High voltage circuit breaker contact wear intelligent online detection and fault prediction system

Abstract

The application relates to the technical field of high-voltage circuit breaker monitoring, and discloses a high-voltage circuit breaker contact wear intelligent online detection and fault prediction system, which comprises a contact wear online detection module, a data acquisition and processing module, a wear degree intelligent analysis module, a fault risk prediction module, an alarm and output module, a data storage and management module and a user interaction module; when the high-voltage circuit breaker contact wear state is online detected, the electrical, mechanical and thermal parameters of the contact are synchronously collected through a multi-sensor fusion technology, and a wavelet transform and Kalman filter multi-stage signal processing algorithm is adopted to perform adaptive filtering and optimization on the original detection signal, so that the influence of arc, electromagnetic noise transient strong interference on the detection signal can be inhibited, the accuracy of contact contact resistance and stroke key parameter calculation can be ensured, the accuracy of contact wear state evaluation can be improved, and the evaluation error caused by signal interference can be reduced.

Description

Technical Field

[0001] This invention relates to the field of high-voltage circuit breaker monitoring technology, specifically to a high-voltage circuit breaker contact wear intelligent online detection and fault prediction system. Background Technology

[0002] High-voltage circuit breakers are the most important control and protection devices to ensure the safe and stable operation of the power grid. They are numerous and extremely complex in the power grid. High-voltage circuit breakers undertake the dual tasks of control and protection in the power system. With the trend of global power grid interconnection, the scale of the power grid is growing larger and larger. Therefore, the position of high-voltage circuit breakers in the power system is becoming more and more important.

[0003] Currently, in the online detection and fault prediction of contact wear in high-voltage circuit breakers, due to the various complex operating conditions such as strong electric arcs, electromagnetic interference, and mechanical vibrations during circuit breaker operation, the existing detection systems cannot detect and compensate for the instantaneous strong interference on the detection signals in real time when monitoring the contact wear status. When the monitoring signals are contaminated by electric arcs and electromagnetic noise during the acquisition process, it will cause large calculation errors in key parameters such as contact resistance and stroke, making it impossible to guarantee the accuracy of wear status assessment.

[0004] Therefore, a smart online detection and fault prediction system for high-voltage circuit breaker contact wear is proposed to solve the above problems. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent online detection and fault prediction system for high-voltage circuit breaker contact wear, which solves the problem mentioned in the background technology that the calculation errors of key parameters such as contact resistance and stroke are large and cannot guarantee the accuracy of wear condition assessment.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent online detection and fault prediction system for high-voltage circuit breaker contact wear, the system comprising:

[0007] The online contact wear detection module monitors the electrical, mechanical, and thermal parameters of the high-voltage circuit breaker contacts in real time through multi-sensor fusion technology. The electrical parameters include contact resistance, breaking current waveform, and arc energy; the mechanical parameters include contact stroke, operating force, and vibration signal; and the thermal parameters include contact surface temperature and temperature rise rate. It also outputs multi-source sensor signals.

[0008] The data acquisition and processing module receives the signals from the multi-source sensors, performs signal filtering, noise reduction, and feature extraction through wavelet transform and Kalman filtering algorithms, and generates standardized contact wear detection data.

[0009] The intelligent wear analysis module calculates the wear amount, wear rate and remaining life index of the contact based on the standardized contact wear detection data and combined with the deep neural network model in the machine learning algorithm, and outputs the wear analysis results.

[0010] The fault risk prediction module, based on the wear analysis results, combined with historical wear data and environmental factors, predicts the contact fault risk level and fault time point, and generates a risk prediction report.

[0011] The alarm and output module triggers an audible and visual alarm, generates a maintenance suggestion report, and outputs it to the monitoring center through the communication interface when the fault risk level in the risk prediction report exceeds the preset threshold.

[0012] The data storage and management module stores historical detection data, model parameters and user configuration information of the wear degree intelligent analysis module and the fault risk prediction module, supports efficient querying and fault recovery of large data volumes, and provides data access and management services.

[0013] The user interaction module provides a graphical interface that displays contact status, wear trends, and alarm information in real time. It supports user-defined thresholds and manual operation, and is used to receive manual input commands and provide feedback on system status information.

[0014] Preferably, the online contact wear detection module includes an electrical parameter detection unit, a mechanical parameter detection unit, and a thermal parameter detection unit;

[0015] The electrical parameter detection unit collects the contact resistance and arc parameters of the high-voltage circuit breaker during operation using a high-precision current transformer and voltage sensor, and generates electrical waveform data.

[0016] The mechanical parameter detection unit measures the contact stroke and vibration signal of the operating mechanism through a laser displacement sensor and an accelerometer, and generates mechanical dynamic data.

