Unmanned aerial vehicle-based power transmission line inspection visual early warning system and method

Abstract

The application discloses a power transmission line inspection visual early warning system and method based on a UAV, relates to the technical field of power transmission line inspection, and breaks the limitation that static equipment attributes and dynamic environment information are separated in traditional inspection analysis by introducing a multi-source correlation deep coding mechanism, so that dynamic vulnerability assessment of power transmission equipment in a specific environment context is realized.The deep correlation coding module can accurately quantify the influence weight of external environmental factors such as strong wind, high temperature or high humidity on different types and different service life of power transmission equipment, so as to output a high-dimensional equipment-environment interaction feature vector, so that the risk assessment has high context dependency.The problem that all the equipment is uniformly analyzed without distinction in the prior art, resulting in resource waste and risk omission is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power transmission line inspection, in particular to a power transmission line inspection visual early warning system and method based on a UAV. BACKGROUND

[0002] With the deep application of UAV technology in the field of power inspection, power transmission line visual inspection has moved from simple image acquisition to intelligent analysis. The current mainstream inspection system generally adopts a non-discriminatory comprehensive scanning strategy, that is, regardless of the environmental conditions or the service state of all tower, insulator, conductor and other device components, a unified standard image recognition and defect identification process is performed. This mode can ensure the coverage, but it seriously ignores the finiteness of inspection resources and the non-uniformity of risk distribution. Especially under complex weather conditions, the dynamic risk of key components is significantly increased, and the analysis priority of static components should be correspondingly reduced.

[0003] Prior art such as patent application numbers CN117519240A, CN117521800A, CN118781095A / B, CN117809191A / B and UAV and power transmission line inspection related invention application patents can be seen that the existing system mostly collects device static attributes + real-time environmental data, predicts failure probability through a conventional time series model, and generates a UAV inspection instruction based on a single factor such as voltage level or failure probability. The existing system still mechanically allocates the same amount of computing resources for full analysis, resulting in waste of computing resources and delay in responding to key risks. At the same time, the existing method does not couple device attributes and real-time environmental data for modeling, resulting in the use of the same criteria for newly commissioned devices and devices near the end of life, which can easily lead to excessive analysis of low-risk devices or omission of analysis of high-risk devices, making it difficult to support precise and differentiated operation and maintenance decisions.

[0004] In addition, the early warning result is not associated with the power grid topology and fault conduction path, which cannot evaluate the cascading effect of a single point defect on the entire line and even the regional power grid, resulting in a lack of scientific basis for operation and maintenance resource scheduling. Finally, the analysis strategy of the existing system is fixed and cannot be dynamically optimized according to historical re-inspection feedback and new failure modes, making it difficult to form a closed-loop iteration capability of perception-decision-execution-feedback. The above problems are increasingly prominent in the background of frequent extreme weather and accelerated aging of power grid equipment, and there is an urgent need for a UAV power transmission line visual inspection method that integrates multi-source correlation, time series prediction and hierarchical early warning to achieve risk identification precision, resource scheduling efficiency and intelligent operation and maintenance decisions. SUMMARY

[0005] The purpose of the present application is to provide a power transmission line inspection visual early warning system and method based on a UAV, which solves the problems in the background art.

[0006] To solve the above technical problems, the present application adopts the following technical solutions: the first aspect of the present application provides a power transmission line inspection visual early warning system based on a UAV, comprising: a multi-source data aggregation module for acquiring multi-source associated data, the multi-source associated data including a power transmission equipment static attribute data set, a real-time environment perception data set, and a historical defect maintenance data set;

[0007] An associated feature deep coding module is used to deeply associate and encode the multi-source associated data to generate an equipment-environment interaction feature vector;

[0008] A time series risk prediction module is used to input the equipment-environment interaction feature vector into a time series risk prediction model to predict the probability distribution of each type of defect in the pre-defined defect type set of the power transmission equipment within a future pre-set time window;

[0009] A hierarchical early warning and task redirection module is used to determine the hierarchical early warning level of the power transmission equipment in the current inspection target area based on the predicted probability distribution and in combination with the equipment criticality evaluation index; and generate a UAV inspection resource scheduling instruction corresponding to the risk level according to the hierarchical early warning level;

[0010] A display terminal is used to display the hierarchical early warning level of the power transmission equipment in the current inspection target area in the entire three-dimensional visual model;

[0011] A UAV payload control module is used to execute the UAV inspection resource scheduling instruction.

