequence, wherein the Q-learning algorithm expression is as follows:

[0086] In the formula, is a technical index of long-term expected return of action a under system state s, is a system state at time t, is an action at time t, is a learning rate, is an instant reward function, is a discount factor; The expression of the Q-learning algorithm training termination condition is as follows:

[0087] In the formula, is a long-term expected return value of action a under system state s at the iteration, For the first In the next iteration, the long-term expected return of action a under system state s is... For a preset set of system states, A pre-defined set of actions to be executed; The optimal decision output algorithm expression is as follows:

[0088] In the formula, For optimal decision, Indicates the preset action space Select to The action with the highest value is executed. , This is a technical indicator representing the long-term expected return of action a under system state s.

[0089] It should be noted that the methods for setting the processing cost weighting factor, safety monitoring weighting factor, and misjudgment weighting factor are as follows: Calculate real-time processing cost, safe runtime, and number of false alarms within a time interval. The cumulative impact area of ​​the real-time processing cost, safe runtime, and false alarm count is obtained by integrating the components within the range. The ratio of the cumulative impact area is the ratio of the processing cost weighting factor, the safety monitoring weighting factor, and the false alarm weighting factor. The expression for the cumulative impact area table is as follows:

[0090] In the formula, To calculate the cumulative impact area of ​​real-time processing costs, As the starting point of the time interval, The end of the time interval, To reduce real-time processing costs The function formed For safe operating time The function formed Number of false alarms The function formed The cumulative impact area during safe operation. This represents the cumulative area affected by the number of false alarms.

[0091] The UAV inspection module is used to dynamically optimize the initial route based on the optimal decision and real-time environmental information to obtain the inspection route of the UAV, and feed the inspection route back to the intelligent noise reduction module. In this embodiment of the invention, the UAV inspection module performs the following actions to dynamically optimize the initial flight path based on the optimal decision and real-time environmental information to obtain the UAV inspection flight path: The method for dynamically optimizing the initial route based on the optimal decision and real-time environmental information is as follows: The route optimization algorithm for the initial route optimizes the initial route to obtain the following route optimization algorithm based on the initial route:

[0092] In the formula, The optimized path planning cost function, Distance weights As energy consumption weight, As time weight, To misjudge the weight, The three-dimensional Euclidean distance of the drone from the starting point to the destination. For the energy consumption of drone flight, The cumulative time from takeoff to landing. The error rate coefficient is the coefficient for misjudgment. The total number of waypoints for the initial route. For waypoint sequence index, For the first The longitude coordinates of each waypoint For the first Latitude coordinates of each waypoint For the first The altitude coordinates of each waypoint Based on power consumption, For speed coefficient, Let t be the flight speed of the drone at time t.

[0093] It should be noted that the route optimization algorithm based on the initial route is used to obtain the optimized path planning cost function. The specific process of obtaining the UAV inspection route based on the optimized path planning cost function is as follows: First, the system sets weight factors according to the task requirements. The initial waypoint set is generated based on the location of the inspection target. Then, the improved A algorithm is used to calculate the path branch with the minimum cumulative cost. In each iteration step, the algorithm calculates the total cost, comprehensively considering the three-dimensional Euclidean length of the flight distance, energy consumption, cumulative time, and number of misjudgment triggers. At the same time, it forces the avoidance of obstacle coordinates and satisfies the endurance constraint. When a sudden obstacle occurs, the system dynamically adjusts the local path, updates the misjudgment coefficient through real-time sensor data, and quickly replans. The optimized waypoint sequence is output after simulation verification by a digital twin model, and finally forms a UAV inspection route.

[0094] The data analysis module is used to visualize the fault location and generate a detection report for the overhead line based on a health index algorithm.

[0095] In this embodiment of the invention, the health index algorithm expression is as follows:

[0096] In the formula, health index, the weight factor of the first class of perception parameters, the normalization function of the first class of perception data, the measurement value of the first class of sensors at time , the area pollution coefficient, the cumulative high humidity duration, the PM2.5 value, the wire material aging coefficient, the temperature change rate, the wire absolute temperature, the time change amount, the activation energy of the wire material, the Boltzmann constant, the minimum value of the sensor measurement value interval, the maximum value of the sensor measurement value interval.