[0017] The thermal parameter detection unit monitors the temperature distribution and thermal field changes on the contact surface using an infrared thermal imager and thermocouples, and generates thermal imaging data.

[0018] Preferably, the data acquisition and processing module includes a multi-source data acquisition unit, a signal preprocessing unit, and a feature extraction unit;

[0019] The multi-source data acquisition unit is used to simultaneously acquire the electrical waveform data, mechanical dynamic data and thermal imaging data, and generate a digital signal stream through an analog-to-digital converter;

[0020] The signal preprocessing unit performs noise reduction and smoothing on the digital signal stream based on wavelet transform and Kalman filtering algorithms, eliminating environmental interference and sensor errors, and generating purified sensor data.

[0021] The feature extraction unit extracts key features of contact wear from the purification sensor data, including contact resistance change rate, stroke deviation, vibration spectrum features and temperature gradient, and generates standardized contact wear detection data.

[0022] Preferably, the intelligent wear analysis module includes a wear calculation unit, a life assessment unit, and an adaptive learning unit;

[0023] The wear calculation unit, based on standardized contact wear detection data, uses a deep neural network model to calculate the cumulative wear amount and real-time wear rate of the contact through forward propagation, and generates a wear degree index.

[0024] The life assessment unit, combining historical operating data of the high-voltage circuit breaker and material fatigue model, assesses the remaining service life and health status of the contacts and generates a life prediction report.

[0025] The adaptive learning unit dynamically updates the neural network parameters through an incremental learning mechanism, enabling the model to adapt to different circuit breaker models and operating conditions.

[0026] Preferably, the fault risk prediction module includes a risk level assessment unit, a fault time prediction unit, and an environmental factor integration unit;

[0027] The risk level assessment unit, based on the wear analysis results and the fault history database, uses a support vector machine algorithm to classify the contact fault risk level, including low risk, medium risk and high risk.

[0028] The fault time prediction unit uses a time series analysis model to predict the possible time points when contact faults may occur and generates fault time axis data.

[0029] The environmental factor integration unit is used to integrate environmental parameters such as humidity, dust concentration, and load changes to correct the parameters of the time series analysis model.

[0030] Preferably, the alarm and output module includes an alarm triggering unit, a report generation unit, and a communication interface unit;

[0031] When the fault risk level is high, the alarm triggering unit automatically triggers a local audible and visual alarm and a remote SMS notification.

[0032] The report generation unit generates a structured maintenance recommendation report based on the wear analysis results and the risk prediction report, including recommended replacement time and maintenance steps;

[0033] The communication interface unit transmits alarm information and reports to the power grid monitoring system and mobile terminal via Ethernet and wireless communication protocols.

[0034] Preferably, the data storage and management module adopts a distributed architecture, including a real-time database unit and a historical data archiving unit;

[0035] The real-time database unit is used to store sensor data and wear analysis results within the current detection cycle, and supports high-speed read / write and real-time query.

[0036] The historical data archiving unit is used to compress and store long-term historical data, and generate wear trend analysis reports based on data mining technology, providing training data for the prediction models in the wear degree intelligent analysis module and the fault risk prediction module.

[0037] The data storage and management module also integrates data encryption and access control mechanisms to protect the security and integrity of stored data.

[0038] Preferably, the user interaction module includes a visual interface unit and an operation control unit;

[0039] The visualization interface unit displays dynamic curves of contact wear, risk heat maps, and alarm logs through a web interface and mobile application, supporting multi-dimensional data visualization.

[0040] The operation control unit allows users to manually start the detection process, adjust alarm thresholds, and export detection reports, and ensures operational security through authentication.

[0041] The user interaction module works in conjunction with the alarm and output module to receive alarm information and support manual confirmation and processing.

[0042] Preferably, the wavelet transform and Kalman filter algorithms used in the signal preprocessing unit are implemented through a multi-level filtering mechanism, specifically including the following processing steps:

[0043] Primary filtering stage: A moving average filter is used to perform preliminary smoothing of the original sensor signal;

[0044] Intermediate filtering stage: A threshold denoising algorithm based on wavelet transform is used to adaptively filter signal components in different frequency bands;

[0045] Advanced filtering stage: Predictive correction of dynamic signals using Kalman filters;

[0046] Signal verification stage: The correlation coefficient analysis method is used to verify the consistency between the filtered signal and the original signal.

