State monitoring method, device and medium for high-voltage isolating switch operating mechanism

Abstract

The application discloses a kind of state monitoring method, equipment and medium for high-voltage disconnecting switch operating mechanism;Method includes the following steps: obtaining the multimodal data of high-voltage disconnecting switch, the multimodal data at least includes video data and corner data;The multimodal data is extracted to obtain multimodal feature;The multimodal feature is input into multimodal feature fusion diagnostic model, and the working state of high-voltage disconnecting switch operating mechanism is judged according to the multidimensional feature sequence output by the multimodal feature fusion diagnostic model.The application ensures that multimodal data is strictly aligned in time dimension through synchronous trigger mechanism;Then extract multimodal feature and input multimodal feature fusion diagnostic model to output multidimensional feature sequence, so as to realize the fine analysis and intelligent diagnosis of high-voltage disconnecting switch operating mechanism running state.

Description

Technical Field

[0001] This invention relates to the field of intelligent fault diagnosis technology, specifically to a method, device, and medium for monitoring the condition of a high-voltage disconnector switch operating mechanism. Background Technology

[0002] High-voltage disconnect switches, as key equipment in power systems, primarily function to isolate circuits, ensure maintenance safety, and cooperate with circuit breakers to complete switching operations. Their core function is to disconnect circuits under conditions of no current or extremely low current, thereby ensuring the safety of power equipment maintenance work. However, their operating mechanisms operate under complex conditions for extended periods, making them prone to failure due to mechanical wear, lubrication failure, component deformation, or environmental corrosion. Therefore, how to effectively monitor these faults in real time has become an important research direction.

[0003] Specifically, existing monitoring methods generally rely on periodic checks of single electrical quantities, such as monitoring motor current, voltage, or the on / off signals of limit switches to determine whether the mechanism has completed its action; when the current curve shows abnormal peaks or waveform distortion, the system will determine that there is a potential fault. However, these methods can only reflect macroscopic anomalies such as motor current fluctuations and action timeouts, but cannot perform quantitative analysis of the subtle mechanical changes in the internal transmission mechanism.

[0004] Especially when the disconnector switch experiences slight jamming, loose transmission components, or poor lubrication in the early stages of abnormal operation, the changes in motor current are often not obvious and are easily affected by external factors such as ambient temperature and load fluctuations. Therefore, the above-mentioned monitoring methods based on electrical quantities are prone to deviations in their judgment results, and may even lead to misjudgment or omission. As a result, it is difficult to effectively monitor the actual fault status of the operating mechanism. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a method, device and medium for monitoring the status of a high-voltage disconnector operating mechanism.

[0006] The first aspect of this invention provides a method for monitoring the status of the operating mechanism of a high-voltage disconnector, comprising the following steps: Acquire multimodal data of the high-voltage disconnector, wherein the multimodal data includes at least video data and rotation angle data; Feature extraction is performed on the multimodal data to obtain multimodal features; the multimodal features include at least video features and corner-turning temporal features; The multimodal features are input into the multimodal feature fusion diagnostic model, and the working status of the high-voltage disconnector operating mechanism is determined based on the multidimensional feature sequence output by the multimodal feature fusion diagnostic model.

[0007] Furthermore, the multimodal data of the high-voltage disconnector is acquired through a preset sensor; wherein, the video data is acquired through a preset camera, and the rotation angle data is acquired through a preset angle sensor.

[0008] Furthermore, the multi-mode data of the high-voltage disconnector is acquired through a synchronous triggering mechanism, specifically including the following steps: Upon receiving the operation command from the high-voltage disconnect switch, the preset camera and preset angle sensor are simultaneously triggered; The video data is acquired based on the preset camera, and the corner data is acquired based on the angle sensor; The video data and the cornering data are time-synchronized and calibrated so that the multimodal data have a unified timestamp reference.

[0009] Furthermore, the feature extraction of the multimodal data includes video feature extraction; the video feature extraction specifically includes the following steps: The video data is divided into multiple video frames; The target components in the high-voltage disconnect switch are located in multiple video frames based on computer vision algorithms; the target components include the main disconnect switch, the connecting rod, and the target drive shaft. The target component is tracked frame by frame using optical flow, and the visual motion features corresponding to the target component are quantified and calculated. The visual motion features include at least the opening and closing angular velocity features of the main switch, the displacement curve features, and the motion stability features.

[0010] Furthermore, the feature extraction of the multimodal data includes corner temporal feature extraction; the corner temporal feature extraction specifically includes the following steps: The corner data is preprocessed by filtering and noise reduction. The rotation timing features of the driving core state are calculated based on the preprocessed rotation data; the rotation timing features include spindle angular velocity features, spindle acceleration features, and positioning features; the positioning features are used to characterize the accuracy of the component's action endpoint reaching the preset target position.

[0011] Furthermore, the multimodal feature fusion diagnostic model is trained through the following steps: Acquire training sample data corresponding to the high-voltage disconnect switch under multiple operating modes; the multiple operating modes include normal operation mode, foreign object jamming mode, transmission component disengagement mode, motor stall mode, and component loosening mode; A multimodal neural network structure is established based on the training sample data; the multimodal neural network structure specifically includes a convolutional neural network layer for learning video sample features, a long short-term memory network layer for learning temporal features of corner samples, and a fully connected layer for outputting the multidimensional feature sequence; The training sample data is input into the multimodal neural network structure; the parameters of the multimodal neural network structure are adjusted according to its output. Return to the step of inputting the training sample data into the multimodal neural network structure, and re-input the training sample data into the multimodal neural network structure until the output of the multimodal neural network structure converges to a preset range; After the output of the multimodal neural network structure converges to a preset range, the multimodal feature fusion diagnostic model is constructed based on the multimodal neural network structure.