[0097] It should be noted that the method for generating a detection report based on the health index is as follows: first, based on the preset health level threshold, the health index is classified into a state to obtain a health rating of the line, wherein the health rating includes excellent, medium and good; based on historical fault data, risk correlation analysis is performed on the current health index to obtain main risk factors; based on the health rating and the risk factors, the overhead line is prioritized for maintenance to generate a detection report.

[0098] Further, the area pollution coefficient is graded according to the PM2.5 concentration, and the low level is when the PM2.5 concentration is less than 0.8, the medium level is when the PM2.5 concentration is less than 1.2, and the high level is when the PM2.5 concentration is greater than 1.2.

[0099] It should be noted that the Boltzmann constant is a physical constant related to temperature and energy, and the value is J / K.

[0100] Compared with the prior art, the present application has the following beneficial effects: 1. By dynamically adjusting the weight coefficient through the adaptive filtering algorithm, electromagnetic interference and sensor noise in power analysis data are effectively eliminated, and the data signal-to-noise ratio is significantly improved, providing high-confidence basic data for subsequent fault analysis.

[0101] 2. Combined with multi-physical field coupling analysis of electricity, heat, force, etc., three-dimensional temperature field distribution data is generated to accurately identify composite fault sites such as wire overheating and deformation, avoid the limitations of single physical field analysis, and improve the comprehensiveness and accuracy of fault location.

[0102] 3. The system adjusts the decision and inspection path through real-time environmental information and dynamic algorithm, adapts to complex and variable operating environment, reduces the false negative and false positive rates, and improves the robustness and practicality of the system.

[0103] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0104] The embodiments of the present application can acquire and process related data based on artificial intelligence technology. Artificial intelligence is the use of digital computers or computer-controlled machines to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0105] Finally, 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, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A distribution network overhead line detection system, the system comprising an intelligent denoising module, a composite positioning module, a digital twin construction module, a decision module, a UAV inspection module, and a data analysis module, wherein: the intelligent denoising module is configured to obtain power analysis data of an overhead line based on a preset initial flight route, and to denoise the power analysis data by using an adaptive filtering algorithm to obtain accurate analysis data of the overhead line; the composite positioning module is configured to perform multi-physical field coupling analysis on the accurate analysis data and physical field data of the overhead line to obtain electric-thermal power distribution data of the overhead line, and to locate a fault site in the overhead line based on the electric-thermal power distribution data; the digital twin construction module is configured to construct a digital twin model of the overhead line, and to map the fault site and the electric-thermal power distribution data to the digital twin model; the decision module is configured to construct a knowledge graph of the overhead line based on perception data of the overhead line, to infer the electric-thermal power distribution data based on the knowledge graph to obtain a fault type of the overhead line, and to generate an optimal decision of the overhead line by using a reinforcement learning algorithm; the UAV inspection module is configured to dynamically optimize the initial flight route based on the optimal decision and real-time environmental information to obtain an inspection flight route of a UAV, and to feed back the inspection flight route to the intelligent denoising module; and the data analysis module is configured to visualize the fault site, and to generate a detection report of the overhead line based on a health index algorithm. The adaptive filtering algorithm is as follows: an error signal of the adaptive filtering algorithm and a gain vector of the adaptive filtering algorithm are used to update a weight coefficient vector of an adaptive filter. The updating of the weight coefficient vector of the adaptive filter based on the error signal of the adaptive filtering algorithm and the gain vector of the adaptive filtering algorithm comprises: an expression of the error signal of the adaptive filtering algorithm is as follows: an expression of the gain vector of the adaptive filtering algorithm is as follows: and the weight coefficient vector is updated based on a weight coefficient update formula, the error signal, and the gain vector, wherein an expression of the weight coefficient update formula is as follows: the composite positioning module, in performing the multi-physical field coupling analysis on the accurate analysis data and the physical field data of the overhead line to obtain the electric-thermal power distribution data of the overhead line, and locating the fault site in the overhead line based on the electric-thermal power distribution data, is specifically configured to: a specific process of the multi-physical field coupling analysis is as follows: the accurate analysis data and the physical field data of the overhead line are substituted into a multi-physical field coupling equation to obtain a coupling analysis result of the multi-physical field coupling analysis, wherein an expression of the multi-physical field coupling equation is as follows: three-dimensional temperature field distribution data of the overhead line is generated based on the coupling analysis result to obtain the electric-thermal power distribution data of the overhead line; and the three-dimensional temperature field distribution data of the overhead line is generated based on the coupling analysis result to obtain the electric-thermal power distribution data of the overhead line. ​ ​ ​ ​ ​ ​ 2. A distribution overhead line detection system as claimed in claim 1, wherein, ​ ， wherein, is an output value of the adaptive filter, is a transpose of a weight coefficient vector, is an input signal vector of the adaptive filter at time n, is an i-th element in the weight coefficient vector of the adaptive filter, is a value of the input signal vector of the adaptive filter at time n, wherein the input signal is power analysis data, and the power analysis data comprises: three-phase current waveform, color heat map, and conductor sag data; ​ 3. A distribution overhead line detection system as claimed in claim 2, wherein, ​ ​ ， wherein is an error signal, is a desired signal, is an output value of the adaptive filter; ​ ， wherein is an inverse correlation matrix, is a forgetting factor, is a gain vector, is a transpose of an input signal vector of the adaptive filter, is a weight coefficient vector of the adaptive filter, is an error signal, is an input signal vector of the adaptive filter at time n. ​ ， wherein is the updated weight coefficient vector, is the weight coefficient vector, is the gain vector, is the error signal.

