Direct current insulator state diagnosis method and system based on unmanned aerial vehicle

Abstract

The invention relates to the technical field of power system detection, and discloses a DC insulator state diagnosis method and system based on an unmanned aerial vehicle, and the method comprises the steps: collecting a voltage signal and temperature and image data of a target DC insulator in real time through the unmanned aerial vehicle, and carrying out the temperature compensation correction of the voltage signal through the temperature and image data, according to the correction result, the temperature and image data and historical inspection multi-dimensional data of the target DC insulator, constructing an insulator space-time correlation map; analyzing the insulator space-time correlation map, the correction result and the temperature and image data by adopting an identification prediction model obtained by training a dynamic learning rate adjustment mechanism based on space-time characteristics to obtain a real-time state diagnosis result and a degradation trend prediction result of the target direct current insulator; a traditional single-point static detection mode is broken through, and through temperature compensation, time-space correlation modeling and fusion deep learning, the accuracy, environmental adaptability and foresight of direct current insulator state diagnosis are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of power system testing technology, and in particular to a method and system for diagnosing the condition of DC insulators based on unmanned aerial vehicles (UAVs). Background Technology

[0002] In power transmission, the performance of DC insulators is crucial for ensuring the stable operation of transmission lines. With the continuous expansion of the power grid and the increasing demands for power reliability, voltage measurement and condition monitoring of DC insulators have become key aspects of power maintenance.

[0003] Currently, drones and optical voltage sensors are used for target identification, thereby locating insulators and measuring their voltage. However, this method does not consider the influence of ambient temperature on the measurement accuracy of optical voltage sensors, resulting in errors in the measurement results under different operating conditions. Furthermore, it only performs static detection at a single time point and lacks correlation analysis of the historical state data of insulators, making it impossible to capture the spatiotemporal patterns of their state evolution.

[0004] Therefore, it is evident that solving the problems of poor environmental adaptability and static-only capability of existing DC insulator testing technologies has become a pressing technical issue for those skilled in the art. Summary of the Invention

[0005] This invention provides a method and system for diagnosing the condition of DC insulators based on unmanned aerial vehicles (UAVs), which solves the problems of poor environmental adaptability and static detection capability of existing DC insulator testing technologies.

[0006] To address the aforementioned technical problems, the first aspect of this invention provides a DC insulator condition diagnosis system based on a drone, comprising a drone and an intelligent sensing module, a voltage correction module, a map construction module, and a data processing module mounted on the drone; wherein, The intelligent sensing module is used to collect voltage signals, temperature and image data of the target DC insulator in real time during the inspection process of the UAV; The voltage correction module is used to perform temperature compensation correction on the voltage signal based on the temperature and image data to obtain a corrected voltage signal. The map construction module is used to construct a spatiotemporal correlation map of the insulator based on the temperature and image data, the correction voltage signal, and the historical inspection multi-dimensional data of the target DC insulator. The data processing module is used to analyze the spatiotemporal correlation map of the insulator, the correction voltage signal, and the temperature and image data using an identification and prediction model to obtain the real-time status diagnosis results and degradation trend prediction results of the target DC insulator; the identification and prediction model is trained through a dynamic learning rate adjustment mechanism based on spatiotemporal features.

[0007] As one preferred embodiment, the intelligent sensing module includes: An ultrasonic ranging unit is used to collect the inspection trajectory data of the UAV and the distance data between the UAV and the target DC insulator when the UAV flies to the detection point of the target DC insulator, and to control the distance between the UAV and the target DC insulator within a preset range based on the distance data; An optical voltage sensor is used to acquire the voltage signal of the target DC insulator in real time during the inspection process of the UAV; A temperature sensor is used to collect the ambient temperature of the target DC insulator in real time. A camera system is used to acquire in real time thermal images of the temperature distribution and surface images of the target DC insulator during the inspection process of the UAV, as the temperature and image data.

[0008] As one preferred embodiment, the optical voltage sensor includes: An electric field sensing probe is used to sense the first potential information around the target DC insulator and convert the first potential information into a first electrical signal. A light source used to emit light signals; Pockels effect crystal, used to modulate the optical signal according to the first electrical signal, so as to change the polarization state or phase of the optical signal; An optical analysis component is used to detect the polarization or phase change of the optical signal and convert the polarization or phase change into a change in light intensity; A photoelectric detection component is used to extract second potential information from the change in light intensity and obtain a second electrical signal based on the second potential information. The signal processing sub-component is used to amplify and filter the second electrical signal, and to obtain the voltage signal of the target DC insulator using a differential measurement method.

[0009] As one preferred embodiment, the voltage correction module includes: A measurement error calculation unit is used to determine the measurement error caused by the ambient temperature based on the ambient temperature and the temperature coefficient of the temperature sensor. A temperature compensation correction unit is used to correct the voltage signal based on the measurement error to obtain the corrected voltage signal.

[0010] As one preferred embodiment, the map construction module includes: A node attribute setting unit is used to set the target DC insulator as a target node and set the attribute vector of the target node; The data association unit is used to construct spatiotemporal anchor points based on the inspection trajectory data, and to associate the historical inspection multi-dimensional data with the spatiotemporal anchor points to construct historical nodes; the historical inspection multi-dimensional data includes historical distance data, historical voltage signals, historical ambient temperature, historical temperature and image data, and historical status diagnosis results; A time edge construction unit is used to establish a time edge between the target node and the historical node, and to construct the edge weight of the time edge based on the similarity between the attribute vectors in the target node and the historical node and the inspection time difference. The spatial edge construction unit is used to take the target nodes corresponding to the target DC insulators that belong to the same inspection and the same insulator string as spatial nodes, and establish spatial edges between the spatial nodes; the electrical association weight and thermal association weight are determined based on the voltage gradient and temperature distribution between adjacent spatial nodes, respectively, and the edge weight of the spatial edge is calculated by combining the basic geometric weight determined according to the physical installation sequence of the spatial nodes. The map generation unit is used to integrate the temperature and image data, the correction voltage signal and the historical inspection multi-dimensional data based on the time edge and its edge weight, the spatial edge and its edge weight, the target node and its attribute vector, to obtain the spatiotemporal correlation map of the insulator.