[0012] The second aspect of the present application provides a method for the power transmission line inspection visual early warning system based on a UAV, comprising: step one, acquiring multi-source associated data, the multi-source associated data including a power transmission equipment static attribute data set, a real-time environment perception data set, and a historical defect maintenance data set;

[0013] Step two, deeply associating and encoding the multi-source associated data to generate an equipment-environment interaction feature vector;

[0014] Step three, inputting the equipment-environment interaction feature vector into a time series risk prediction model to predict the probability distribution of each type of defect in the pre-defined defect type set of the power transmission equipment within a future pre-set time window;

[0015] Step four, determining the hierarchical early warning level of the power transmission equipment in the current inspection target area based on the predicted probability distribution and in combination with the equipment criticality evaluation index; and generating a UAV inspection resource scheduling instruction corresponding to the risk level according to the hierarchical early warning level;

[0016] Step five, display the hierarchical early warning level of the power transmission equipment in the current inspection target area in the entire three-dimensional visualization model; and execute the unmanned aerial vehicle inspection resource scheduling instruction.

[0017] The beneficial effects of the present application are as follows: first, the present application breaks the limitation of separating static equipment attributes and dynamic environmental information in traditional inspection analysis by introducing a multi-source associated deep coding mechanism, and realizes dynamic vulnerability assessment of power transmission equipment in a specific environmental context. The problem of wasting resources and missing risks caused by indiscriminate uniform analysis of all equipment in the prior art is solved.

[0018] The present application adopts a time series risk prediction model based on a causal diagram convolution structure. This model not only captures the dependency of historical data in the time dimension, but also models the propagation and influence of potential defects between devices in the power network topology through a graph attention mechanism, realizing accurate prediction of the probability distribution of specific defect types in the future time. This prediction capability changes the inspection work from passive post-analysis to active risk avoidance, ensuring that the inspection task is triggered before the risk appears.

[0019] The present application constructs a comprehensive risk index, combining the predicted dynamic risk probability and the preset equipment criticality evaluation index, and implements hierarchical early warning on the inspection target equipment. Ensure that the inspection resources can be allocated to the high-risk target that needs attention in real time and dynamically, significantly improving the overall inspection efficiency and the pertinence of risk discovery.

[0020] The present application directly converts the early warning result into an executable unmanned aerial vehicle load control instruction, realizes closed-loop control from data analysis, risk decision to physical execution, makes the data acquisition strategy completely match the real-time risk priority, and ensures that the highest quality diagnostic data is obtained in high-risk scenarios, thereby improving the intelligence and engineering reliability of the inspection system. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiment or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0022] Figure 1 It is a schematic diagram of the system structure of the present application.

[0023] Figure 2 It is a schematic diagram of the method implementation step flow of the present application. DETAILED DESCRIPTION

[0024] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described, obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0025] With reference to Figure 1 The first aspect of the present application provides a UAV-based transmission line inspection visual early warning system and method, comprising: a multi-source data aggregation module for acquiring multi-source associated data, the multi-source associated data including a transmission equipment static attribute data set, a real-time environment perception data set and a historical defect maintenance data set;

[0026] In specific embodiments of the present application, the multi-source associated data acquisition specifically includes: acquiring the transmission equipment static attribute data set, and the static attribute data set covers transmission line type, tower pole model, equipment manufacturing year, installation location coordinates, insulator string type, conductor cross-sectional area and the last maintenance time;

[0027] The real-time environment perception data set is acquired through a real-time environment sensor array or an external data interface, and the real-time environment perception data set covers wind speed, wind direction, environmental temperature, humidity, air pressure, rainfall, light intensity and air quality index;

[0028] The historical defect maintenance data set is acquired, and the historical defect maintenance data set covers historical defect type, occurrence time, severity, repair measure and post-repair operation time.

[0029] A correlation feature deep coding module is used for deep correlation coding of the multi-source associated data to generate a device-environment interaction feature vector.

[0030] In specific embodiments of the present application, the multi-source associated data is deep correlation coded to generate a device-environment interaction feature vector, which specifically includes: the non-numerical features in the transmission equipment static attribute data set are embedded vectorized by a device attribute encoder, and connected with numerical features to obtain a static attribute feature tensor.

[0031] It should be noted that the non-numerical features such as tower pole type and insulator string type, and the device attribute encoder such as a one-hot encoder and a Word2Vec word embedding model are selected according to requirements in use, such as one-hot encoding for tower pole type and Word2Vec word embedding model for text information.

[0032] The multi-dimensional environment parameters in the real-time environment perception data set are processed in time domain fusion by an environment state encoder to obtain an environment state feature tensor;

[0033] It should be noted that the environment state encoder is a time sequence convolution network or a long short-term memory network.

[0034] The static attribute feature tensor and the environment state feature tensor are input into a cross-modal feature interaction layer in combination with the association rule of the historical defect type and the corresponding environment parameter in the historical defect maintenance data set, the cross-modal feature interaction layer is configured with an attention weighting mechanism, which is used for preferentially increasing the coupling weight calculation proportion of the environment parameter related to the high-frequency defect, and then calculating the coupling weight of the influence of the environment state on the potential defect of the static attribute;

[0035] Based on the coupling weight, the static attribute feature is dynamically adjusted and fused to generate the device-environment interaction feature vector.