4. A distribution overhead line detection system as claimed in claim 1, wherein, ​ ​ ， wherein is the dielectric constant, is the charge density, is the gradient operator, is the electric potential, is the current intensity, is the resistance, is the voltage, is the heat generated by the electric field, is the hot melt ratio, is the wire temperature variation, is the time variation, is the wire temperature, is the thermal conductivity, is the convective heat transfer coefficient, is the ambient temperature; ​ Based on the electric thermal force distribution data, a three-dimensional physical field model of the overhead line is constructed, and when a site exceeding a preset maximum temperature threshold of the overhead line is detected in the three-dimensional physical field model, the site is recorded as a fault site.

5. A distribution overhead line detection system as claimed in claim 4, wherein, The digital twin construction module is configured to: obtain a digital model of the overhead line, perform spatial mapping of the digital model and the overhead line to obtain a digital twin model of the overhead line, and map the fault site and the electric thermal force distribution data to the digital twin model.

6. A distribution overhead line detection system as claimed in claim 4, wherein, The decision module is configured to: perform data integrity processing and data feature construction processing on the perception data of the overhead line to obtain feature data of the overhead line, and construct a knowledge graph based on the feature data, wherein the knowledge graph is constructed according to the following method: define devices in the overhead line as entities, construct relationships between the entities to obtain device association relationships in the overhead line, and store the entities and the device association relationships in a database to obtain a knowledge graph; perform triple confidence reasoning on the electric thermal force distribution data based on the knowledge graph to obtain a fault type, wherein an expression of a triple confidence score in the triple confidence reasoning is as follows: ， wherein, is a triple confidence score for the knowledge graph, is a head entity for the knowledge graph, is the association relation, is a tail entity for the knowledge graph, is a low-dimensional vector for the head entity, is a head entity vector for the knowledge graph is a transpose of, is a low-dimensional vector for the tail entity, is a matrix for the association relation; a transformation formula expression of the triple confidence score is as follows: ， In the formula, is a probability conversion result of the triple confidence score, is a triple confidence score of the knowledge graph, a fault type reasoning function is constructed based on the probability conversion result, and reasoning is performed based on the fault type reasoning function to obtain a fault type; wherein an expression of the fault type reasoning function is as follows: ， In the formula, is the fault type of the final reasoning, is a preset fault type set, is a set of triples associated with the current electric and thermal power distribution data in the knowledge graph, is a probability conversion result of the triple confidence score, is a fault type weight function, is the value of t, which is the fault type with the highest number of occurrences in the fault type, represents an entity in the historical data due to the fault type occurred, wherein the mathematical expression of the fault type weight function is as follows: ， wherein, is a fault type weight function, is an entity in the historical data for the fault type occurred, is a set of preset fault types.