[0011] As one preferred embodiment, the recognition and prediction model employs a fusion structure of a one-dimensional convolutional neural network, gated recurrent units, and graph convolutional networks; wherein, The data processing module is specifically used for: Local features in the corrected voltage signal are extracted using the one-dimensional convolutional neural network. The gated loop unit is used to capture the temporal characteristics of the correction voltage signal and the temperature and image data; The graph convolutional network is used to mine the spatial correlation features in the spatiotemporal correlation map of the insulator; The local features, temporal features, and spatial correlation features are processed by a fully connected layer and a Softmax function to output the real-time state diagnosis results and degradation trend prediction results of the target DC insulator.

[0012] As one preferred embodiment, the dynamic learning rate adjustment mechanism based on spatiotemporal features includes: The feature vector of the target DC insulator is determined based on the spatiotemporal correlation map of the insulator, and the intensity of feature change is determined based on the feature vector; The learning rate is initialized, and when the intensity of the feature change is greater than a first preset change threshold, the initialized learning rate is dynamically adjusted to a first preset multiple of the current value; when the intensity of the feature change is less than a second preset change threshold, the initialized learning rate is dynamically adjusted to a second preset multiple of the current value.

[0013] A second aspect of the present invention provides a method for diagnosing the condition of DC insulators based on unmanned aerial vehicles (UAVs), comprising: During the inspection process, the drone collects voltage signals, temperature, and image data of the target DC insulator in real time. The voltage signal is temperature-compensated and corrected based on the temperature and image data to obtain a corrected voltage signal. Based on the temperature and image data, the correction voltage signal, and the historical inspection multi-dimensional data of the target DC insulator, a spatiotemporal correlation map of the insulator is constructed. An identification and prediction model is used to analyze the spatiotemporal correlation map of the insulator, the correction voltage signal, and the temperature and image data to obtain the real-time status diagnosis results and degradation trend prediction results of the target DC insulator; the identification and prediction model is trained through a dynamic learning rate adjustment mechanism based on spatiotemporal features.

[0014] A third aspect of the present invention provides an electronic device including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the above-described method for diagnosing the condition of DC insulators based on unmanned aerial vehicles.

[0015] A fourth aspect of the present invention provides a computer-readable storage medium comprising a stored computer program, wherein when the device containing the computer-readable storage medium executes the computer program, it implements the above-described method for diagnosing the condition of DC insulators based on unmanned aerial vehicles.

[0016] Compared with the prior art, the beneficial effects of the embodiments of the present invention are as follows: By introducing temperature compensation correction for the voltage signal, the influence of ambient temperature on the measurement results of the optical voltage sensor can be eliminated in real time, significantly improving the accuracy of voltage measurement data and enhancing the system's adaptability to different climatic conditions, thus solving the problem of poor environmental adaptability of existing DC insulator detection technology. By constructing a spatiotemporal correlation map of insulators that integrates time and space dimensions, the model can learn the spatiotemporal laws of insulator state evolution. Combined with the identification and prediction model, it can not only achieve real-time state diagnosis but also predict the degradation trend for several future cycles, solving the problem that existing DC insulator detection technology can only perform static detection. The model is trained using a dynamic learning rate adjustment mechanism based on spatiotemporal features, which accelerates the model convergence speed and balances the training objectives of state classification and trend prediction, further reducing false alarms, false negatives, and trend prediction bias. Attached Figure Description

[0017] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a structural diagram of a DC insulator condition diagnosis system based on an unmanned aerial vehicle (UAV) according to a certain embodiment of the present invention; Figure 2 This is a flowchart of a DC insulator condition diagnosis method based on an unmanned aerial vehicle (UAV) according to a certain embodiment of the present invention; Figure 3 This is a structural diagram of an electronic device provided in a certain embodiment of the present invention; Figure label: Among them, 10 is the intelligent sensing module; 20 is the voltage correction module; 30 is the map construction module; 40 is the data processing module; 5000 is the electronic device; 5001 is the processor; 5002 is the bus; 5003 is the memory; and 5004 is the transceiver. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings and examples. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0020] In the description of this application, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items. Those skilled in the art will be able to understand the specific meaning of the above terms in this application according to the specific circumstances.

[0021] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is merely for describing specific embodiments and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0022] In one embodiment, such as Figure 1 As shown, the first aspect of the present invention provides a DC insulator condition diagnosis system based on a drone, including a drone and an intelligent sensing module, a voltage correction module, a map construction module, and a data processing module mounted on the drone; wherein, The intelligent sensing module is used to collect voltage signals, temperature, and image data of the target DC insulator in real time during the inspection process of the UAV. The intelligent sensing module is the "sense" of the system, integrating a high-precision optical voltage sensor, a dual-light (visible light / infrared) gimbal camera, an ultrasonic / laser rangefinder, a temperature and humidity sensor, and an environmental sensor to ensure that electrical, thermal, visual, and distance information of the target DC insulator is acquired simultaneously from multiple angles and in multiple bands.

[0023] In one embodiment, the intelligent sensing module includes: An ultrasonic ranging unit is used to collect the inspection trajectory data of the UAV and the distance data between the UAV and the target DC insulator when the UAV flies to the detection point of the target DC insulator, and to control the distance between the UAV and the target DC insulator within a preset range based on the distance data; An optical voltage sensor is used to acquire the voltage signal of the target DC insulator in real time during the inspection process of the UAV; A temperature sensor is used to collect the ambient temperature of the target DC insulator in real time. A camera system is used to acquire in real time thermal images of the temperature distribution and surface images of the target DC insulator during the inspection process of the UAV, as the temperature and image data.