[0036] In specific embodiments of the application, the attention weighting mechanism configured in the cross-modal feature interaction layer specifically includes: mapping the static attribute feature tensor into a query matrix;

[0037] It should be noted that the attention weight is allocated based on the association rule of the device static attribute feature and the defect risk, for each key attribute dimension of the device static attribute feature tensor, in combination with the anti-environment interference ability reflected by the device static attribute, such as the weak wind resistance of the old tower pole and the strong moisture resistance of the new moisture-resistant insulator, and referring to the association frequency of the attribute and the environment-induced defect in the historical defect data, the high attention weight is allocated to the attribute dimension of the defect according to the association frequency, and the low attention weight is allocated to the dimension with strong anti-interference ability.

[0038] The query matrix is constructed according to the dimension matching principle, that is, the attention weights of each key attribute dimension are arranged one by one according to the dimension order of the static attribute feature tensor to form a query matrix; the dimension of the query matrix is consistent with the key attribute dimension of the static attribute feature tensor, and the matrix element value quantitatively represents the attention degree of the device to the corresponding attribute related environment, which provides core data support for subsequent calculation of the association degree of the static attribute and the environment state.

[0039] The environment state feature tensor is mapped into a key matrix and a value matrix;

[0040] It should be noted that when mapping the key matrix, a fully connected layer is used as the mapping function, and the core attributes of the power transmission equipment susceptible to environmental influences such as tower material wind resistance, conductor heat resistance, and insulator moisture resistance are first determined. Then, the time sequence features of each environmental parameter in the environmental state feature tensor are mapped into vectors matching the dimensions of the core attributes. Each vector represents the influence correlation dimension of the corresponding environmental parameter on each core attribute. All vectors are arranged according to the environmental parameter-core attribute dimension to form the key matrix.

[0041] When mapping the value matrix, a fully connected layer with ReLU activation function is also used as the mapping function. The statistical rules of the influence intensity of environmental parameters on equipment attributes in historical defect maintenance data are combined, such as the influence intensity value of wind speed on tower material wind resistance. The time sequence features of each environmental parameter in the environmental state feature tensor are quantified into actual influence intensity values of the corresponding core attributes, arranged according to the same environmental parameter-core attribute dimension as the key matrix, to form the value matrix, ensuring that the key matrix and the value matrix are dimensionally matched, providing data support for subsequent correlation degree calculation.

[0042] By calculating the dot product of the query matrix and the key matrix, the correlation degree score matrix of the static attribute features to the environmental state features is obtained.

[0043] By performing flexible maximum value normalization processing on the correlation degree score matrix, the coupling weight matrix is obtained.

[0044] The coupling weight matrix is applied to the value matrix for weighted summation to obtain the environment-weighted static feature representation.

[0045] The environment-weighted static feature representation is connected and fused with the original static attribute feature tensor through residual connection to obtain the device-environment interaction feature vector.

[0046] It should be noted that the query matrix is mapped from the device static attribute feature tensor, which represents which environmental influences the current device is looking for in a simple way. For example, for an old angle steel tower manufactured in 2018, the query matrix is: tower material wind resistance, conductor heat resistance, and other equipment attributes. Assuming it is simplified to a 1-row 3-column vector:

[0047] ;

[0048] Where 0.9 represents the tower pole wind resistance requirement, 0.7 represents the conductor heat resistance requirement, and 0.2 represents the insulator moisture resistance requirement. The larger the value, the higher the attention of the equipment to that attribute, i.e., the more susceptible it is to the environmental influence corresponding to that attribute.

[0049] Key matrix is mapped from the environmental state feature tensor, which represents which influencing factors and their intensities in the current environment. For example, the real-time environment is a wind speed of 5 levels, a temperature of 32°C, and a humidity of 40%, and the key matrix is 3 rows and 3 columns, each row corresponding to an environmental factor, and each column corresponding to a concerned attribute of the device: ;

[0050] respectively represent the influence value of wind speed on the wind resistance of tower material, the influence value of wind speed on the heat resistance of conductor, and the influence value of wind speed on the moisture resistance of insulator;

[0051] respectively represent the influence value of temperature on the wind resistance of tower material, the influence value of temperature on the heat resistance of conductor, and the influence value of temperature on the moisture resistance of insulator;

[0052] respectively represent the influence value of humidity on the wind resistance of tower material, the influence value of humidity on the heat resistance of conductor, and the influence value of humidity on the moisture resistance of insulator;

[0053] The correlation degree score matrix of the static attribute feature to the environmental state feature is calculated as follows: the correlation degree score of wind speed to the device , the correlation degree score of temperature to the device , and the correlation degree score of humidity to the device The final correlation degree score matrix is obtained, and the coupling weight matrix is obtained after normalization, which respectively represents the weight of wind speed, the weight of temperature, and the weight of humidity.