[0047] Preferably, the adaptive learning unit adopts an incremental learning mechanism, specifically implemented as follows:

[0048] Model initialization phase: A deep neural network model is pre-trained based on historical circuit breaker operation data to establish an initial wear analysis model;

[0049] Online update phase: Network weights are dynamically adjusted using standardized contact wear detection data collected in real time, and parameters are optimized using the gradient descent algorithm;

[0050] Forgetting mechanism design: Introduce the elastic weight consolidation algorithm to constrain the weight updates related to historical tasks in the neural network;

[0051] Performance evaluation phase: Periodically calculate the prediction accuracy of the deep neural network model and the error rate of real-time data. When the error exceeds the threshold, trigger the model reconstruction process.

[0052] Compared with existing technologies, this invention provides an intelligent online detection and fault prediction system for high-voltage circuit breaker contact wear, which has the following advantages:

[0053] 1. In this invention, when performing online detection of the wear condition of high-voltage circuit breaker contacts, the electrical, mechanical, and thermal parameters of the contacts are simultaneously collected through multi-sensor fusion technology. The original detection signal is adaptively filtered and optimized using wavelet transform and Kalman filtering multi-level signal processing algorithms. This can suppress the influence of instantaneous strong interference from electric arcs and electromagnetic noise on the detection signal, ensure the accuracy of the calculation of key parameters such as contact resistance and stroke, improve the accuracy of contact wear condition assessment, and reduce assessment errors caused by signal interference.

[0054] 2. In this invention, when analyzing the mechanical characteristics of high-voltage circuit breaker contacts, the intelligent wear analysis module performs real-time analysis of the inherent logical relationship of multi-source sensor data. This allows for indirect identification of whether the monitoring status of vibration and stroke sensors is abnormal. When abnormal or deviating data is detected, the alarm and output module triggers a warning signal in real time, prompting sensor calibration and data compensation. This reduces the possibility of inaccurate contact motion characteristic analysis benchmarks caused by sensor position offsets, and provides real-time warnings when sensor monitoring status is abnormal, ensuring the reliability of the contact wear analysis benchmark.

[0055] 3. In this invention, when conducting a comprehensive assessment and prediction of the health status of high-voltage circuit breaker contacts, multi-source sensor data is fused and input into a deep neural network model to achieve comprehensive learning and hierarchical identification of multiple wear modes, including electrical wear, mechanical wear, and thermal wear. Based on an incremental learning mechanism, the model can adapt to different circuit breaker operating conditions, enabling the system to achieve intelligent fusion diagnosis and prediction of multiple wear modes of contacts, reducing misjudgments of complex wear mechanisms, and improving the accuracy of contact remaining life prediction and the reliability of fault early warning. Attached Figure Description

[0056] Figure 1 This is a schematic diagram of the intelligent online detection and fault prediction system for high-voltage circuit breaker contact wear according to the present invention. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] For specific implementation examples, please refer to: Figure 1 A high-voltage circuit breaker contact wear intelligent online detection and fault prediction system, the system includes:

[0059] The online contact wear detection module monitors the electrical, mechanical, and thermal parameters of the high-voltage circuit breaker contacts in real time through multi-sensor fusion technology. The electrical parameters include contact resistance, breaking current waveform, and arc energy; the mechanical parameters include contact stroke, operating force, and vibration signal; and the thermal parameters include contact surface temperature and temperature rise rate. It also outputs multi-source sensor signals.

[0060] The data acquisition and processing module receives signals from multiple sources of sensors, performs signal filtering, noise reduction and feature extraction through wavelet transform and Kalman filtering algorithms, and generates standardized contact wear detection data.

[0061] The intelligent wear analysis module calculates the wear amount, wear rate and remaining life index of the contact based on standardized contact wear detection data and combined with the deep neural network model in the machine learning algorithm, and outputs the wear analysis results.

[0062] The fault risk prediction module, based on wear analysis results, combined with historical wear data and environmental factors, predicts the contact fault risk level and fault time point, and generates a risk prediction report.

[0063] The alarm and output module triggers an audible and visual alarm, generates a maintenance suggestion report, and outputs it to the monitoring center through the communication interface when the fault risk level in the risk prediction report exceeds the preset threshold.

[0064] The data storage and management module stores historical detection data, model parameters and user configuration information of the wear degree intelligent analysis module and the fault risk prediction module, supports efficient querying and fault recovery of large data volumes, and provides data access and management services.

[0065] The user interaction module provides a graphical interface that displays contact status, wear trends, and alarm information in real time. It supports user-defined thresholds and manual operation, and is used to receive manual input commands and provide feedback on system status information.

[0066] The online contact wear detection module includes an electrical parameter detection unit, a mechanical parameter detection unit, and a thermal parameter detection unit;

[0067] The electrical parameter detection unit collects contact resistance and arc parameters of the high-voltage circuit breaker during operation using high-precision current transformers and voltage sensors, and generates electrical waveform data.