[0024] Specifically, operators control a drone via a ground station to fly to the vicinity of the DC insulator string to be inspected. Using the ultrasonic ranging unit included in the intelligent sensing module, the drone is controlled to move at a preset distance from the target DC insulator string (i.e., the inspection point). The drone's control system automatically records the GPS position trajectory, altitude, attitude, time, and other data of this flight as inspection trajectory data. Because the distance between the drone and the target DC insulator string needs to remain stable to ensure the optical voltage sensor is at the optimal measurement distance, the ultrasonic ranging unit includes an ultrasonic sensor and a PID controller. The ultrasonic sensor measures the distance between the drone and the target DC insulator string in real time according to the ranging formula. The PID controller controls the distance between the drone and the target DC insulator string within a preset range (e.g., 1.5 meters) based on a distance control algorithm or existing trajectory generation algorithm.

[0025] The optical voltage sensor included in the intelligent sensing module senses the electric field around the insulator based on the Pockels effect. It converts the refractive index change of the electro-optic crystal into a light intensity signal, then inversely calculates the electric field strength, and combines this with the insulator's geometric parameters to obtain the voltage signal of the target DC insulator. This process is expressed by the following formula: In the formula, is the electric field-free refractive index of Pockels crystal; The electro-optic coefficient is related to the crystal material. The electric field strength inside the crystal; The incident light intensity; The thickness of the crystal; The wavelength of light; The geometric parameters of the target DC insulator (i.e., the thickness of the target DC insulator along the direction of the electric field inside the crystal). The terminal voltage of the target DC insulator.

[0026] The optical path design adopts an optimized anti-interference optical path structure and combines a multi-frequency sampling strategy to adapt to the voltage measurement requirements of different types of insulators.

[0027] This invention uses a temperature sensor contained in an intelligent sensing module to collect the temperature of the environment in which the target DC insulator is located in real time; the camera system in the intelligent sensing module consists of a visible light camera and an infrared camera: the visible light camera collects images of the insulator surface (reflecting appearance defects), and the infrared camera collects thermal images of the insulator temperature distribution (reflecting temperature anomalies), and the two together constitute the temperature and image data of the target DC insulator.

[0028] In one embodiment, the optical voltage sensor includes: An electric field sensing probe is used to sense the first potential information around the target DC insulator and convert the first potential information into a first electrical signal. A light source used to emit light signals; Pockels effect crystal, used to modulate the optical signal according to the first electrical signal, so as to change the polarization state or phase of the optical signal; An optical analysis component is used to detect the polarization or phase change of the optical signal and convert the polarization or phase change into a change in light intensity; A photoelectric detection component is used to extract second potential information from the change in light intensity and obtain a second electrical signal based on the second potential information. The signal processing sub-component is used to amplify and filter the second electrical signal, and to obtain the voltage signal of the target DC insulator using a differential measurement method.

[0029] Specifically, this invention employs a hemispherical metal electrode (approximately 3 cm in diameter) mounted at the end of an insulating rod, positioned near the high-voltage end of the insulator under test (approximately 10 cm away), to sense the electric field strength around the insulator and output an induced voltage signal proportional to the potential; a distributed feedback (DFB) laser diode with a wavelength of 1550 nm is used as the light source, with an output power stabilized at 10 mW; a lithium niobate (LiNbO3) crystal with dimensions of 10 mm × 2 mm × 2 mm (light propagation direction × width × thickness), z-axis cut, and antireflection coatings at both ends is selected as the Pockels effect crystal; the optical analysis component consists of a polarization beam splitter (PBS) and a λ / 4 waveplate, converting the polarization state change into the intensity difference of two orthogonally polarized lights; the optical detection component uses a pair of InGaAs photodiodes to receive the two polarized lights respectively, outputting a differential photocurrent signal; the signal processing unit includes a transimpedance amplifier, a low-pass filter (cutoff frequency 1 kHz), and a differential amplifier, ultimately outputting the voltage signal of the target DC insulator.

[0030] The electric field generated at the high-voltage end of the insulator induces a voltage V on the metal electrode of the probe. pThe light is transmitted via a shielded cable to the parallel plate electrodes on both sides of the Pockels crystal, forming an axially modulated electric field inside the crystal. The laser light is input to the crystal via a single-mode fiber, and the crystal generates a linear electro-optic effect (Pockels effect) under the action of the electric field, resulting in a phase delay in the linearly polarized light passing through the crystal. After the outgoing light passes through a λ / 4 waveplate and a PBS, the phase delay is converted into a change in light intensity. The optical signal is converted into a current signal by a photodiode, amplified by transimpedance, and then sent to a differential amplifier to obtain a voltage. After being low-pass filtered, the voltage is converted into the actual voltage of the insulator by geometric parameters (determined by the probe position and the insulator structure).

[0031] This invention utilizes optical principles and an insulated electric field sensing probe, eliminating the direct electrical connection between the sensor and the measured high-voltage DC insulator, effectively avoiding the risk of high-voltage breakdown and improving equipment and personnel safety. The optical signal is transmitted through optical fibers or optical components, unaffected by strong electromagnetic field interference, making it suitable for high electromagnetic noise environments such as high-voltage DC converter stations, and providing stable and reliable measurement results. The Pockels effect crystal exhibits fast response speed and good linearity to electric fields; combined with differential measurement methods, it significantly reduces common-mode noise such as temperature drift and light source fluctuations, achieving high-precision DC voltage measurement. The electric field sensing probe can be designed as a compact, non-contact structure, eliminating the need for a hard connection to the insulator, facilitating on-site installation and maintenance, and is particularly suitable for monitoring insulator voltage on operational DC lines.