[0054] The value matrix is the actual influence degree data of environmental factors, i.e., the specific influence value of each environmental factor such as wind speed and temperature on the attribute of the device, which is the core data for subsequent calculation of how the environment changes the device feature. Exemplarily, the value matrix is 3 rows and 3 columns, each row corresponds to an environmental factor, and each column corresponds to an attribute of the device, and the numerical value is the actual influence intensity of the environment on the attribute: ;

[0055] respectively represent the actual influence value of wind speed on the wind resistance of tower material, the actual influence value of wind speed on the heat resistance of conductor, and the actual influence value of wind speed on the moisture resistance of insulator;

[0056] respectively represent the actual influence value of temperature on the wind resistance of tower material, the actual influence value of temperature on the heat resistance of conductor, and the actual influence value of temperature on the moisture resistance of insulator;

[0057] respectively represent the actual influence value of humidity on the wind resistance of tower material, the actual influence value of humidity on the heat resistance of conductor, and the actual influence value of humidity on the moisture resistance of insulator.

[0058] The environment-weighted static feature representation is calculated as: tower material wind resistance property: ; conductor heat resistance property: ; insulator moisture resistance property: The environment-weighted static feature representation is obtained, that is, the impact quantitative value of the current environment on the wind resistance property of the tower material, the heat resistance property of the conductor and the moisture resistance property of the insulator.

[0059] The environment-weighted static feature representation is fused with the original static attribute feature tensor through residual connection to obtain the device-environment interaction feature vector: assuming that the original static attribute feature tensor is the attribute value of the device itself, reflecting the wind resistance upper limit of the tower material of the old angle steel tower, the heat resistance upper limit of the conductor itself and the moisture resistance upper limit of the insulator itself, for example, the basic attributes of the old angle steel tower are: , wherein 0.6 represents the basic wind resistance property value, 0.5 represents the basic heat resistance property value, and 0.8 represents the basic moisture resistance property value, then the fused tower material wind resistance property value is calculated as ; the fused conductor heat resistance property value is: ; the fused insulator moisture resistance property value is: , and finally the device-environment interaction feature vector is obtained, representing the dynamic risk correlation degree of the tower material wind resistance, the conductor heat resistance and the insulator moisture resistance.

[0060] The application breaks the limitation of separating static device attributes and dynamic environment information in traditional inspection analysis by introducing a multi-source association deep coding mechanism, and realizes dynamic vulnerability assessment of the power transmission device in a specific environmental context. The deep association coding module can accurately quantify the influence weight of external environmental factors such as strong wind, high temperature or high humidity on different types and different service life of power transmission devices, so as to output a high-dimensional device-environment interaction feature vector, so that the risk assessment has high context dependence. The problem of wasting resources and missing risks caused by indiscriminate unified analysis of all devices in the prior art is solved.

[0061] The time sequence risk prediction module is configured to input the device-environment interaction feature vector into a time sequence risk prediction model to predict the probability distribution of each type of defect in the pre-defined defect type set of the power transmission device within a future preset time window.

[0062] In specific embodiments of the application, the device-environment interaction feature vector is input into a time sequence risk prediction model to predict the probability distribution of each type of defect in the pre-defined defect type set of the power transmission device within a future preset time window, which specifically includes: the time sequence risk prediction model is configured as a causal graph convolution time sequence prediction structure, and the causal graph convolution time sequence prediction structure takes a historical device-environment interaction feature sequence as input and outputs a defect probability distribution of a future time window.

[0063] It should be noted that the training process of the time series risk prediction model needs to input the historical equipment-environment interaction feature sequence and the defect occurrence time-defect type-severity data in the historical defect maintenance data set corresponding to the sequence, learn the mapping rule of equipment-environment state-defect development through the correspondence between the two, and the details of logic and attention calculation are well known to those skilled in the art, which will not be described in detail here.

[0064] The causal graph convolution time series prediction structure includes a plurality of causal convolution modules, which are used to capture the dependence relationship in the time dimension and ensure that the prediction process is based only on historical information.

[0065] It should be noted that the causal graph convolution time series prediction structure refers to an integrated time series prediction architecture that integrates causal convolution modules and graph attention layers, and the core content is to capture the time dimension dependence of power transmission equipment risk and the defect propagation relationship between equipment through the cooperation of the two modules.

[0066] The causal graph convolution time series prediction structure includes a graph attention layer, which is used to model the propagation and influence relationship of the defect state between adjacent equipment in the power transmission network.

[0067] The time series risk prediction model performs prediction operation and outputs a probability matrix, the dimension of the probability matrix corresponds to the product of the preset time window length and the number of the predefined defect type set.

[0068] It should be noted that the modeling of the defect propagation and influence relationship between equipment includes the graph representation of the power grid topology, for example, modeling the power transmission network as an undirected graph G=(V,E), where the node V is a power transmission tower or equipment, and the edge E is the physical connection between the equipment, such as shared conductors, etc., and the edge weight is based on the distance between the two equipment, such as distance <500 meters weight 1, 500-1000 meters weight 0.5, >1000 meters weight 0.