[0068] The mechanical parameter detection unit measures the contact stroke and vibration signal of the operating mechanism through a laser displacement sensor and an accelerometer, and generates mechanical dynamic data.

[0069] The thermal parameter detection unit monitors the temperature distribution and thermal field changes on the contact surface using an infrared thermal imager and thermocouples, and generates thermal imaging data.

[0070] The data acquisition and processing module includes a multi-source data acquisition unit, a signal preprocessing unit, and a feature extraction unit;

[0071] A multi-source data acquisition unit is used to simultaneously acquire electrical waveform data, mechanical dynamic data, and thermal imaging data, and generate a digital signal stream through an analog-to-digital converter;

[0072] The signal preprocessing unit performs noise reduction and smoothing on the digital signal stream based on wavelet transform and Kalman filtering algorithms, eliminating environmental interference and sensor errors, and generating purified sensor data.

[0073] The feature extraction unit extracts key features of contact wear from the purification sensor data, including contact resistance change rate, stroke deviation, vibration spectrum characteristics and temperature gradient, and generates standardized contact wear detection data.

[0074] The intelligent wear analysis module includes a wear calculation unit, a life assessment unit, and an adaptive learning unit.

[0075] The wear calculation unit, based on standardized contact wear detection data, uses a deep neural network model to calculate the cumulative wear and real-time wear rate of the contact through forward propagation, generating a wear degree index. Its forward propagation calculation process is implemented in the following steps:

[0076] Input layer processing: Standardized contact wear detection data are used as feature vectors and input into the network's input layer;

[0077] Hidden layer computation: Data passes sequentially through two fully connected hidden layers. Each hidden layer undergoes a nonlinear transformation by linearly weighted summation and the application of an activation function.

[0078] Output layer calculation: The output of the last hidden layer is transformed by the output layer to generate continuous numerical predictions of contact wear amount and wear rate index;

[0079] The life assessment unit combines historical operating data of high-voltage circuit breakers with material fatigue models to assess the remaining service life and health status of the contacts and generate a life prediction report.

[0080] The adaptive learning unit dynamically updates the neural network parameters through an incremental learning mechanism, enabling the model to adapt to different circuit breaker models and operating conditions.

[0081] The failure risk prediction module includes a risk level assessment unit, a failure time prediction unit, and an environmental factor integration unit.

[0082] The risk level assessment unit, based on wear analysis results and a fault history database, uses a support vector machine algorithm to classify contact fault risk levels into low, medium, and high risks. This includes the following steps:

[0083] Feature space mapping: Mapping the input wear level index to a high-dimensional feature space through a kernel function;

[0084] Optimal hyperplane solution: Find a hyperplane with the largest margin in the high-dimensional feature space to separate data points with different risk levels;

[0085] Classification decision: Based on the position of the new wear data relative to the hyperplane in the feature space, determine its risk level;

[0086] The fault time prediction unit uses a time series analysis model to predict the possible time points of contact faults and generates fault time axis data. The time series analysis model used to predict the contact fault time points includes the following steps:

[0087] Sequence stabilization: Differentiate the historical wear data sequence to eliminate trends and obtain a stationary time series;

[0088] Model parameter identification: Based on the autocorrelation function and partial autocorrelation function, identify the autoregression order, difference order and moving average order of the ARIMA model;

[0089] Model fitting and prediction: The maximum likelihood estimation method is used to fit the model parameters and extrapolate to predict the wear amount at future time points. When the predicted value exceeds the failure threshold, the time point is recorded as the failure time point.

[0090] The environmental factors integration unit is used to integrate environmental parameters such as humidity, dust concentration, and load changes to correct parameters in the time series analysis model.

[0091] The alarm and output module includes an alarm triggering unit, a report generation unit, and a communication interface unit;

[0092] The alarm triggering unit automatically triggers a local audible and visual alarm and a remote SMS notification when the fault risk level is high.

[0093] The report generation unit generates a structured maintenance recommendation report based on wear analysis results and risk prediction reports, including recommended replacement times and maintenance steps;

[0094] The communication interface unit transmits alarm information and reports to the power grid monitoring system and mobile terminals via Ethernet and wireless communication protocols.

[0095] The data storage and management module adopts a distributed architecture, including a real-time database unit and a historical data archiving unit;

[0096] The real-time database unit is used to store sensor data and wear analysis results within the current detection cycle, and supports high-speed read / write and real-time query.

[0097] The historical data archiving unit is used to compress and store long-term historical data, and generate wear trend analysis reports based on data mining technology, providing training data for the prediction models in the wear degree intelligent analysis module and the failure risk prediction module;

[0098] The data storage and management module also integrates data encryption and access control mechanisms to protect the security and integrity of stored data.