[0032] The voltage correction module is used to perform temperature compensation correction on the voltage signal based on the temperature and image data to obtain a corrected voltage signal. Specifically, the present invention achieves voltage signal correction by calculating the measurement error of temperature and compensating for the temperature of voltage.

[0033] In one embodiment, the voltage correction module includes: A measurement error calculation unit is used to determine the measurement error caused by the ambient temperature based on the ambient temperature and the temperature coefficient of the temperature sensor. A temperature compensation correction unit is used to correct the voltage signal based on the measurement error to obtain the corrected voltage signal.

[0034] This invention determines the measurement error of temperature data by using real-time ambient temperature and a temperature sensor that collects the temperature data. This process is expressed by the following formula: In the formula, This is the temperature coefficient of the temperature sensor (unit: V / ℃, used to characterize the voltage error caused by a 1℃ temperature shift, and is an inherent parameter of the temperature sensor). For calibration temperature; This refers to the ambient temperature collected in real time.

[0035] The difference between the original measured insulator voltage signal and the measurement error was then used as the temperature compensation correction result for the voltage based on the temperature data—the corrected voltage signal—eliminating the measurement error introduced by temperature drift.

[0036] This invention introduces temperature compensation correction for voltage, thereby eliminating the influence of ambient temperature on the measurement results of the optical voltage sensor in real time, significantly improving the accuracy of voltage measurement data and enhancing the system's adaptability to different climatic conditions.

[0037] The map construction module is used to construct a spatiotemporal correlation map of the insulator based on the temperature and image data, the correction voltage signal, and the historical inspection multi-dimensional data of the target DC insulator. Specifically, this invention uses the inspection trajectory data of the UAV as a link to correlate the real-time acquired correction voltage signal, insulator temperature and image data, distance data, and historical inspection multi-dimensional data (including historical distance, historical voltage, historical temperature, historical images, and historical state diagnosis results). Based on the correlated data, a spatiotemporal correlation map of the insulator is constructed, thereby clearly presenting the state evolution trajectory of the insulator at different times and locations, providing data support for predicting the degradation trend.

[0038] In one embodiment, the atlas construction module includes: A node attribute setting unit is used to set the target DC insulator as a target node and set the attribute vector of the target node; The data association unit is used to construct spatiotemporal anchor points based on the inspection trajectory data, and to associate the historical inspection multi-dimensional data with the spatiotemporal anchor points to construct historical nodes; the historical inspection multi-dimensional data includes historical distance data, historical voltage signals, historical ambient temperature, historical temperature and image data, and historical status diagnosis results; A time edge construction unit is used to establish a time edge between the target node and the historical node, and to construct the edge weight of the time edge based on the similarity between the attribute vectors in the target node and the historical node and the inspection time difference. The spatial edge construction unit is used to take the target nodes corresponding to the target DC insulators that belong to the same inspection and the same insulator string as spatial nodes, and establish spatial edges between the spatial nodes; the electrical association weight and thermal association weight are determined based on the voltage gradient and temperature distribution between adjacent spatial nodes, respectively, and the edge weight of the spatial edge is calculated by combining the basic geometric weight determined according to the physical installation sequence of the spatial nodes. The map generation unit is used to integrate the temperature and image data, the correction voltage signal and the historical inspection multi-dimensional data based on the time edge and its edge weight, the spatial edge and its edge weight, the target node and its attribute vector, to obtain the spatiotemporal correlation map of the insulator.

[0039] Specifically, this invention uses each target DC insulator and its state snapshot at a specific inspection moment as a target node, and sets the attribute vector of the target node. That is, each target node is represented by a high-dimensional feature vector containing multimodal data, such as electrical attributes: corrected voltage signal, electric field strength feature value; thermal attributes: average temperature, maximum temperature, temperature variance, and temperature gradient of hot spot area of ​​the insulator extracted from infrared thermal image; visual attributes: depth features extracted from visible light image through pre-trained convolutional neural network (such as ResNet), which encodes appearance conditions such as degree of contamination, surface cracks, and glaze peeling; spatiotemporal metadata: inspection timestamp, GPS coordinates, UAV attitude, ambient temperature and humidity; historical diagnostic labels: status labels given by the system at the last inspection (such as "normal", "slight deterioration"), etc.

[0040] This invention labels the attribute vector of the target node with a unique insulator identifier (e.g., tower number + phase + serial number). Then, it uses the Douglas-Peucker algorithm (or other existing classic trajectory simplification algorithms) to remove redundant coordinate points (retaining key inflection points) from the recorded inspection trajectory data (GPS coordinates + timestamp). This generates spatiotemporal anchor points—each containing a unique identifier and core attributes (longitude, latitude, altitude, inspection task)—serving as the spatiotemporal reference for all subsequent data associations. Subsequently, using these spatiotemporal anchor points as a reference, it retrieves relevant historical inspection multi-dimensional data for the insulator from the historical database, including historical distance data, historical voltage signals, historical ambient temperatures, historical temperature and image data, and historical condition diagnostic results from the past N (e.g., 10) inspections. Historical nodes corresponding to this historical data are then created.

[0041] Choose any target node as the current node, and establish a directed "time edge" between the current node and each of its historical nodes, with the direction pointing from the old time point to the new time point. Simultaneously, considering the smoothness and correlation of state evolution, calculate the edge weight of the time edge. This process is expressed by the following formula: Wtime(i, j) =α×exp(-Δt / τ)+(1-α)×sim(Fi, Fj) In the formula, Wtime(i, j) is the edge weight between the i-th target node and the j-th historical node of the target node; α is a hyperparameter balancing time decay and state similarity, preferably 0.5, which can be adjusted according to actual needs; Δt is the time difference between two inspections; τ is the time decay constant, set according to the typical cycle of insulator deterioration; sim(Fi, Fj) is the similarity (such as cosine similarity) between the attribute vectors of the i-th target node and the j-th historical node of the target node, used to measure the severity of state change.