[0012] Furthermore, the step of determining the operating state of the high-voltage disconnector operating mechanism based on the multi-dimensional feature sequence output by the multi-modal feature fusion diagnostic model specifically includes the following steps: Calculate the confidence value of the multidimensional feature sequence for each of the operating modes; Determine whether any of the confidence values ​​is greater than or equal to a preset threshold; If it is confirmed that any value in the confidence level is greater than or equal to the preset threshold, and the operating mode corresponding to the confidence level is not a normal operating mode, then the operating mode corresponding to the confidence level is output as the fault type corresponding to the target high-voltage disconnect switch.

[0013] Furthermore, after the step of determining the working state of the high-voltage disconnector operating mechanism based on the multi-dimensional feature sequence output by the multi-modal feature fusion diagnostic model, the following steps are also included: A fault report for the high-voltage disconnector operating mechanism is generated based on the fault diagnosis results. The fault report includes equipment information of the high-voltage disconnector, fault type, characteristic data screenshots, and handling suggestions. The fault report of the disconnector switch operating mechanism is sent to the corresponding maintenance personnel of the high-voltage disconnector switch according to a preset method; the preset method includes at least PDF export, email push and SMS push.

[0014] Another aspect of the present invention discloses an electronic device, including a processor, a communication bus, a user interface, a network interface, and a memory. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the above-described method for monitoring the status of a high-voltage disconnector operating mechanism.

[0015] In another aspect, the present invention discloses a computer-readable storage medium storing a program that is executed by a processor to implement the above-described method for monitoring the status of a high-voltage disconnector operating mechanism.

[0016] This invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the aforementioned method.

[0017] The embodiments of the present invention have the following beneficial effects: The present invention provides a method, device, and medium for monitoring the state of a high-voltage disconnector switch operating mechanism. Through a synchronous triggering mechanism, a camera and angle sensor are simultaneously activated upon receiving an operation command to acquire video and rotation data, and time synchronization calibration is performed to ensure strict alignment of multimodal data in the time dimension. Subsequently, after extracting visual motion features from the video and rotation time-series features from the rotation data, the present invention inputs these multimodal features into a multimodal feature fusion diagnostic model to output a multidimensional feature sequence, thereby achieving refined analysis and intelligent diagnosis of the operating state of the high-voltage disconnector switch operating mechanism. Simultaneously, based on training sample data from multiple operating modes, the present invention can construct a multimodal neural network including convolutional neural networks and long short-term memory networks. Through parameter optimization training convergence, it improves the accuracy and generalization ability of identifying faults such as foreign object jamming and transmission component disengagement, and comprehensively evaluates the dynamic consistency and structural stability during the switch operation process.

[0018] Additional aspects and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description or may be learned by practice of the invention. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the steps of a status monitoring method for the operating mechanism of a high-voltage disconnect switch according to the present invention; Figure 2 This is a schematic diagram of the feature learning process of the multimodal feature fusion diagnostic model in the method of the present invention; Figure 3This is a schematic diagram of the modules in which the method of the present invention is applied to the system; Figure 4 This is a schematic diagram of the structure of an electronic device provided by the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0022] High-voltage disconnect switches are indispensable key equipment in power systems, mainly used for isolating circuits, ensuring maintenance safety, and cooperating with circuit breakers to complete switching operations. Their core function is to disconnect circuits under conditions of no current or extremely low current, ensuring safety during power equipment maintenance. The operating mechanism of disconnect switches is constantly exposed to complex operating conditions, making it susceptible to malfunctions due to mechanical wear, lubrication failure, component deformation, or environmental corrosion. Therefore, real-time and effective monitoring of these malfunctions has become a major research focus.

[0023] Existing methods often rely on monitoring single electrical quantities for periodic fault diagnosis. These methods determine whether the mechanism has completed its action by monitoring motor current, voltage, or the on / off signals of limit switches. When abnormal peaks or waveform distortions are detected in the current curve, the system identifies a potential fault. However, these methods only reflect macroscopic anomalies such as motor current fluctuations and timeouts, and cannot quantitatively analyze the subtle mechanical changes within the high-voltage disconnector's internal transmission mechanism. Especially under early abnormal conditions such as slight jamming, loose transmission components, or poor lubrication, the judgment results of these methods are often biased, even leading to misjudgments or omissions, due to the insignificant changes in motor current or the influence of external factors such as ambient temperature and load fluctuations. This makes it impossible to effectively monitor the actual faults in the disconnector's operating mechanism.

[0024] To overcome the above problems or at least provide some solutions, the first embodiment of the present invention provides a method for monitoring the status of the operating mechanism of a high-voltage disconnector, such as... Figure 1 As shown, it includes the following steps: S1. Acquire multi-mode data of the high-voltage disconnector, including at least video data and rotation angle data; S2. Extract features from the multimodal data to obtain multimodal features; the multimodal features include at least video features and corner-to-corner temporal features; S3. Input the multimodal features into the multimodal feature fusion diagnostic model, and determine the working status of the high-voltage disconnector operating mechanism based on the multidimensional feature sequence output by the multimodal feature fusion diagnostic model.