[0069] The calculation method of the attention weight is that the attention weight of the graph attention layer is calculated by the cosine similarity of the node features, for example, when predicting the flashover probability of the insulator of tower A, the graph attention layer gives the attention weight of tower B, which is adjacent to tower A, has a distance of 300 meters and shares 1 segment of conductor, is 0.6, and gives the attention weight of tower C, which is farther away, has a distance of 800 meters, is 0.2, so that the prediction result can reflect the influence of the insulator state of tower B on tower A.

[0070] It also needs to be explained that the core function of the time sequence risk prediction model is to predict the future probability based on the historical feature sequence, which has been maturely applied in the power transmission line defect prediction scene, such as transformer state prediction and line icing risk prediction. The input and output matching logic of such a model is well known to those skilled in the art. In addition, the future preset time window can be combined with the defect development cycle of the power transmission equipment. It is known in the art that defects such as insulator flashover and conductor dancing need to be reasonably set for several hours to several tens of hours of environmental influence accumulation. Based on the above, the technical logic, data basis and parameter setting of the time sequence risk prediction model can be independently understood and reproduced by those skilled in the art based on existing knowledge, and will not be described in detail here.

[0071] In specific embodiments of the present application, the set of predefined defect types includes insulator flashover risk, conductor dancing risk, hardware connection loosening risk, tower material structure stress overrun risk, and hot spot abnormal heating risk.

[0072] The present application adopts a time sequence risk prediction model based on a causal graph convolution structure. This model not only captures the dependence of historical data in the time dimension, but also models the propagation and influence of potential defects between devices in the power transmission network topology through a graph attention mechanism, achieving accurate prediction of the probability distribution of the occurrence of a specific defect type in the future time. This prediction capability enables the inspection work to be changed from passive post-analysis to active risk avoidance, ensuring that the inspection task is triggered before the risk appears.

[0073] The hierarchical warning and task redirection module is used to determine the hierarchical warning level of the power transmission equipment in the current inspection target area based on the predicted probability distribution and the device criticality evaluation index; and generate a UAV inspection resource scheduling instruction corresponding to the risk level according to the hierarchical warning level.

[0074] In specific embodiments of the present application, the hierarchical warning level of the power transmission equipment in the current inspection target area is determined based on the predicted probability distribution and the device criticality evaluation index, which specifically includes: for each predefined defect type, extracting the maximum predicted probability value in the probability matrix and recording it as a dynamic risk probability value.

[0075] It needs to be explained that the purpose of this step is to focus on the highest risk period of each defect type and avoid hiding the key risks due to the average probability.

[0076] The device criticality evaluation index of the power transmission equipment is obtained, which is pre-set based on the power transmission line voltage level, the importance of the area to which it belongs, and the vulnerability of the power grid topology, and the device criticality evaluation index is applicable to all predefined defect types of the same device.

[0077] calculating a comprehensive risk index by a risk quantification function, the comprehensive risk index being equal to the dynamic risk probability value multiplied by a weighted sum of the equipment criticality evaluation indexes, the comprehensive risk index being calculated as: wherein is a dynamic risk probability value, , , are respectively a first, a second, and a third criticality factor, , are respectively preset weights of the first, the second, and the third criticality factors, for example , , are respectively 0.4, 0.3, and 0.3;

[0078] It should be noted that the preset weights of the first, the second, and the third criticality factors are determined in advance: referring to a weight distribution benchmark of criticality factors in the power industry, combining regional power transmission fault data, statistically calculating the influence proportion of each criticality factor on the corresponding fault occurrence, and determining by an analytic hierarchy process, so as to ensure that the weights match the actual fault influence law and are not arbitrarily set, and can be directly used for equipment criticality evaluation.

[0079] According to the interval of the comprehensive risk index corresponding to the first, the second, and the third early warning levels of each predefined defect type stored in the data warehouse, the hierarchical early warning level is determined to be: the first early warning, the second early warning, or the third early warning, and the highest early warning level among all defect types is selected as the hierarchical early warning level of the power transmission equipment in the current inspection target area.

[0080] It should be noted that the interval of the comprehensive risk index corresponding to the first, the second, and the third early warning levels is calibrated in combination with the statistical results of historical defect maintenance data sets, and can be determined based on conventional historical data statistical calibration methods in the prior art: selecting historical defect maintenance data of the same type and same service environment as the current inspection target equipment, calculating the index value corresponding to the historical defect type according to the comprehensive risk index calculation formula disclosed in the present application, and then using existing statistical clustering or threshold division means, such as the K-means algorithm and the percentile method, to determine the index interval of each early warning level, and finally storing in the data warehouse after conventional data verification and adjustment. The specific statistical calibration details belong to the prior art and are not described in detail here.