[0042] This invention treats all target nodes belonging to the same insulator string collected in the same inspection task as spatial nodes, and constructs spatial edges by analyzing the spatial dependencies between them. The voltage gradient (i.e., the absolute value of the voltage difference) between adjacent spatial nodes is calculated. Under normal circumstances, the voltage gradient between adjacent insulators should be within a small range. This invention sets an upper limit ΔVmax for the normal voltage gradient (obtained from historical data or theoretical calculations). Therefore, the electrical association weight can be defined as: Wc(m,n) =exp(-ΔV / ΔVmax) In the formula, Wc(m,n) represents the electrical association weight between the m-th and n-th spatial nodes; ΔV represents the voltage gradient. Thus, when the voltage gradient is close to 0, the weight is close to 1; when the voltage gradient exceeds the normal range, the weight decreases; when the gradient is within the normal range, a strong weighted edge is established to represent the continuity of the electrical distribution. If the voltage of an insulator is abnormal (too high or too low), the weight of its electrical edge with adjacent insulators will significantly decrease or show a negative correlation, marking an abnormal point.

[0043] This invention extracts the temperature distribution of insulators from infrared thermal images and uses the average temperature to calculate the temperature difference between spatial nodes corresponding to two insulators. An upper limit ΔTmax for the normal temperature difference is set (obtained from historical data or theoretical calculations). Therefore, the thermal correlation weight can be defined as: Wt(m,n) =exp(-ΔT / ΔTmax) In the formula, Wt(m,n) is the thermal association weight between the m-th spatial node and the n-th spatial node; ΔT is the temperature difference. A simple temperature gradient is used to establish spatial edges for nodes that are physically adjacent and whose temperature distribution conforms to the heat conduction law. Locally overheated nodes will destroy the thermal association edge weight between them and their surrounding nodes.

[0044] Since the insulator string is physically connected, adjacent insulator pieces have an inherent connection relationship. Therefore, this invention sets the basic geometric weights based on the position of the insulator pieces in the string. For example, if insulators m and n are physically adjacent, then Wg(m,n) = 1; if they are separated by one insulator, then Wg(m,n) = 0.5; if they are separated by two or more insulators, then the weight is 0. The edge weight of the spatial edge is a weighted sum of the electrical association weight, thermal association weight, and basic geometric weight, reflecting the comprehensive spatial coupling strength of insulators m and n in the multiphysics field. The sum of the three weights is 1, preferably 0.4, 0.4, and 0.2, which can be adjusted according to actual needs.

[0045] Ultimately, all nodes (target nodes and associated historical nodes) and their attribute vectors (with data), temporal edges (with edge weights), and spatial edges (with edge weights) are integrated to form a graph data structure—the insulator spatiotemporal association graph. This graph can be stored and visualized using a graph database (such as Neo4j). Furthermore, the graph is not static. After each new inspection, new nodes are added: a new node is created for each insulator inspected; temporal edges are updated: a temporal edge is established between the new node and its respective historical node; and spatial edges are updated: new spatial edges are established between all new nodes inspected. Simultaneously, the system can re-evaluate the weights of spatial edges between nodes in recent inspections and fine-tune them based on long-term patterns.

[0046] This invention expands the data points of a single insulator into a correlation network that integrates time series and spatial topology, greatly improving the comprehensiveness and accuracy of state assessment. By constructing time and spatial edges and scientifically quantifying their weights, it helps to identify progressive defects and associated fault risks that are difficult to detect using traditional methods at an early stage. The graph structure intuitively displays the data support and correlation paths for all diagnostic conclusions, enhancing the interpretability and reliability of diagnostic decisions, achieving deep integration of multi-source heterogeneous data, and fully exploring the potential value of the data.

[0047] The data processing module is used to analyze the spatiotemporal correlation map of the insulator, the correction voltage signal, and the temperature and image data using an identification and prediction model to obtain the real-time status diagnosis results and degradation trend prediction results of the target DC insulator; the identification and prediction model is trained through a dynamic learning rate adjustment mechanism based on spatiotemporal features; wherein, the identification and prediction model adopts a fusion structure of one-dimensional convolutional neural network, gated recurrent unit and graph convolutional network; In one embodiment, the data processing module is specifically used for: Local features in the corrected voltage signal are extracted using the one-dimensional convolutional neural network. The gated loop unit is used to capture the temporal characteristics of the correction voltage signal and the temperature and image data; The graph convolutional network is used to mine the spatial correlation features in the spatiotemporal correlation map of the insulator; The local features, temporal features, and spatial correlation features are processed by a fully connected layer and a Softmax function to output the real-time state diagnosis results and degradation trend prediction results of the target DC insulator.

[0048] Specifically, the spatiotemporal correlation map of the insulator, the corrected voltage signal, and the temperature and image data are input into the convolutional neural network of the recognition and prediction model through the input layer. This model uses a three-layer convolutional structure, with each layer having a kernel size of 3×1, a stride of 1, and an activation function of LeakyReLU (α=0.1) to extract local features from the corrected voltage signal. Max pooling with a pooling window of 2×1 is then employed to reduce the feature dimension of the local features. The corrected voltage signal, temperature, and image data are concatenated into a multivariate time series and input into a gated recurrent unit (containing 64 hidden units). The output of the last time step is taken as the temporal feature to mine the temporal dependency of the voltage signal and temperature data. The method employs a two-layer graph convolutional network with ReLU activation to mine spatial correlations and the correlation patterns between historical and real-time data in the spatiotemporal correlation map of insulators, thus obtaining spatial correlation features. Finally, the local features, temporal features, and spatial correlation features are passed through a fully connected layer (output dimension 5) to correspond to five insulator state labels (normal, slight degradation, moderate degradation, severe degradation, and fault) and the predicted degradation trend values ​​for the next three inspection cycles, respectively. The output layer outputs the probability distribution of the insulator state through a Softmax function and the predicted degradation trend value through a linear activation function, thereby obtaining the real-time state diagnosis results and degradation trend prediction results of the target DC insulator.