[0025] The implementation process of each step of this invention is described in detail below: The operating mechanism of a high-voltage disconnector includes mechanical components such as a motor, gearbox, connecting rod, and drive shaft. Because high-voltage disconnectors operate under complex conditions including high humidity, high temperature, dust, and strong electromagnetic interference, they are prone to malfunctions such as jamming, incomplete operation, or disengagement due to mechanical wear, lubrication failure, component deformation, or environmental corrosion. Current technologies typically rely on single electrical quantity monitoring or video detection for fault identification. The former can only reflect macroscopic anomalies such as motor current fluctuations and timeouts, making it difficult to accurately determine the specific location of mechanical failure. The latter is affected by ambient light, obstructions, or image blur, making quantitative analysis of internal transmission characteristics impossible and leading to unstable judgment results.

[0026] Therefore, this embodiment of the invention first acquires multimodal data of the high-voltage disconnector, including video data and rotation angle data. Preferably, the multimodal data includes video data and rotation angle data. The video data reflects the macroscopic motion characteristics of the operating mechanism, including the opening and closing process of the main disconnector, the linkage trajectory, and the integrity of the mechanism's surface structure; the rotation angle data reflects the dynamic characteristics of the drive shaft, including the real-time rotation angle of the main shaft, changes in angular velocity, and the termination angle of the action, thereby achieving precise capture of the mechanical action state.

[0027] In some embodiments, the control system of the high-voltage disconnector sends an action command to the operating mechanism before performing opening and closing operations. Therefore, this embodiment of the invention obtains multi-mode data of the high-voltage disconnector through a synchronous triggering mechanism, specifically including the following steps: S1-1. Upon receiving the operation command from the high-voltage disconnect switch, the preset camera and preset angle sensor are simultaneously triggered; S1-2. Acquire video data based on the preset camera and obtain rotation data based on the angle sensor; S1-3. Time synchronization calibration of video data and corner data is performed to ensure that multimodal data has a unified timestamp reference.

[0028] In this embodiment of the invention, a synchronous triggering unit simultaneously activates a preset camera and a preset angle sensor, enabling both data sources to begin recording at the same time. This ensures that the mechanical movement process recorded in the video corresponds perfectly to the angle change process of the rotation data. The synchronous triggering unit uses a BeiDou / GPS dual-mode timing module, which is connected to the camera and sensor respectively to achieve timestamp synchronization.

[0029] In a preferred embodiment, a preset camera is positioned in an unobstructed location that allows simultaneous observation of the main switch, connecting rod, and drive shaft areas of the operating mechanism. Its frame rate is set to at least 30 frames per second to record the complete movement of the operating mechanism from its initial position to the target position. An angle sensor, mounted on the motor spindle and employing a high-precision magnetic scale structure, measures the spindle's rotation angle in real time during operation, outputting a continuous angle time-series signal. The preset camera is a high-definition industrial camera installed near the operating mechanism of the target high-voltage disconnect switch to acquire dynamic video data of the operating mechanism's movement. Preferably, the preset camera is a CCD-type industrial camera with a 2 / 3-inch CCD chip image sensor, a resolution of at least 3312×2496 pixels, a signal-to-noise ratio greater than 50dB, and automatic exposure and low-light compensation functions to ensure stable output of high-quality video signals even in strong outdoor light, shadow, and low-light environments. The camera's frame rate is set to 30 to 60 frames per second, and the I / O interface uses a 12-pin industrial interface to achieve hardware triggering connection with the time synchronization module. The camera is installed 3 to 5 meters to the side of the operating mechanism housing, with the lens aimed at the moving plane of the main knife switch and the connecting rod, ensuring that the field of view covers the knife switch arm, the transmission connecting rod and the main shaft connection point.

[0030] In a preferred embodiment, the angle sensor is a high-precision magnetic scale angle sensor mounted on the main drive shaft of the target high-voltage disconnector operating mechanism, used to collect data on the rotational angle changes of the main shaft during operation. The angle sensor consists of two parts: a magnetic scale and a magnetic head. The magnetic scale is fixedly connected to the outer surface of the main drive shaft and generates a displacement signal as the shaft rotates. The magnetic head reads the magnetic pole changes of the magnetic scale through a Hall element and outputs orthogonal coded pulse signals, thereby achieving high-precision measurement of angular displacement. The resolution of the magnetic scale is preferably 10 μm, corresponding to an angle measurement resolution better than 0.05°, and the sampling frequency is set to 100 Hz to ensure complete capture of rapid opening and closing actions. The orthogonal coded pulses output by the sensor are converted into digital signals by a signal conditioning module and then transmitted to the edge computing unit via an RS485 bus for data storage and analysis.

[0031] S2. Extract features from the multimodal data to obtain multimodal features; the multimodal features include at least video features and corner time-series features.

[0032] In step S2, feature extraction is performed on the multimodal data, including video feature extraction; video feature extraction specifically includes the following steps: S2-A1. Divide the video data into multiple video frames; S2-A2. Based on computer vision algorithms, target components in a high-voltage disconnector are located in multiple video frames; the target components include the main disconnector, connecting rod, and target drive shaft. S2-A3. The target component is tracked frame by frame using optical flow method, and the visual motion characteristics corresponding to the target component are quantified and calculated; the visual motion characteristics include at least the opening and closing angular velocity characteristics of the main switch, the displacement curve characteristics, and the motion stability characteristics.