[0099] The user interaction module includes a visual interface unit and an operation control unit;

[0100] The visualization interface unit displays dynamic curves of contact wear, risk heat maps, and alarm logs through a web interface and mobile application, supporting multi-dimensional data visualization.

[0101] The operation control unit allows users to manually start the detection process, adjust alarm thresholds, and export detection reports, while ensuring operational security through authentication.

[0102] The user interaction module works in conjunction with the alarm and output module to receive alarm information and support manual confirmation and processing.

[0103] The wavelet transform and Kalman filter algorithms used in the signal preprocessing unit are implemented through a multi-stage filtering mechanism, specifically including the following processing steps:

[0104] Primary filtering stage: A moving average filter is used to perform preliminary smoothing of the original sensor signal;

[0105] Intermediate filtering stage: A threshold denoising algorithm based on wavelet transform adaptively filters signal components in different frequency bands, specifically including the following steps:

[0106] Wavelet decomposition: Select appropriate wavelet basis functions and decomposition levels, perform discrete wavelet transform on the input digital signal stream, and obtain the high-frequency coefficients and low-frequency coefficients of each level;

[0107] Threshold processing: for high-frequency coefficients Quantization is performed using an adaptive threshold function, which dynamically adjusts the threshold value based on the signal-to-noise level. Calculate using the following formula:

[0108] ;

[0109] in Indicates the first Threshold for layer wavelet decomposition, The noise standard deviation is estimated based on the high-frequency detail coefficients of the signal. For the first The number of layer wavelet coefficients, This is the layer index for wavelet decomposition;

[0110] Wavelet reconstruction: Using the high-frequency coefficients after thresholding and the original low-frequency coefficients, perform discrete wavelet inverse transform to reconstruct the denoised signal;

[0111] State prediction: Based on the system state equation of the high-voltage circuit breaker contact movement and the state estimate of the previous moment, predict the system state at the current moment;

[0112] Advanced filtering stage: Predictive correction of dynamic signals is performed using a Kalman filter, specifically including the following steps:

[0113] State prediction: Based on the system state equations governing the movement of the high-voltage circuit breaker contacts, and using the state estimate from the previous moment, predict the system state at the current moment. The calculation formula is as follows:

[0114] ;

[0115] in Indicates the first The prior state estimate vector at time t. Indicates the first The state transition matrix at time t, Indicates the first The posterior state estimate vector at time t. For time indexing;

[0116] Covariance prediction: updating the covariance matrix of the predicted state values The calculation formula is as follows:

[0117] ;

[0118] in This represents the prior estimation error covariance matrix. Indicates the first The posterior estimation error covariance matrix at time 1. Represents the process noise covariance matrix;

[0119] Kalman gain calculation: Calculate the Kalman gain at the current time by combining the system observation equations and the prediction covariance. The calculation formula is as follows:

[0120] ;

[0121] in Represents the observation matrix. Represents the observation noise covariance matrix;

[0122] State Update: Using Kalman gain, the system state prediction is compared with the actual sensor observations at the current moment. By performing weighted fusion, the optimal state estimate is obtained. The calculation formula is as follows:

[0123] ;

[0124] in Indicates the first The posterior state estimate vector at time t. Indicates the first The actual sensor observation vector at any given time;

[0125] Covariance Update: Updates the covariance matrix of the state estimates. To prepare for predictions at the next moment, the calculation formula is as follows:

[0126] ;

[0127] in Represents the identity matrix. Indicates the first The posterior estimation error covariance matrix at time t;

[0128] Signal verification stage: The correlation coefficient analysis method is used to verify the consistency between the filtered signal and the original signal.

[0129] The adaptive learning unit employs an incremental learning mechanism, which is implemented in the following ways:

[0130] Model initialization phase: A deep neural network model is pre-trained based on historical circuit breaker operation data to establish an initial wear analysis model;

[0131] Online update phase: The network weights are dynamically adjusted using standardized contact wear detection data collected in real time, and the parameters are optimized using the gradient descent algorithm, including the following steps:

[0132] Small-batch data sampling: Sequentially extract small batches of samples from real-time collected contact wear data;

[0133] Loss function calculation: Input a small batch of samples into the current network model and calculate the loss function between the model's predicted values ​​and the actual values;

[0134] Backpropagation gradient calculation: Using the backpropagation algorithm, calculate the gradient of the loss function with respect to the weights of each layer of the network;

[0135] Weight Iterative Update: Based on the gradient, the network weight parameters are updated along the reverse direction of the gradient using the stochastic gradient descent algorithm. This weight iterative update process is implemented through the following specific steps:

[0136] Learning rate setting: Set the initial learning rate η to control the step size of weight updates;