[0049] This invention integrates three networks—one-dimensional convolutional neural networks, gated recurrent units, and graph convolutional networks—to simultaneously process voltage signal waveforms, time-series data, and topological graph structure data. This enables comprehensive, multi-level feature extraction of insulator states, significantly improving the accuracy and reliability of diagnosis and prediction. The model design considers the complete lifecycle of insulator health, enabling real-time assessment of the current state (classification diagnosis) and prediction of future degradation trends (regression prediction), providing complete decision support for preventative maintenance, from current state perception to future risk warning. Through spatiotemporal feature joint modeling, the model can understand the spatial dependencies and temporal evolution of insulators within the string, effectively identifying the propagation effects and time-varying trends of local anomalies in the insulator string. End-to-end intelligent analysis reduces reliance on human expert experience, lowers the operational threshold, and improves the consistency and efficiency of diagnosis.

[0050] In one embodiment, the dynamic learning rate adjustment mechanism based on spatiotemporal features includes: The feature vector of the target DC insulator is determined based on the spatiotemporal correlation map of the insulator, and the intensity of feature change is determined based on the feature vector; The learning rate is initialized, and when the intensity of the feature change is greater than a first preset change threshold, the initialized learning rate is dynamically adjusted to a first preset multiple of the current value; when the intensity of the feature change is less than a second preset change threshold, the initialized learning rate is dynamically adjusted to a second preset multiple of the current value.

[0051] Specifically, the identification and prediction model uses an adaptive Adam optimizer during model training and introduces a dynamic learning rate adjustment mechanism based on spatiotemporal features, dynamically adjusting the learning rate according to the intensity of feature changes in the spatiotemporal correlation map of the insulator.

[0052] In this embodiment, the present invention determines the intensity of feature change by calculating the variance of node features in the spatiotemporal correlation graph. The formula for the intensity of feature change is: In the formula, denoted as the intensity of characteristic changes during the t-th inspection; N represents the number of nodes in the spatiotemporal correlation map of the insulator. This is the feature vector (including voltage, temperature, and image features) of the i-th insulator in the t-th inspection. Let be the average feature vector of the t-th inspection.

[0053] The invention initializes the learning rate to 0.001. When the intensity of feature change is greater than the first preset change threshold (0.3), the learning rate is adjusted to 1.2 times the current value; when the intensity of feature change is less than the second preset change threshold (0.1), the learning rate is adjusted to 0.8 times the current value; otherwise, the learning rate remains unchanged to accelerate convergence and avoid local optima.

[0054] This invention utilizes a dynamic learning rate adjustment mechanism based on spatiotemporal features, enabling the model training process to adaptively adjust the learning step size according to the spatiotemporal variations in data distribution. This improves training efficiency, avoids local optima, and enhances the model's generalization ability and convergence stability.

[0055] Furthermore, the loss function for the identification prediction model employs a combined loss function of weighted cross-entropy loss and trend prediction loss, where the trend prediction loss is used to constrain the deviation between the predicted and actual trends. (Combined loss function) The formula is as follows: In the formula, The weighted cross-entropy loss; Loss due to trend prediction; The balance coefficient is 0.3; the weighted cross-entropy loss formula is: In the formula, M is the number of state categories; Category weights; For true labels (0 or 1); Predict the probability of class j for the model.

[0056] The model training dataset includes 8,000 sets of historically collected "voltage-temperature-image-trajectory-state" joint data. Using 5-fold cross-validation, the validation set state diagnosis accuracy reached 97.8%, and the degradation trend prediction error was less than 5%, which significantly improved the insulator state identification accuracy and trend prediction reliability compared with traditional methods.

[0057] For example, after obtaining the real-time status diagnosis results and deterioration trend prediction results of the insulator, the present invention compresses and encrypts them, and sends the compressed and encrypted data to the remote monitoring center for corresponding decryption and decompression.

[0058] This application's embodiments, by collecting insulator distance data, insulator voltage signals, insulator temperature, and image data, combined with temperature compensation correction, spatiotemporal correlation maps, and machine learning models, break through the traditional single-time-point static detection mode, improving the accuracy, environmental adaptability, and forward-looking nature of DC insulator condition diagnosis. The proposed temperature compensation and correction can compensate and correct the insulator voltage signal based on the measurement error caused by temperature, eliminating the influence of ambient temperature on the measurement results. By constructing an insulator spatiotemporal correlation map, the spatiotemporal evolution law of the insulator state is clearly presented, providing data support for trend prediction. The proposed identification and prediction model adopts a fusion structure of one-dimensional convolutional neural network-gated recurrent unit-graph convolutional network, simultaneously extracting local spatial features, temporal dependency features, and spatial correlation features, overcoming the problem of weak feature extraction ability of traditional models, and maintaining high sensitivity even in complex environments. The proposed adaptive Adam optimizer and combined loss function further improve the model's convergence speed and prediction accuracy, reducing false alarm rate, false negative rate, and trend prediction bias. The real-time insulator condition diagnosis results and degradation trend prediction results are compressed and encrypted, improving the security and efficiency of data transmission.