[0033] In this embodiment of the invention, after the video data is divided into multiple video frames, the target components of the target high-voltage disconnect switch are located in the multiple video frames based on computer vision algorithms. The target components include the main disconnect switch, the connecting rod, and the target drive shaft. The target components are tracked frame by frame using optical flow method, and the visual motion characteristics corresponding to the target components are quantitatively calculated, including the opening and closing angular velocity characteristics of the main disconnect switch, the displacement curve characteristics, and the motion stability characteristics.

[0034] In step S2, feature extraction is performed on the multimodal data, including corner time series feature extraction; the corner time series feature extraction specifically includes the following steps: S2-B1. Preprocessing of corner data through filtering and noise reduction; S2-B2. Calculate the rotation timing characteristics of the core state based on the preprocessed rotation data; the rotation timing characteristics include spindle angular velocity characteristics, spindle acceleration characteristics, and positioning characteristics; the positioning characteristics are used to characterize the accuracy of the component's action endpoint reaching the preset target position.

[0035] For example, the present invention first smooths and suppresses noise in the acquired corner time series signal, assuming the original corner sequence is... The sampling time interval is The total number of sampling frames is T. Noise suppression is achieved through one-dimensional Kalman filtering. The state equation and observation equation are as follows: ; in, The angle estimate at time t. Angular velocity is the state variable. and The system noise and observation noise are respectively, both following a zero-mean Gaussian distribution; observation sequence This refers to the angle observation value output by the sensor.

[0036] The filtering result yields a smoothed angle sequence. The principal axis angular velocity characteristic is calculated using the angle difference between adjacent time points: ; in, The main spindle angular velocity characteristic is used to reflect the change in the spindle rotation rate, and the positive and negative values ​​correspond to the clockwise and counterclockwise directions; The time interval is provided by the synchronization triggering unit.

[0037] The principal axis acceleration characteristics are obtained through angular velocity difference: ; in, The main axis acceleration characteristic is used to reflect the rate of change of angular velocity and the smoothness of the rotation process.

[0038] At the moment of termination of the action Obtain the spindle termination angle. The preset target angle is Define the feature for: ; in The positioning feature is used to characterize the accuracy with which the end point of a component's movement reaches the preset target position. The value range is (0,1], and the closer the value is to 1, the more accurate it is.

[0039] The above calculations yielded... , and These correspond to the spindle angular velocity characteristics, spindle acceleration characteristics, and positioning characteristics, respectively. Together, they constitute the drive state characteristic set of the drive core state, which is used to reflect the dynamic response characteristics inside the operating mechanism.

[0040] In some embodiments, the heterogeneous data collected synchronously can be denoised, standardized, and structured to extract key feature parameters that accurately reflect the operating status of the disconnector switch operating mechanism, providing a unified data input basis for subsequent multimodal feature fusion and fault diagnosis.

[0041] In this embodiment of the invention, external motion information from the video channel and internal driving information from the corner channel are jointly modeled on a unified time scale to form a high-dimensional feature expression that can comprehensively characterize the operating state of the disconnector switch operating mechanism.

[0042] S3. Input the multimodal features into the multimodal feature fusion diagnostic model, and determine the working status of the high-voltage disconnector operating mechanism based on the multidimensional feature sequence output by the multimodal feature fusion diagnostic model.

[0043] Preferably, the multimodal feature fusion diagnostic model used in this embodiment of the invention is trained through the following steps: S3-1. Obtain training sample data for high-voltage disconnect switches under various operating modes; the various operating modes include normal operation mode, foreign object jamming mode, transmission component disengagement mode, motor stall mode, and component loosening mode; S3-2. Establish a multimodal neural network structure based on training sample data; the multimodal neural network structure specifically includes a convolutional neural network layer for learning video sample features, a long short-term memory network layer for learning temporal features of corner samples, and a fully connected layer for outputting multidimensional feature sequences; S3-3. Input the training sample data into the multimodal neural network structure; adjust the parameters of the multimodal neural network structure according to its output; S3-4. Return to the step of inputting training sample data into the multimodal neural network structure, and re-input the training sample data into the multimodal neural network structure until the output of the multimodal neural network structure converges to the preset range; S3-5. After the output of the multimodal neural network structure converges to a preset range, construct a multimodal feature fusion diagnostic model based on the multimodal neural network structure.

[0044] In this embodiment of the invention, to construct a multimodal feature fusion diagnostic model, offline training sample datasets of the target high-voltage disconnect switch under various operating modes are first collected. These operating modes include normal operation mode, foreign object jamming mode, transmission component disengagement mode, motor stall mode, and component loosening mode. Video sample sequences are simultaneously collected under each mode. With corner sample sequence The corresponding visual motion feature sequences and corner temporal feature sequences are obtained through the processing steps S1 and S2, respectively. Each sample data is labeled with a running mode label. Each integer corresponds to one of the five operating modes mentioned above. The training set is denoted as . ,in For the sample size, Indicates the first The video feature matrix of the sample, Indicates the first The corner feature matrix of the sample, Set its corresponding tag.

[0045] The feature learning process for a multimodal feature fusion diagnostic model is as follows: Figure 3 As shown. Based on the above training sample data, a multimodal neural network structure is established. This network consists of a visual branch, a driving branch, and a fully connected output layer. The visual branch uses a convolutional neural network (CNN) structure for video sample feature learning, and its feature extraction function is defined as: ; in The output feature vector of the visual branch. The input video sample feature matrix, and These represent the kernel weights and bias parameters, respectively, and the function... This represents a composite mapping of convolution, non-linear activation, and pooling. The size of the convolution kernel window is set according to the spatial resolution of the video features. The number of convolutional layers is generally three to five. The output of each layer is activated by the ReLU function before entering the next convolution or pooling layer.