[0081] In specific embodiments of the present application, the pre-setting process of the equipment criticality evaluation index specifically includes: determining the first criticality factor based on the voltage level of the power transmission line, the value of the first criticality factor being positively correlated with the fault influence range corresponding to the voltage level of the power transmission line.

[0082] It should be noted that the higher the voltage level, the higher the value of the first criticality factor, and the voltage level is, for example, extra-high voltage, high voltage, medium voltage, and low voltage.

[0083] determining a second criticality factor based on the importance of the area to which the power transmission equipment belongs, the value of the second criticality factor being positively correlated with the load type in the area and the economic and social loss caused by power failure;

[0084] It should be noted that the higher the proportion of important loads in the area and the greater the loss caused by power failure, the higher the value of the second criticality factor. The composite types are important loads, ordinary loads and temporary loads. The important loads are directly related to public safety and livelihood protection or major activities such as hospitals, airports and railway stations. The ordinary loads are, for example, residential areas and ordinary shops. The temporary loads are, for example, temporary power consumption in construction;

[0085] determining a third criticality factor based on the vulnerability of the power grid topology, the value of the third criticality factor being positively correlated with the node role of the power transmission equipment in the power grid topology and the risk of causing cascading power failure;

[0086] It should be noted that the wider the range of the equipment as a hub node and the wider the range of the failure chain, the higher the value of the third criticality factor. The node roles are hub nodes, branch nodes and end nodes. The hub nodes have the highest risk of causing cascading power failure, followed by branch nodes, and end nodes have the lowest risk.

[0087] The first criticality factor, the second criticality factor and the third criticality factor are weighted and fused according to a preset weight to obtain the equipment criticality evaluation index.

[0088] The present application constructs a comprehensive risk index, combines the predicted dynamic risk probability and the preset equipment criticality evaluation index, and implements graded early warning on the inspection target equipment. The graded early warning mechanism accurately divides the inspection resource scheduling instructions into first, second and third levels, and presets completely different unmanned aerial vehicle flight parameters and sensor configuration parameters for each level. It ensures that the inspection resources can be allocated to the high-risk target that needs attention in real time and dynamically, significantly improving the overall inspection efficiency and the pertinence of risk discovery.

[0089] In specific embodiments of the present application, generating an unmanned aerial vehicle inspection resource scheduling instruction corresponding to the risk level according to the graded early warning level specifically includes: when the early warning level is first early warning, generating a high-priority directional fine inspection instruction;

[0090] For example, the high-priority directional fine inspection instruction requires the unmanned aerial vehicle to immediately adjust the flight path to the target equipment, and to use a visible light with a resolution of ≥1920x1080, a thermal imager with a sensitivity of ≤0.05℃ and an ultraviolet corona instrument with a sampling rate of ≥30Hz to perform full-load collaborative scanning. At the same time, it is required to reduce the flight speed to a preset low speed value and continue hovering for at least three minutes for data acquisition;

[0091] When the early warning level is a second-level early warning, a medium-priority key monitoring inspection instruction is generated;

[0092] Illustratively, the medium-priority key monitoring inspection instruction requires the UAV to increase the duration of the low-altitude segment flight when passing through the target device, and to zoom or adjust the gain of the sensor corresponding to the high-risk defect type;

[0093] When the early warning level is a third-level early warning, a low-priority routine cruise instruction is generated.

[0094] Illustratively, the low-priority routine cruise instruction requires the UAV to perform full-line scanning at a preset cruise speed and standard flight path, and only raw data collected is analyzed in the background.

[0095] The present application directly converts the early warning result into an executable UAV payload control instruction, realizing closed-loop control from data analysis, risk decision-making to physical execution. Through the flight parameter adjustment interface and the sensor configuration interface, the system can adjust the flight speed, zoom ratio and spectral selection of the UAV in real time according to the early warning level, so that the data acquisition strategy completely matches the real-time risk priority, ensuring that the highest quality diagnostic data is obtained in high-risk scenarios, thereby improving the intelligence and engineering reliability of the inspection system.

[0096] A display terminal is used to display the hierarchical early warning level of the power transmission device in the current inspection target area in the entire three-dimensional visualization model;

[0097] A UAV payload control module is used to execute the UAV inspection resource scheduling instruction.

[0098] Referring to Figure 2 The second aspect of the present application provides a method for executing the UAV-based power transmission line inspection visualization early warning system described in the present application, comprising: step one, acquiring multi-source associated data, the multi-source associated data including a power transmission device static attribute data set, a real-time environment perception data set and a historical defect maintenance data set;

[0099] Step two, deeply associating and coding the multi-source associated data to generate a device-environment interaction feature vector;

[0100] Step three, inputting the device-environment interaction feature vector into a time-series risk prediction model to predict the probability distribution of each type of defect in the set of predefined defect types occurring to the power transmission device within a future preset time window;

[0101] Step four, based on the predicted probability distribution, combining a device criticality evaluation index, determining the hierarchical early warning level of the power transmission device in the current inspection target area; according to the hierarchical early warning level, generating a UAV inspection resource scheduling instruction corresponding to the risk level;

[0102] Step five, display the hierarchical early warning level of the power transmission equipment in the current inspection target area in the entire three-dimensional visualization model; and execute the unmanned aerial vehicle inspection resource scheduling instruction.