[0137] Gradient calculation: Calculate the loss function for the current mini-batch of data. Regarding model weights gradient ;

[0138] Weight update execution: The weight parameters in the network are iteratively updated according to the following update rules:

[0139] ;

[0140] in The value of the loss function. and They represent the first The weight in the th... The values ​​before and after the next iteration update For weight parameter index, Index for iteration count, The learning rate;

[0141] Forgetting Mechanism Design: An elastic weight consolidation algorithm is introduced to constrain the updates of weights related to historical tasks in the neural network. This includes the following steps:

[0142] Importance Measurement: Before learning a new task, the importance of each weight parameter in the old task is calculated based on the Fisher information matrix. The weight parameters are... Importance metric Calculate using the following formula:

[0143] ;

[0144] in The first term of the neural network One weight parameter, This represents the importance metric value on the diagonal of the corresponding Fisher information matrix. This represents the old task dataset. Indicates that under given parameters Log-likelihood function of old task data Represents the mathematical expectation;

[0145] Regularization constraint construction: Importance measure As a regularization coefficient, an elastic weight consolidation regularization term based on weighted squared difference is constructed and added to the new task loss function. Specifically, it is used to constrain the weight parameters that are important for historical tasks from their learned values ​​when the model learns new tasks.

[0146] Constraint optimization: When optimizing the objective function of a new task, the changes in the weights are constrained by regularization terms, thereby reinforcing the memory of the old task;

[0147] Performance evaluation phase: Periodically calculate the prediction accuracy of the deep neural network model against the error rate of real-time data. When the error exceeds a threshold, trigger the model reconstruction process, which includes the following steps:

[0148] Prediction generation: Input the currently collected contact wear data into the updated deep neural network model to obtain the predicted wear amount;

[0149] Error Calculation: The error index between the predicted value and the actual wear value obtained through measurement and calibration within the corresponding time period is calculated. Specifically, a combination of mean absolute error and root mean square error is used for quantitative evaluation. The calculation process is as follows:

[0150] Mean absolute error calculation:

[0151] ;

[0152] in The mean absolute error, Indicates the first The actual wear value obtained from measurement calibration This represents the corresponding model prediction value. To count the total number of data points within a given time period, Index the data points;

[0153] Root mean square error calculation:

[0154] ;

[0155] in This is the root mean square error;

[0156] Threshold comparison: The calculated error rate is compared with a preset accuracy threshold. If the error rate exceeds the threshold for several consecutive periods, the model reconstruction process is triggered.

[0157] The operation steps of the intelligent online detection and fault prediction system for high-voltage circuit breaker contact wear are as follows:

[0158] Step 1: Synchronous acquisition of multi-source sensor data

[0159] The system synchronously collects the status parameters of the high-voltage circuit breaker contacts using multiple sensors integrated on the track inspection vehicle. The electrical parameter detection unit employs high-precision current transformers and voltage sensors to monitor contact resistance, breaking current waveform, and arc energy in real time. The mechanical parameter detection unit measures contact stroke, operating force, and mechanical vibration signals using laser displacement sensors and accelerometers. The thermal parameter detection unit captures the surface temperature distribution and temperature rise rate of the contacts using infrared thermal imagers and thermocouples. All sensor data is synchronously aggregated by the data acquisition unit to form a multi-source sensor signal describing the overall condition of the contacts.

[0160] Step 2: Intelligent Signal Preprocessing and Feature Extraction

[0161] The acquired multi-source sensor signals enter the data acquisition and processing module for in-depth purification and feature extraction. The signal preprocessing unit first uses a moving average filter for preliminary smoothing to suppress high-frequency random noise; then, a wavelet transform-based threshold denoising algorithm is used to adaptively filter signals in specific frequency bands to separate noise from the signal; finally, a Kalman filter is used to predict and correct dynamic signals, reducing measurement uncertainty. The feature extraction unit then accurately extracts key wear features from the purified signals, including contact resistance change rate, stroke dynamic deviation, vibration spectrum features, and temperature gradient field, generating a standardized contact wear detection dataset.

[0162] Step 3: Intelligent Wear Analysis and Life Assessment

[0163] Standardized test data is input into the wear level intelligent analysis module for in-depth calculation. The wear calculation unit uses a pre-trained deep neural network model to calculate and output the cumulative wear and instantaneous wear rate of the contacts in real time through forward propagation. The life assessment unit combines historical operating data and material fatigue models to comprehensively calculate the remaining service life of the contacts and generate a health status report. At the same time, the adaptive learning unit dynamically updates the neural network weight parameters through an incremental learning mechanism to ensure that the model can continuously adapt to different circuit breaker models and changing operating conditions, maintaining analysis accuracy.