[0059] This application proposes a UAV-based DC insulator condition diagnosis system to address the shortcomings of existing DC insulator detection technologies, which suffer from poor environmental adaptability and are limited to static detection. By introducing temperature compensation correction for voltage signals, the system eliminates the influence of ambient temperature on the measurement results of optical voltage sensors in real time, significantly improving the accuracy of voltage measurement data and enhancing the system's adaptability to different climatic conditions, thus solving the problem of poor environmental adaptability in existing DC insulator detection technologies. Furthermore, by constructing a spatiotemporal correlation map of insulators that integrates time and space dimensions, the model can learn the spatiotemporal patterns of insulator state evolution. Combined with an identification and prediction model, this system not only enables real-time condition diagnosis but also predicts degradation trends over several future cycles, overcoming the limitation of static detection in existing DC insulator detection technologies. Finally, a dynamic learning rate adjustment mechanism based on spatiotemporal features is used to train the model, accelerating model convergence and balancing the training objectives of state classification and trend prediction, further reducing false alarms, false negatives, and trend prediction bias.

[0060] It should be noted that the various modules in the aforementioned UAV-based DC insulator condition diagnosis system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0061] In another embodiment, such as Figure 1 As shown, a second aspect of the present invention provides a method for diagnosing the condition of DC insulators based on unmanned aerial vehicles (UAVs), comprising: S1. During the inspection process of the UAV, the voltage signal, temperature and image data of the target DC insulator are collected in real time; S2. Perform temperature compensation correction on the voltage signal based on the temperature and image data to obtain a corrected voltage signal; S3. Based on the temperature and image data, the correction voltage signal, and the historical inspection multi-dimensional data of the target DC insulator, construct a spatiotemporal correlation map of the insulator; S4. The spatiotemporal correlation map of the insulator, the correction voltage signal, and the temperature and image data are analyzed using an identification and prediction model to obtain the real-time status diagnosis results and degradation trend prediction results of the target DC insulator; the identification and prediction model is trained through a dynamic learning rate adjustment mechanism based on spatiotemporal features.

[0062] It should be noted that although the steps in the flowchart above are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise explicitly stated in this document, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. For specific limitations on a UAV-based DC insulator condition diagnosis method, please refer to the limitations on a UAV-based DC insulator condition diagnosis system above. The two have the same function and role, and will not be repeated here.

[0063] A third aspect of the present invention provides an electronic device comprising: Processor, memory, and bus; The bus is used to connect the processor and the memory; The memory is used to store operation instructions; The processor is configured to execute instructions by calling the operation instructions, causing the processor to perform operations corresponding to the UAV-based DC insulator condition diagnosis method shown in the second aspect of this application.

[0064] In one alternative embodiment, an electronic device is provided, such as Figure 3 As shown, Figure 3 The illustrated electronic device 5000 includes a processor 5001 and a memory 5003. The processor 5001 and the memory 5003 are connected, for example, via a bus 5002. Optionally, the electronic device 5000 may also include a transceiver 5004. It should be noted that in practical applications, the transceiver 5004 is not limited to one type, and the structure of this electronic device 5000 does not constitute a limitation on the embodiments of this application.

[0065] Processor 5001 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 5001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0066] Bus 5002 may include a path for transmitting information between the aforementioned components. Bus 5002 may be a PCI bus or an EISA bus, etc. Bus 5002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0067] The memory 5003 may be a ROM or other type of static storage device capable of storing static information and instructions, RAM or other type of dynamic storage device capable of storing information and instructions, or it may be an EEPROM, CD-ROM or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0068] The memory 5003 is used to store application code that executes the scheme of this application, and its execution is controlled by the processor 5001. The processor 5001 is used to execute the application code stored in the memory 5003 to implement the content shown in any of the foregoing method embodiments.

[0069] Among them, electronic devices include, but are not limited to: mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), and in-vehicle terminals (such as in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers.

[0070] The fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements a method for diagnosing the condition of a DC insulator based on an unmanned aerial vehicle (UAV) as shown in the second aspect of this application.

[0071] Another embodiment of this application provides a computer-readable storage medium storing a computer program that, when run on a computer, enables the computer to execute the corresponding content in the aforementioned method embodiments.

[0072] Furthermore, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0073] In summary, this invention relates to the field of power system detection technology, and discloses a method and system for DC insulator condition diagnosis based on unmanned aerial vehicles (UAVs). The method involves real-time acquisition of voltage signals, temperature, and image data of the target DC insulator using a UAV, followed by temperature compensation correction of the voltage signals using the temperature and image data. Based on the correction results, temperature and image data, and historical multi-dimensional inspection data of the target DC insulator, a spatiotemporal correlation map of the insulator is constructed. A recognition and prediction model trained using a dynamic learning rate adjustment mechanism based on spatiotemporal features is employed to analyze the spatiotemporal correlation map, correction results, and temperature and image data, yielding real-time condition diagnosis results and degradation trend prediction results for the target DC insulator. This method breaks through the traditional single-point static detection mode, significantly improving the accuracy, environmental adaptability, and forward-looking nature of DC insulator condition diagnosis through temperature compensation, spatiotemporal correlation modeling, and deep learning fusion.

[0074] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on its differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0075] The embodiments described above are merely preferred embodiments of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the technical principles of this invention, and these improvements and substitutions should also be considered within the scope of protection of this application. Therefore, the scope of protection of this patent application should be determined by the scope of the claims.

Claims

1. A DC insulator condition diagnosis system based on unmanned aerial vehicles (UAVs), characterized in that, This includes a drone and an intelligent sensing module, voltage correction module, map construction module, and data processing module mounted on the drone; wherein, The intelligent sensing module is used to collect voltage signals, temperature and image data of the target DC insulator in real time during the inspection process of the UAV; The voltage correction module is used to perform temperature compensation correction on the voltage signal based on the temperature and image data to obtain a corrected voltage signal. The map construction module is used to construct a spatiotemporal correlation map of the insulator based on the temperature and image data, the correction voltage signal, and the historical inspection multi-dimensional data of the target DC insulator. The data processing module is used to analyze the spatiotemporal correlation map of the insulator, the correction voltage signal, and the temperature and image data using an identification and prediction model to obtain the real-time status diagnosis results and degradation trend prediction results of the target DC insulator; the identification and prediction model is trained through a dynamic learning rate adjustment mechanism based on spatiotemporal features.