[0046] The driving branch uses a Long Short-Term Memory (LSTM) network structure for learning temporal features of corner samples, and its state update process is defined as follows: ； ； ； ； ； ； In the formula, For a moment The driving state characteristic input vector is composed of the principal axis angular velocity characteristics. Main shaft acceleration characteristics and positioning characteristics composition; This is the hidden state output vector of the LSTM; This is the unit state; These are the input gate, the forget gate, and the output gate, respectively. Candidate cell state; symbol This indicates element-wise multiplication; Represents the Sigmoid function; It is the hyperbolic tangent function; and This is the weight matrix. This is the bias vector. Through this recursive relationship, LSTM can learn the dynamic dependence and nonlinear changes of corner data in the time dimension.

[0047] Feature vectors output by the visual branch and the driving branch With the hidden state at the final moment In the feature fusion layer, features are concatenated to obtain a joint feature representation: ; The semicolon ";" indicates a vector-level concatenation operation. This is a multimodal joint feature vector. Input is fed into a fully connected layer to output a multidimensional feature sequence: ; in The output is a multidimensional feature sequence. These are the weight matrix and bias vector of the fully connected layer, respectively. The activation function is typically ReLU or Sigmoid. This output is used in subsequent classification layers to calculate confidence values ​​and perform diagnostic inference.

[0048] The model training uses the cross-entropy loss function: ; in The total number of operating modes, For the model to predict the first The sample belongs to the first The probability of a class This is the one-hot encoding corresponding to the actual label.

[0049] The model parameters are iteratively optimized by running a correction mechanism, and the weights are updated using the Adam optimization algorithm. The parameter adjustment rules are as follows: ; in The model parameter set, including wait; The learning rate has a range of values. ; These are the first and second moment estimates of the gradient, respectively. To prevent small constants from being divided by zero, a correction mechanism is implemented to dynamically adjust the learning rate based on the loss trend on the validation set after each training epoch. And when performance degrades, it rolls back to the previous optimal parameter state to prevent overfitting.

[0050] In step S3, the operating status of the high-voltage disconnector operating mechanism is determined based on the multi-dimensional feature sequence output by the multi-modal feature fusion diagnostic model. This specifically includes the following steps: S3-6. Calculate the confidence value of the multidimensional feature sequence for each operating mode; S3-7. Determine whether any confidence value is greater than or equal to a preset threshold; S3-8. If it is confirmed that any value in the confidence level is greater than or equal to the preset threshold, and the operating mode corresponding to the confidence level is not the normal operating mode, then output the operating mode corresponding to the confidence level as the fault type corresponding to the target high-voltage disconnector.

[0051] In this embodiment of the invention, after the model training is completed, the real-time collected feature sequences are inferred and calculated to generate diagnostic conclusions, thereby realizing the mapping of multimodal data to fault types.

[0052] Preferably, based on the multimodal feature fusion diagnostic model, the confidence value of the multidimensional feature sequence corresponding to each operating mode is calculated; it is determined whether any value in the confidence value is greater than or equal to a preset threshold; if it is confirmed that any value in the confidence value is greater than or equal to the preset threshold, and the operating mode corresponding to the confidence value is not a normal operating mode, then the operating mode corresponding to the confidence value is output as the fault type corresponding to the target high-voltage disconnector.

[0053] Specifically, multidimensional feature sequences The input is fed into a trained and converged multimodal feature fusion diagnostic model, and the confidence score set for each operating mode is calculated through the model's output layer. ,in This represents the total number of operating modes, including normal operation modes and various failure modes. The model output layer uses the Softmax activation function, which converts the linear output vector... The confidence score, mapped to probabilistic form, is defined as follows: ; in To determine whether the target high-voltage disconnect switch belongs to the first category in the model Confidence values ​​for each operating mode; Multidimensional feature sequences After linear mapping of the fully connected layer, the first The component reflects the model's response to the first component. The original response strength of the class; the denominator is the sum of the indexed scores of all classes, used for normalization to ensure that all confidence values ​​satisfy... .

[0054] Subsequently, the set of confidence values ​​is evaluated to detect whether any value is greater than or equal to a preset threshold. The situation. Threshold. The value range is (0,1), and it is used to limit the minimum confidence level of the model output, for example... This indicates that a pattern is considered valid only if the model's prediction confidence level for a given pattern is not less than 90%. Define the decision function. as follows: ; in, This represents the index of the operating mode corresponding to the confidence value that meets the conditions. An index representing the normal operating mode. If and If so, the operating mode corresponding to the confidence level value is output as the fault type; if If the target high-voltage disconnect switch is found to be in a normal state, then the inference output can be determined to be normal. To ensure the stability of the inference output, a moving average confidence level can be introduced. To suppress instantaneous fluctuations, it is defined as: ; in The length of the sliding window. This is the current time step. It is used to smooth the diagnostic results of multiple consecutive frames of data, thereby maintaining the temporal continuity of the output results in the event of noise or transient anomalies.

[0055] The model inference process is as follows: Input multidimensional feature sequence Calculate the linear output vector ,in and These are the output layer weight matrix and bias vector, respectively; Substitute into the Softmax function to calculate the confidence level Compare all With threshold If the maximum confidence level Greater than or equal to And the corresponding mode is not the normal operating mode. If the condition is met, the corresponding pattern label is output as the fault type; otherwise, the normal state is output. Using this algorithm, the model can classify and reason about the input multimodal features under various operating modes, and achieve reliable automatic fault determination using a confidence threshold as the decision boundary.