7. A distribution overhead line detection system as claimed in claim 6, wherein, the data integrity processing and the data feature construction processing on the perception data of the overhead line are performed according to the following method: S701, predicting missing values of the perception data to obtain missing values of the perception data, and filling the missing values to obtain one-stage complete data of the perception data; S702, calculating Z-score of each data in the one-stage complete data, recording the perception data with an absolute value of the Z-score greater than 3 as an abnormal value, and deleting the abnormal value to obtain two-stage complete data of the perception data, wherein the Z-score is a deviation degree of data from a mean value; S703, performing repeatability screening on the two-stage complete data by a similarity matching algorithm to obtain complete data of the perception data; S704, extracting features of the complete data by a statistical feature construction method to obtain feature data of the perception data.

8. A distribution overhead line detection system as claimed in claim 6, wherein, The decision module is configured to: construct a reinforcement learning model based on the probability conversion result; the reinforcement learning algorithm in the reinforcement learning model is as follows: ， In the formula, For instant reward function, To reduce real-time processing costs, The preset cost threshold, For safe operating time, For the monitoring cycle, The number of times the report was missed. Number of false alarms To handle cost weighting factors, As a weighting factor for safety monitoring, To misjudge the weighting factor, The underreporting ratio, For false alarm ratio, Security level; substitute the reward function into a Q-learning algorithm for training to obtain a technical index, and substitute the technical index into an optimal decision output algorithm to obtain an optimal decision action sequence, wherein an expression of the Q-learning algorithm is as follows: ， wherein is a technical indicator of long-term expected return of performing action a at system state s, is a system state at time t, is a performed action at time t, is a learning rate, is an immediate reward function, is a discount factor; an expression of a training termination condition of the Q-learning algorithm is as follows: ， wherein, is the long-term expected return value of performing action a in system state s at the is the long-term expected return value of performing action a in system state s at the is a set of predetermined system states, is a set of predetermined actions. an expression of the optimal decision output algorithm is as follows: ， In the formula, is the optimal decision, represents selecting the execution action that maximizes the value in the preset action space , is a technical index of the long-term expected return of the execution action a under the system state s.

9. A distribution overhead line detection system as claimed in claim 1, wherein, The unmanned aerial vehicle inspection module performs dynamic optimization on the initial flight path based on the optimal decision and real-time environment information to obtain an inspection flight path of the unmanned aerial vehicle, and is specifically used for: The method for dynamically optimizing the initial flight path based on the optimal decision and real-time environment information is as follows: a flight path optimization algorithm of the initial flight path optimizes the initial flight path to obtain a flight path optimization algorithm based on the initial flight path as follows: ， In the formula, The optimized path planning cost function, Distance weights As energy consumption weight, As time weight, To misjudge the weight, The three-dimensional Euclidean distance of the drone from the starting point to the destination. For the energy consumption of drone flight, The cumulative time from takeoff to landing. The error rate coefficient is the coefficient for misjudgment. The total number of waypoints for the initial route. For waypoint sequence index, For the first The longitude coordinates of each waypoint For the first Latitude coordinates of each waypoint For the first The altitude coordinates of each waypoint Based on power consumption, For speed coefficient, Let t be the flight speed of the drone at time t.

10. A distribution overhead line detection system as claimed in claim 1, wherein, The health index algorithm expression is as follows: ， In the formula, health index, is the first weight factor of the class of perception parameters, is the first normalization function of the class of perception data, is the first measurement value of the class of sensors at time , is the area pollution coefficient, is the cumulative high humidity duration, is the PM2.5 value, is the wire material aging coefficient, is the temperature change rate, is the wire absolute temperature, is the time change amount, is the activation energy of the wire material, is the Boltzmann constant, is the minimum value of the sensor measurement value interval, is the maximum value of the sensor measurement value interval.

Images