[0164] Step 4: Fault Risk Prediction and Early Warning Decision

[0165] Based on wear analysis results, the fault risk prediction module performs a multi-dimensional fault risk assessment. The risk level assessment unit uses a support vector machine algorithm to classify the current risk level against a historical fault database. The fault time prediction unit applies a time series analysis model to extrapolate and predict the timing of potential fault occurrences. The environmental factor integration unit simultaneously incorporates external parameters such as humidity and dust concentration to correct the prediction model in real time. When the assessed risk level exceeds a preset threshold, the system immediately triggers an early warning decision process.

[0166] Step 5: Alarm Information Generation and Interactive Management

[0167] The alarm and output module generates a structured response based on the early warning decision. The alarm triggering unit simultaneously activates local audible and visual alarms and remote SMS notifications. The report generation unit automatically generates a detailed maintenance suggestion report, clearly specifying the recommended replacement time and specific maintenance steps. All alarm information and reports are transmitted to the upper-level monitoring system via the communication interface unit. Meanwhile, the user interaction module provides a graphical interface, displaying real-time contact status trends and alarm logs, and supports authorized users to adjust thresholds and perform manual detection operations, forming a complete human-machine collaborative decision-making closed loop.

[0168] Step Six: Persistent Data Storage and System Support

[0169] Throughout the system's operation, the data storage and management module provides full-cycle data support. The real-time database unit stores dynamic data for the current detection period at high speed, supporting millisecond-level queries. The historical data archiving unit compresses, stores, and analyzes massive amounts of historical data, providing a foundation for training data for the predictive model. The integrated data encryption and access control mechanisms in this module comprehensively ensure data security and integrity, forming the cornerstone of the system's stable operation.

[0170] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0171] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A high-voltage circuit breaker contact wear intelligent online detection and fault prediction system, characterized in that: The system includes: The online contact wear detection module monitors the electrical, mechanical, and thermal parameters of the high-voltage circuit breaker contacts in real time through multi-sensor fusion technology. The electrical parameters include contact resistance, breaking current waveform, and arc energy; the mechanical parameters include contact stroke, operating force, and vibration signal; and the thermal parameters include contact surface temperature and temperature rise rate. It also outputs multi-source sensor signals. The data acquisition and processing module receives the signals from the multi-source sensors, performs signal filtering, noise reduction, and feature extraction through wavelet transform and Kalman filtering algorithms, and generates standardized contact wear detection data. The intelligent wear analysis module calculates the wear amount, wear rate and remaining life index of the contact based on the standardized contact wear detection data and combined with the deep neural network model in the machine learning algorithm, and outputs the wear analysis results. Specifically, it includes a wear amount calculation unit, a life assessment unit and an adaptive learning unit. The wear calculation unit, based on standardized contact wear detection data, uses a deep neural network model to calculate the cumulative wear amount and real-time wear rate of the contact through forward propagation, and generates a wear degree index. The life assessment unit, combining historical operating data of the high-voltage circuit breaker and material fatigue model, assesses the remaining service life and health status of the contacts and generates a life prediction report. The adaptive learning unit dynamically updates the neural network parameters through an incremental learning mechanism, enabling the model to adapt to different circuit breaker models and operating conditions. The fault risk prediction module, based on the wear analysis results, combined with historical wear data and environmental factors, predicts the contact fault risk level and fault time point, and generates a risk prediction report. The alarm and output module triggers an audible and visual alarm, generates a maintenance suggestion report, and outputs it to the monitoring center through the communication interface when the fault risk level in the risk prediction report exceeds the preset threshold. The data storage and management module stores historical detection data, model parameters and user configuration information of the wear degree intelligent analysis module and the fault risk prediction module, supports efficient querying and fault recovery of large data volumes, and provides data access and management services. The user interaction module provides a graphical interface that displays contact status, wear trends, and alarm information in real time. It supports user-defined thresholds and manual operation, and is used to receive manual input commands and provide feedback on system status information.

2. The intelligent online detection and fault prediction system for high-voltage circuit breaker contact wear according to claim 1, characterized in that: The online contact wear detection module includes an electrical parameter detection unit, a mechanical parameter detection unit, and a thermal parameter detection unit; The electrical parameter detection unit collects the contact resistance and arc parameters of the high-voltage circuit breaker during operation using a high-precision current transformer and voltage sensor, and generates electrical waveform data. The mechanical parameter detection unit measures the contact stroke and vibration signal of the operating mechanism through a laser displacement sensor and an accelerometer, and generates mechanical dynamic data. The thermal parameter detection unit monitors the temperature distribution and thermal field changes on the contact surface using an infrared thermal imager and thermocouples, and generates thermal imaging data.