[0056] In some embodiments, after determining the working state of the high-voltage disconnector operating mechanism based on the multi-dimensional feature sequence output by the multi-modal feature fusion diagnostic model, the following steps are also included: S3-9. Generate a fault report for the high-voltage disconnector operating mechanism based on the fault diagnosis results; the fault report for the disconnector operating mechanism includes equipment information of the high-voltage disconnector, fault type, characteristic data screenshots, and handling suggestions; S3-10. Send the fault report of the disconnector switch operating mechanism to the corresponding maintenance personnel of the high-voltage disconnector switch according to the preset method; the preset method includes at least PDF export, email push and SMS push.

[0057] In this embodiment of the invention, after generating the fault diagnosis result, a fault report of the disconnector switch operating mechanism is automatically generated based on the output operating mode and its corresponding confidence information. The fault report generation process includes three stages: result parsing, information integration, and formatted output. First, the operating status of the target high-voltage disconnector switch is determined based on the output results of the multimodal feature fusion diagnostic model. If the operating mode is determined to be a normal operating mode, a normal operating report is generated; if it is determined to be an abnormal mode, the corresponding fault cause description and key feature indicators are extracted based on the fault type label.

[0058] During the information integration phase, representative feature data fragments are extracted from the multidimensional feature sequence. The visual motion features in the video channel and the driving state features in the corner channel are time-aligned, and image data of key frames before and after the fault occurs are automatically captured to generate feature data screenshots as auxiliary diagnostic basis. At the same time, the equipment information of the target disconnect switch is retrieved from the equipment database, including equipment number, installation location, model parameters, running time and maintenance records, to ensure that the report content is complete and traceable.

[0059] During the report generation phase, equipment information, fault type, confidence level, characteristic data screenshots, and handling suggestions are integrated into a structured report with a unified format. Handling suggestions are automatically generated based on the fault type. For example, when the fault type is "foreign object jamming mode," maintenance suggestions to remove foreign objects or check the lubrication status of the transmission mechanism are added to the report; when the fault type is "motor stall mode," a prompt to check the motor control circuit and mechanical blockage is given. The report can be appended with a timestamp, diagnostic number, and model version number for subsequent tracking and comparison.

[0060] The generated fault reports of the disconnector switch operating mechanism support multiple push methods and are automatically distributed according to the operation and maintenance strategy: on the one hand, the reports are archived and stored in the operation and maintenance platform database as PDF files for subsequent statistical analysis and model retraining; on the other hand, the reports can be sent to the registered email addresses of the target operation and maintenance personnel via email push, and a report summary and viewing link can be sent via SMS push, ensuring that key fault information can be transmitted to on-site or remote maintenance personnel in the shortest possible time, realizing intelligent closed-loop fault management of high-voltage disconnectors.

[0061] In summary, the embodiments of the present invention have at least the following beneficial effects: 1. This method acquires multimodal data of the target high-voltage disconnector switch at a unified timestamp reference through a synchronous triggering mechanism; preprocesses the multimodal data, extracting visual motion features from the video data and driving state features from the cornering data; inputs the visual motion features and driving state features into a multimodal feature fusion diagnostic model, and outputs a multidimensional feature sequence after feature fusion; based on the multidimensional feature sequence, the multimodal feature fusion diagnostic model outputs the corresponding fault diagnosis results of the target high-voltage disconnector switch, thereby achieving refined analysis and intelligent diagnosis of the operating status of the high-voltage disconnector switch operating mechanism. This method can fuse visual motion features reflecting the external motion state of the mechanism and driving state features reflecting the internal transmission state at a unified time scale, comprehensively evaluating the dynamic consistency and structural stability during the switch operation process.

[0062] 2. Obtain the operation command corresponding to the target high-voltage disconnect switch, and simultaneously start the preset camera and preset angle sensor at the same time as receiving the operation command; acquire video data based on the start of the preset camera, and acquire corner data based on the angle sensor; perform time synchronization calibration on the video data and corner data to build a unified timestamp reference, thereby ensuring strict alignment of different modal data in the time dimension and achieving consistent matching between video signal and corner signal.

[0063] 3. Obtain training sample data for the target high-voltage disconnect switch under various operating modes, including normal operation, foreign object jamming, transmission component disengagement, motor stall, and component loosening. Based on the training sample data, establish a multimodal neural network structure, including a convolutional neural network layer for video sample feature learning, a long short-term memory network layer for learning temporal features of corner samples, and a fully connected layer for outputting multidimensional feature sequences. Optimize the parameters of the convolutional neural network layer, long short-term memory network layer, and fully connected layer through a correction mechanism until training convergence, and construct a multimodal feature fusion diagnostic model to achieve stable identification and classification under various operating modes, thereby improving the accuracy and generalization ability of fault diagnosis.

[0064] like Figure 3 The diagram shows a schematic of a status monitoring device for a disconnector switch operating mechanism according to an embodiment of this application. The device includes an acquisition module 21 and a processing module 22. The acquisition module 21 is used to acquire multimodal data of the target high-voltage disconnector at a unified timestamp reference through a synchronous triggering mechanism. The multimodal data includes video data and corner data. The module performs preprocessing operations on the multimodal data and extracts visual motion features from the video data and driving state features from the corner data.