[0103] It should be added that the above-mentioned formula can be translated into a standard value without unit or a parameter of the same dimension that can be superimposed by the principle of dimensional consistency and mathematical standardization methods such as normalization processing, dimensionless parameter conversion or unit system unification, so as to eliminate the interference of different dimensions on the operation logic, so that the formula has mathematical operation rationality and objective law adaptability while retaining the original data distribution characteristics. The above is only an exemplary embodiment of the present application, and cannot limit the scope of the present application.

[0104] The above embodiments can be realized wholly or partially by software, hardware, firmware or any combination thereof. When realized by software, the above embodiments can be realized in the form of a computer program product wholly or partially. Those skilled in the art can realize that the algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized by hardware or software depends on the specific application and design constraints of the technical solutions, and those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0105] In addition, the functional modules in each embodiment of the present application can be integrated in one processing module, or each module can exist physically independently, or two or more modules can be integrated in one module.

[0106] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any skilled person in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0107] The above is only an example and description of the concept of the present application, and those skilled in the art can make various modifications or supplements or use similar ways to replace the described specific embodiments, as long as they do not deviate from the concept of the present application or exceed the scope defined by the present application, which should be within the protection scope of the present application.

Claims

1. A UAV-based visualized early warning system for power transmission line inspection, characterized in that, include: The multi-source data aggregation module is used to acquire multi-source associated data, which includes power transmission equipment static attribute dataset, real-time environmental perception dataset, and historical defect maintenance dataset. The deep association feature encoding module is used to perform deep association encoding on the multi-source association data to generate device-environment interaction feature vectors; The time-series risk prediction module is used to input the device-environment interaction feature vector into the time-series risk prediction model to predict the probability distribution of various types of defects in a predefined defect type set of power transmission equipment within a future preset time window; The graded early warning and task redirection module is used to determine the graded early warning level of power transmission equipment in the current inspection target area based on the predicted probability distribution and combined with the equipment criticality evaluation index. Based on the aforementioned graded early warning level, generate a drone inspection resource scheduling instruction corresponding to the aforementioned risk level; The display terminal is used to display the graded early warning level of the power transmission equipment in the current inspection target area in the entire three-dimensional visualization model; The UAV payload control module is used to execute the UAV inspection resource scheduling instructions.

2. The UAV-based power transmission line inspection visualization and early warning system according to claim 1, characterized in that, The acquisition of multi-source correlation data specifically includes: Obtain the static attribute dataset of the power transmission equipment, which includes the type of power transmission line, tower model, year of equipment manufacturing, coordinates of installation location, type of insulator string, cross-sectional area of ​​conductor, and time of the most recent maintenance. The real-time environmental sensing dataset is obtained through a real-time environmental sensor array or an external data interface. The real-time environmental sensing dataset includes wind speed, wind direction, ambient temperature, humidity, air pressure, rainfall, light intensity, and air quality index. Obtain the historical defect maintenance dataset, which includes the historical defect type, occurrence time, severity, remediation measures, and post-remediation runtime.

3. The UAV-based power transmission line inspection visualization and early warning system according to claim 1, characterized in that, The step of performing deep association encoding on the multi-source associated data to generate a device-environment interaction feature vector specifically includes: The non-numerical features in the static attribute dataset of the power transmission equipment are embedded and vectorized by the device attribute encoder, and then connected with the numerical features to obtain the static attribute feature tensor. The environmental state feature tensor is obtained by performing temporal fusion processing on the multi-dimensional environmental parameters in the real-time environmental perception dataset through an environmental state encoder. Based on the correlation between historical defect types and corresponding environmental parameters in the historical defect maintenance dataset, the static attribute feature tensor and the environmental state feature tensor are input into the cross-modal feature interaction layer. The cross-modal feature interaction layer is configured with an attention weighting mechanism to prioritize increasing the proportion of coupling weight calculation for environmental parameters related to high-frequency defects, thereby calculating the coupling weight of the influence of environmental state on potential static attribute defects. Based on the coupling weights, the static attribute features are dynamically adjusted and fused to generate the device-environment interaction feature vector.

4. The UAV-based power transmission line inspection visualization and early warning system according to claim 3, characterized in that, The attention weighting mechanism configured in the cross-modal feature interaction layer specifically includes: Map the static attribute feature tensor to a query matrix; The environmental state feature tensor is mapped into a key matrix and a value matrix; By calculating the dot product of the query matrix and the key matrix, the correlation score matrix between static attribute features and environmental state features is obtained. The coupling weight matrix is ​​obtained by performing flexible maximum value normalization on the correlation score matrix; The coupling weight matrix is ​​applied to the value matrix for weighted summation to obtain the static feature representation of the environment after weighting. The device-environment interaction feature vector is obtained by performing residual connection and fusion of the weighted static feature representation of the environment with the original static attribute feature tensor.