3. The intelligent online detection and fault prediction system for high-voltage circuit breaker contact wear according to claim 2, characterized in that: The data acquisition and processing module includes a multi-source data acquisition unit, a signal preprocessing unit, and a feature extraction unit; The multi-source data acquisition unit is used to simultaneously acquire the electrical waveform data, mechanical dynamic data and thermal imaging data, and generate a digital signal stream through an analog-to-digital converter; The signal preprocessing unit performs noise reduction and smoothing on the digital signal stream based on wavelet transform and Kalman filtering algorithms, eliminating environmental interference and sensor errors, and generating purified sensor data. The feature extraction unit extracts key features of contact wear from the purification sensor data, including contact resistance change rate, stroke deviation, vibration spectrum features and temperature gradient, and generates standardized contact wear detection data.

4. The intelligent online detection and fault prediction system for high-voltage circuit breaker contact wear according to claim 1, characterized in that: The fault risk prediction module includes a risk level assessment unit, a fault time prediction unit, and an environmental factor integration unit. The risk level assessment unit, based on the wear analysis results and the fault history database, uses a support vector machine algorithm to classify the contact fault risk level, including low risk, medium risk and high risk. The fault time prediction unit uses a time series analysis model to predict the possible time points when contact faults may occur and generates fault time axis data. The environmental factor integration unit is used to integrate environmental parameters such as humidity, dust concentration, and load changes to correct the parameters of the time series analysis model.

5. The intelligent online detection and fault prediction system for high-voltage circuit breaker contact wear according to claim 1, characterized in that: The alarm and output module includes an alarm triggering unit, a report generation unit, and a communication interface unit; When the fault risk level is high, the alarm triggering unit automatically triggers a local audible and visual alarm and a remote SMS notification. The report generation unit generates a structured maintenance recommendation report based on the wear analysis results and the risk prediction report, including recommended replacement time and maintenance steps; The communication interface unit transmits alarm information and reports to the power grid monitoring system and mobile terminal via Ethernet and wireless communication protocols.

6. The intelligent online detection and fault prediction system for high-voltage circuit breaker contact wear according to claim 1, characterized in that: The data storage and management module adopts a distributed architecture, including a real-time database unit and a historical data archiving unit; The real-time database unit is used to store sensor data and wear analysis results within the current detection cycle, and supports high-speed read / write and real-time query. The historical data archiving unit is used to compress and store long-term historical data, and generate wear trend analysis reports based on data mining technology, providing training data for the prediction models in the wear degree intelligent analysis module and the fault risk prediction module. The data storage and management module also integrates data encryption and access control mechanisms to protect the security and integrity of stored data.

7. The intelligent online detection and fault prediction system for high-voltage circuit breaker contact wear according to claim 1, characterized in that: The user interaction module includes a visual interface unit and an operation control unit; The visualization interface unit displays dynamic curves of contact wear, risk heat maps, and alarm logs through a web interface and mobile application, supporting multi-dimensional data visualization. The operation control unit allows users to manually start the detection process, adjust alarm thresholds, and export detection reports, and ensures operational security through authentication. The user interaction module works in conjunction with the alarm and output module to receive alarm information and support manual confirmation and processing.

8. The intelligent online detection and fault prediction system for high-voltage circuit breaker contact wear according to claim 3, characterized in that: The wavelet transform and Kalman filtering algorithms used in the signal preprocessing unit are implemented through a multi-level filtering mechanism, specifically including the following processing steps: Primary filtering stage: A moving average filter is used to perform preliminary smoothing of the original sensor signal; Intermediate filtering stage: A threshold denoising algorithm based on wavelet transform is used to adaptively filter signal components in different frequency bands; Advanced filtering stage: Predictive correction of dynamic signals using Kalman filters; Signal verification stage: The correlation coefficient analysis method is used to verify the consistency between the filtered signal and the original signal.

9. The intelligent online detection and fault prediction system for high-voltage circuit breaker contact wear according to claim 1, characterized in that: The adaptive learning unit employs an incremental learning mechanism, specifically implemented as follows: Model initialization phase: A deep neural network model is pre-trained based on historical circuit breaker operation data to establish an initial wear analysis model; Online update phase: Network weights are dynamically adjusted using standardized contact wear detection data collected in real time, and parameters are optimized using the gradient descent algorithm; Forgetting mechanism design: Introduce the elastic weight consolidation algorithm to constrain the weight updates related to historical tasks in the neural network; Performance evaluation phase: Periodically calculate the prediction accuracy of the deep neural network model and the error rate of real-time data. When the error exceeds the threshold, trigger the model reconstruction process.

Images