[0065] The processing module 22 is used to input visual motion features and driving state features into the multimodal feature fusion diagnostic model and output the multidimensional feature sequence after feature fusion; based on the multidimensional feature sequence, the fault diagnosis result corresponding to the target high-voltage disconnecting switch is output through the multimodal feature fusion diagnostic model.

[0066] In one possible implementation, the acquisition module 21 is used to acquire multimodal data of the target high-voltage disconnect switch at a unified timestamp reference through a synchronous triggering mechanism. Specifically, it includes: acquiring the operation command corresponding to the target high-voltage disconnect switch, and simultaneously activating a preset camera and a preset angle sensor at the same time as receiving the operation command; acquiring video data based on activating the preset camera, and acquiring corner data based on the angle sensor; and performing time synchronization calibration on the video data and corner data to construct a unified timestamp reference.

[0067] In one possible implementation, the acquisition module 21 is used to extract visual motion features from the video data, specifically including: dividing the video data into multiple video frames, and locating the target components of the target high-voltage disconnect switch in the multiple video frames based on computer vision algorithms; the target components include the main disconnect switch, connecting rod, and target drive shaft; performing frame-by-frame motion tracking of the target components using optical flow method, and quantifying and calculating the visual motion features corresponding to the target components, including the opening and closing angular velocity features of the main disconnect switch, displacement curve features, and motion stability features.

[0068] In one possible implementation, the acquisition module 21 is used to extract the driving state features from the corner data, specifically including: preprocessing the corner data through filtering and noise reduction; calculating the driving state features of the driving core state based on the preprocessed corner data, the driving state features including spindle angular velocity features, spindle acceleration features and positioning features, the positioning features being used to characterize the accuracy of the component's action endpoint reaching the preset target position.

[0069] In one possible implementation, the processing module 22 is used to construct a multimodal feature fusion diagnostic model before inputting visual motion features and driving state features into the multimodal feature fusion diagnostic model and outputting the multidimensional feature sequence after feature fusion. Specifically, this includes: acquiring training sample data corresponding to the target high-voltage disconnect switch under multiple operating modes, including normal operation mode, foreign object jamming mode, transmission component disengagement mode, motor stall mode, and component loosening mode; establishing a multimodal neural network structure based on the training sample data, including a convolutional neural network layer for learning video sample features, a long short-term memory network layer for learning the temporal features of corner samples, and a fully connected layer for outputting the multidimensional feature sequence; optimizing the parameters of the convolutional neural network layer, the long short-term memory network layer, and the fully connected layer through a running correction mechanism until training converges, and constructing the multimodal feature fusion diagnostic model.

[0070] In one possible implementation, the processing module 22 is used to output the fault diagnosis result corresponding to the target high-voltage disconnector based on the multi-dimensional feature sequence and through the multi-modal feature fusion diagnostic model. Specifically, it includes: calculating the confidence value corresponding to each operating mode of the multi-dimensional feature sequence based on the multi-modal feature fusion diagnostic model; determining whether there is any value in the confidence value that is greater than or equal to a preset threshold; if it is confirmed that there is any value in the confidence value that is greater than or equal to the preset threshold, and the operating mode corresponding to the confidence value is not a normal operating mode, then outputting the operating mode corresponding to the confidence value as the fault type corresponding to the target high-voltage disconnector.

[0071] In one possible implementation, after the processing module 22 outputs the fault diagnosis result corresponding to the target high-voltage disconnecting switch based on the multi-dimensional feature sequence and the multi-modal feature fusion diagnostic model, the method further includes: generating a fault report of the disconnecting switch operating mechanism based on the fault diagnosis result, the fault report of the disconnecting switch operating mechanism including equipment information, fault type, feature data screenshots and processing suggestions; and sending the fault report of the disconnecting switch operating mechanism to the target maintenance personnel in a preset manner, the preset manner including PDF export, email push and SMS push.

[0072] It should be noted that the above embodiments of the apparatus are only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0073] This application also provides an electronic device. (See reference...) Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: at least one processor 301, at least one communication bus 302, a user interface 303, at least one network interface 304, and a memory 305.

[0074] The communication bus 302 is used to enable communication between these components.

[0075] The user interface 303 may include a display screen and a camera. Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0076] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0077] The processor 301 may include one or more processing cores. The processor 301 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 305, and by calling data stored in memory 305. Optionally, the processor 301 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 301 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 301 and may be implemented as a separate chip.

[0078] The memory 305 may include random access memory (RAM) or read-only memory. Optionally, the memory 305 may include a non-transitory computer-readable storage medium. The memory 305 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 305 may also be at least one storage device located remotely from the aforementioned processor 301. (Refer to...) Figure 3 The memory 305, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a status monitoring application for the disconnector switch operating mechanism.

[0079] exist Figure 3In the illustrated electronic device, the user interface 303 is primarily used to provide an input interface for the user and acquire user input data; while the processor 301 can be used to call the status monitoring application program for the disconnector switch operating mechanism stored in the memory 305. When executed by one or more processors 301, the electronic device performs one or more of the methods described in the above embodiments. It should be noted that, for the foregoing method embodiments, for the sake of simplicity, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0080] This application also provides a computer-readable storage medium storing instructions. When executed by one or more processors, these instructions cause an electronic device to perform one or more of the methods described in the above embodiments.

[0081] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0082] In the various embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between apparatuses or units may be electrical or other forms.