5. The UAV-based power transmission line inspection visualization and early warning system according to claim 4, characterized in that, The step of inputting the device-environment interaction feature vector into the time-series risk prediction model to predict the probability distribution of various types of defects in the predefined defect type set of power transmission equipment within a future preset time window specifically includes: The temporal risk prediction model is configured as a causal graph convolutional temporal prediction structure, which takes the historical device-environment interaction feature sequence as input and the defect probability distribution of the future time window as output. The causal graph convolutional temporal prediction structure includes a multi-layer causal convolutional module, which is used to capture dependencies in the temporal dimension and ensure that the prediction process is based only on historical information. The causal graph convolutional temporal prediction structure includes a graph attention layer, which is used to model the propagation and influence relationship of defect states between adjacent devices in the power transmission network. The time-series risk prediction model performs prediction operations and outputs a probability matrix. The dimension of the probability matrix corresponds to the preset time window length multiplied by the number of predefined defect type sets.

6. The UAV-based power transmission line inspection visualization and early warning system according to claim 5, characterized in that, The predefined set of defect types includes: insulator flashover risk, conductor galloping risk, hardware connection loosening risk, tower structure stress exceeding limit risk, and hot spot abnormal heating risk.

7. The UAV-based power transmission line inspection visualization and early warning system according to claim 6, characterized in that, Based on the predicted probability distribution and combined with the equipment criticality evaluation index, the classification and early warning levels for power transmission equipment within the current inspection target area are determined as follows: For each predefined defect type, the maximum predicted probability value in the probability matrix is ​​extracted and recorded as the dynamic risk probability value. The criticality evaluation index of the power transmission equipment is obtained, and the criticality evaluation index is preset based on the voltage level of the transmission line, the importance of the region, and the vulnerability of the power grid topology. A comprehensive risk index is calculated using a risk quantification function. The comprehensive risk index is equal to the dynamic risk probability value multiplied by the weighted sum of the equipment criticality evaluation indicators. Based on the range of comprehensive risk indices corresponding to Level 1, Level 2, and Level 3 warnings for each predefined defect type stored in the data warehouse, the graded warning levels are determined as: Level 1, Level 2, or Level 3 warnings. The highest warning level among all defect types is selected as the graded warning level for the power transmission equipment in the current inspection target area.

8. The UAV-based power transmission line inspection visualization and early warning system according to claim 7, characterized in that, Based on the aforementioned graded early warning levels, generating drone inspection resource scheduling instructions corresponding to the risk levels specifically includes: When the warning level is Level 1, a high-priority targeted fine-grained inspection instruction is generated; When the warning level is Level II, a medium-priority key monitoring and inspection instruction is generated. When the warning level is Level 3, a low-priority routine patrol command is generated.

9. The UAV-based power transmission line inspection visualization and early warning system according to claim 7, characterized in that, The pre-setting process for the equipment criticality evaluation indicators specifically includes: The first criticality factor is determined based on the voltage level of the transmission line, and the value of the first criticality factor is positively correlated with the fault impact range corresponding to the voltage level of the transmission line. A second criticality factor is determined based on the importance of the area to which the power transmission equipment is located. The value of the second criticality factor is positively correlated with the load type in the area and the economic and social losses caused by power outages. A third criticality factor is determined based on the vulnerability of the power grid topology. The value of the third criticality factor is positively correlated with the node role of the transmission equipment in the power grid topology and the risk of cascading power outages caused by faults. The first criticality factor, the second criticality factor, and the third criticality factor are weighted and fused according to preset weights to obtain the equipment criticality evaluation index.

10. A method for implementing the UAV-based power transmission line inspection visualization and early warning system according to any one of claims 1-9, characterized in that, include: Step 1: Obtain multi-source correlation data, which includes power transmission equipment static attribute dataset, real-time environmental perception dataset, and historical defect maintenance dataset; Step 2: Perform deep association encoding on the multi-source associated data to generate device-environment interaction feature vectors; Step 3: Input the device-environment interaction feature vector into the time-series risk prediction model to predict the probability distribution of various types of defects in the predefined defect type set of the power transmission equipment within a future preset time window; Step 4: Based on the predicted probability distribution and combined with the equipment criticality evaluation index, determine the graded early warning level for the power transmission equipment in the current inspection target area; Based on the aforementioned graded early warning level, generate a drone inspection resource scheduling instruction corresponding to the aforementioned risk level; Step 5: Display the graded early warning level of the power transmission equipment in the current inspection target area in the entire three-dimensional visualization model; and execute the UAV inspection resource scheduling command.

Images