2. The DC insulator condition diagnosis system based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The intelligent sensing module includes: An ultrasonic ranging unit is used to collect the inspection trajectory data of the UAV and the distance data between the UAV and the target DC insulator when the UAV flies to the detection point of the target DC insulator, and to control the distance between the UAV and the target DC insulator within a preset range based on the distance data; An optical voltage sensor is used to acquire the voltage signal of the target DC insulator in real time during the inspection process of the UAV; A temperature sensor is used to collect the ambient temperature of the target DC insulator in real time. A camera system is used to acquire in real time thermal images of the temperature distribution and surface images of the target DC insulator during the inspection process of the UAV, as the temperature and image data.

3. The DC insulator condition diagnosis system based on unmanned aerial vehicles (UAVs) according to claim 2, characterized in that, The optical voltage sensor includes: An electric field sensing probe is used to sense the first potential information around the target DC insulator and convert the first potential information into a first electrical signal. A light source used to emit light signals; Pockels effect crystal, used to modulate the optical signal according to the first electrical signal, so as to change the polarization state or phase of the optical signal; An optical analysis component is used to detect the polarization or phase change of the optical signal and convert the polarization or phase change into a change in light intensity; A photoelectric detection component is used to extract second potential information from the change in light intensity and obtain a second electrical signal based on the second potential information. The signal processing sub-component is used to amplify and filter the second electrical signal, and to obtain the voltage signal of the target DC insulator using a differential measurement method.

4. The DC insulator condition diagnosis system based on unmanned aerial vehicles (UAVs) according to claim 2, characterized in that, The voltage correction module includes: A measurement error calculation unit is used to determine the measurement error caused by the ambient temperature based on the ambient temperature and the temperature coefficient of the temperature sensor. A temperature compensation correction unit is used to correct the voltage signal based on the measurement error to obtain the corrected voltage signal.

5. A DC insulator condition diagnosis system based on an unmanned aerial vehicle (UAV) according to claim 2, characterized in that, The map construction module includes: A node attribute setting unit is used to set the target DC insulator as a target node and set the attribute vector of the target node; The data association unit is used to construct spatiotemporal anchor points based on the inspection trajectory data, and to associate the historical inspection multi-dimensional data with the spatiotemporal anchor points to construct historical nodes; the historical inspection multi-dimensional data includes historical distance data, historical voltage signals, historical ambient temperature, historical temperature and image data, and historical status diagnosis results; A time edge construction unit is used to establish a time edge between the target node and the historical node, and to construct the edge weight of the time edge based on the similarity between the attribute vectors in the target node and the historical node and the inspection time difference. The spatial edge construction unit is used to take the target nodes corresponding to the target DC insulators that belong to the same inspection and the same insulator string as spatial nodes, and establish spatial edges between the spatial nodes; the electrical association weight and thermal association weight are determined based on the voltage gradient and temperature distribution between adjacent spatial nodes, respectively, and the edge weight of the spatial edge is calculated by combining the basic geometric weight determined according to the physical installation sequence of the spatial nodes. The map generation unit is used to integrate the temperature and image data, the correction voltage signal and the historical inspection multi-dimensional data based on the time edge and its edge weight, the spatial edge and its edge weight, the target node and its attribute vector, to obtain the spatiotemporal correlation map of the insulator.

6. The DC insulator condition diagnosis system based on unmanned aerial vehicles according to claim 1, characterized in that, The recognition and prediction model employs a fusion structure of a one-dimensional convolutional neural network, gated recurrent units, and graph convolutional networks; wherein, The data processing module is specifically used for: Local features in the corrected voltage signal are extracted using the one-dimensional convolutional neural network. The gated loop unit is used to capture the temporal characteristics of the correction voltage signal and the temperature and image data; The graph convolutional network is used to mine the spatial correlation features in the spatiotemporal correlation map of the insulator; The local features, temporal features, and spatial correlation features are processed by a fully connected layer and a Softmax function to output the real-time state diagnosis results and degradation trend prediction results of the target DC insulator.

7. The DC insulator condition diagnosis system based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The dynamic learning rate adjustment mechanism based on spatiotemporal features includes: The feature vector of the target DC insulator is determined based on the spatiotemporal correlation map of the insulator, and the intensity of feature change is determined based on the feature vector; The learning rate is initialized, and when the intensity of the feature change is greater than a first preset change threshold, the initialized learning rate is dynamically adjusted to a first preset multiple of the current value; when the intensity of the feature change is less than a second preset change threshold, the initialized learning rate is dynamically adjusted to a second preset multiple of the current value.

8. A method for diagnosing the condition of DC insulators based on unmanned aerial vehicles (UAVs), characterized in that, include: During the inspection process, the drone collects voltage signals, temperature, and image data of the target DC insulator in real time. The voltage signal is temperature-compensated and corrected based on the temperature and image data to obtain a corrected voltage signal. Based on the temperature and image data, the correction voltage signal, and the historical inspection multi-dimensional data of the target DC insulator, a spatiotemporal correlation map of the insulator is constructed. An identification and prediction model is used to analyze the spatiotemporal correlation map of the insulator, the correction voltage signal, and the temperature and image data to obtain the real-time status diagnosis results and degradation trend prediction results of the target DC insulator; the identification and prediction model is trained through a dynamic learning rate adjustment mechanism based on spatiotemporal features.

9. An electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the UAV-based DC insulator condition diagnosis method as described in claim 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein when the device containing the computer-readable storage medium executes the computer program, it implements the UAV-based DC insulator condition diagnosis method as described in claim 8.

Images