[0083] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0084] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0085] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0086] The above description is merely an exemplary embodiment disclosed in this application and should not be construed as limiting the scope of this application. Any equivalent changes and modifications made in accordance with the teachings of this application shall still fall within the scope of this application.

[0087] This application is intended to cover any variations, uses, or adaptations disclosed herein that follow the general principles disclosed herein and include common knowledge or customary technical means in the art that are not described in this application.

Claims

1. A method for monitoring the condition of a high-voltage disconnector operating mechanism, characterized in that, Includes the following steps: Acquire multimodal data of the high-voltage disconnector, wherein the multimodal data includes at least video data and rotation angle data; Feature extraction is performed on the multimodal data to obtain multimodal features; the multimodal features include at least video features and corner-turning temporal features; The multimodal features are input into the multimodal feature fusion diagnostic model, and the working status of the high-voltage disconnector operating mechanism is determined based on the multidimensional feature sequence output by the multimodal feature fusion diagnostic model.

2. The method for monitoring the status of a high-voltage disconnector operating mechanism according to claim 1, characterized in that, The multimodal data of the high-voltage disconnect switch is acquired through a preset sensor; wherein, the video data is acquired through a preset camera, and the rotation angle data is acquired through a preset angle sensor.

3. The method for monitoring the status of a high-voltage disconnector operating mechanism according to claim 2, characterized in that, The multi-mode data of the high-voltage disconnector is acquired through a synchronous triggering mechanism, specifically including the following steps: Upon receiving the operation command from the high-voltage disconnect switch, the preset camera and preset angle sensor are simultaneously triggered; The video data is acquired based on the preset camera, and the corner data is acquired based on the angle sensor; The video data and the cornering data are time-synchronized and calibrated so that the multimodal data have a unified timestamp reference.

4. The method for monitoring the status of a high-voltage disconnector operating mechanism according to claim 1, characterized in that, The feature extraction of the multimodal data includes video feature extraction; the video feature extraction specifically includes the following steps: The video data is divided into multiple video frames; The target components in the high-voltage disconnect switch are located in multiple video frames based on computer vision algorithms; the target components include the main disconnect switch, the connecting rod, and the target drive shaft. The target component is tracked frame by frame using optical flow, and the visual motion features corresponding to the target component are quantified and calculated. The visual motion features include at least the opening and closing angular velocity features of the main switch, the displacement curve features, and the motion stability features.

5. The method for monitoring the status of a high-voltage disconnector operating mechanism according to claim 1, characterized in that, The feature extraction of the multimodal data includes corner time series feature extraction; the corner time series feature extraction specifically includes the following steps: The corner data is preprocessed by filtering and noise reduction. The rotation timing features of the driving core state are calculated based on the preprocessed rotation data; the rotation timing features include spindle angular velocity features, spindle acceleration features, and positioning features; the positioning features are used to characterize the accuracy of the component's action endpoint reaching the preset target position.

6. The method for monitoring the status of a high-voltage disconnector operating mechanism according to claim 1, characterized in that, The multimodal feature fusion diagnostic model is trained through the following steps: Acquire training sample data corresponding to the high-voltage disconnect switch under multiple operating modes; the multiple operating modes include normal operation mode, foreign object jamming mode, transmission component disengagement mode, motor stall mode, and component loosening mode; A multimodal neural network structure is established based on the training sample data; the multimodal neural network structure specifically includes a convolutional neural network layer for learning video sample features, a long short-term memory network layer for learning temporal features of corner samples, and a fully connected layer for outputting the multidimensional feature sequence; The training sample data is input into the multimodal neural network structure; the parameters of the multimodal neural network structure are adjusted according to its output. Return to the step of inputting the training sample data into the multimodal neural network structure, and re-input the training sample data into the multimodal neural network structure until the output of the multimodal neural network structure converges to a preset range; After the output of the multimodal neural network structure converges to a preset range, the multimodal feature fusion diagnostic model is constructed based on the multimodal neural network structure.

7. The method for monitoring the status of a high-voltage disconnector operating mechanism according to claim 1, characterized in that, The step of determining the operating status of the high-voltage disconnector operating mechanism based on the multi-dimensional feature sequence output by the multi-modal feature fusion diagnostic model specifically includes the following steps: Calculate the confidence value of the multidimensional feature sequence for each of the operating modes; Determine whether any of the confidence values ​​is greater than or equal to a preset threshold; If it is confirmed that any value in the confidence level is greater than or equal to the preset threshold, and the operating mode corresponding to the confidence level is not a normal operating mode, then the operating mode corresponding to the confidence level is output as the fault type corresponding to the target high-voltage disconnect switch.

8. The method for monitoring the status of a high-voltage disconnector operating mechanism according to claim 1, characterized in that, After the step of determining the working state of the high-voltage disconnector operating mechanism based on the multi-dimensional feature sequence output by the multi-modal feature fusion diagnostic model, the following steps are also included: A fault report for the high-voltage disconnector operating mechanism is generated based on the fault diagnosis results. The fault report includes equipment information of the high-voltage disconnector, fault type, characteristic data screenshots, and handling suggestions. The fault report of the disconnector switch operating mechanism is sent to the corresponding maintenance personnel of the high-voltage disconnector switch according to a preset method; the preset method includes at least PDF export, email push and SMS push.

9. An electronic device, characterized in that, The device includes a processor, a communication bus, a user interface, a network interface, and a memory. The memory is used to store instructions. The user interface and the network interface are used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1 to 7.

Images