Fault dynamic identification method and device of repulsion valve circuit breaker under noisy background

Abstract

The application discloses a kind of repulsion valve type circuit breaker fault dynamic identification method and device in noisy background, it is related to high voltage circuit breaker technical field, specific steps are: to multi-source signal is synchronously collected and preprocessed, obtains displacement time signal, gas composition concentration signal in arc extinguishing chamber and opening and closing coil current signal;Displacement time signal, gas composition concentration signal in arc extinguishing chamber and opening and closing coil current signal are carried out feature extraction, standardization, data alignment and feature fusion operation, generate fusion feature;Fusion feature is input into the robust support vector machine model trained and is carried out abnormal identification, and the failure category is output.The application realizes the synchronous collection and analysis of mechanical displacement, gas composition and driving coil current signal by fusing displacement detection module, SF6 gas detection module and repulsion valve coil detection module, and improves the comprehensiveness of state evaluation.

Description

Technical Field

[0001] This invention relates to the field of high-voltage circuit breaker technology, and more specifically to a method and apparatus for dynamic fault identification of repulsion valve circuit breakers under noisy conditions. Background Technology

[0002] Currently, online monitoring technology for high-voltage circuit breakers has evolved into a comprehensive system integrating multi-source sensing and intelligent diagnosis, constructing a health assessment system for equipment through multi-dimensional parameters such as displacement, current, and gas state. Wired invasive detection schemes, represented by magnetic scale displacement sensing and SF6 gas state monitoring, combined with threshold comparison and shallow machine learning algorithms, have initially achieved online diagnosis of the mechanical characteristics and gas sealing status of circuit breakers. However, existing technologies still face prominent problems such as fragmented multi-dimensional monitoring elements, interference of sensing methods with the circuit breaker's performance, and insufficient ability to analyze complex fault modes. Specifically: traditional displacement sensing uses an invasive architecture coupled with a magnetic scale and transmission shaft, leading to increased mechanical load and distortion of motion characteristics; SF6 gas state monitoring only focuses on physical quantities such as temperature, pressure, and density, failing to effectively capture chemical characteristics such as the composition of decomposition products, resulting in missed detection of critical faults such as arc-extinguishing chamber contact erosion; existing diagnostic algorithms rely excessively on data quality and threshold comparison, making it difficult to cope with the fusion of multi-source heterogeneous data and the identification of complex fault modes in noisy environments, especially lacking the ability to predict latent faults. Therefore, how to construct a fault identification system based on a non-intrusive multi-dimensional sensing network and improve the accuracy of fault identification of high-voltage circuit breakers has become an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0003] In view of this, the present invention provides a method and apparatus for dynamic fault identification of repulsion valve circuit breakers under noisy background, which overcomes the above-mentioned defects.

[0004] To achieve the above objectives, the present invention adopts the following technical solution:

[0005] A method for dynamic fault identification of repulsion valve circuit breakers under noisy backgrounds, comprising the following steps:

[0006] Multi-source signals are synchronously acquired and preprocessed to obtain displacement time signals, gas composition concentration signals in the arc extinguishing chamber, and opening and closing coil current signals;

[0007] The displacement-time signal, the gas composition concentration signal in the arc-extinguishing chamber, and the opening and closing coil current signal are subjected to feature extraction, standardization, data alignment, and feature fusion operations to generate fused features.

[0008] The fused features are input into a trained robust support vector machine model for anomaly identification, and the fault category is output.

[0009] Optionally, the step of obtaining the fusion feature is as follows:

[0010] Based on the displacement-time signal, p-dimensional displacement features are extracted, and based on the gas composition concentration signal in the arc extinguishing chamber and the current signal of the opening and closing coil, mp-dimensional gas-current features are jointly extracted to form an m-dimensional original feature vector.

[0011] The displacement characteristics and the gas-current characteristics are standardized to obtain standard displacement characteristics and standard gas-current characteristics;

[0012] The standard displacement features are used as the first set of variables, and the standard gas-current features are used as the second set of variables. The covariance matrix of the two sets of variables is calculated by canonical correlation analysis, and the generalized eigenvalue problem is solved to obtain the typical weight vector. The fused features are generated based on the typical weight vector.

[0013] Optionally, the construction steps of the robust support vector machine model are as follows:

[0014] Acquire multiple sample signals and calculate the mean and variance of the multiple sample signals;

[0015] A set of probability distributions for noisy information is defined based on mean and variance;

[0016] An initial robust chance-constrained support vector machine model is constructed based on the set of probability distributions.

[0017] The parameters of the initial robust chance-constrained support vector machine model are optimized using an adaptive genetic algorithm to generate the robust support vector machine model.

[0018] Optionally, the expression for the initial robust chance-constrained support vector machine model is:

[0019]

[0020] The transformed expression is:

[0021]

[0022] In the formula, w is the weight vector; b is the bias term; ξ is the bias term. k y is a slack variable; C is the penalty coefficient; k Let x be the label of the k-th sample; k Let be the feature vector of the k-th sample; m be the total number of samples; η be the confidence level; u k Let be the mean vector of the k-th sample; The value is calculated from the confidence level.

[0023] Optionally, the fault categories include: jamming during opening and closing, insufficient travel of the mechanism, abnormal opening and closing speed, contact wear, poor contact between moving and stationary contacts, deterioration of the insulation of the arc-extinguishing chamber, contact erosion, aging of the opening and closing coil, open circuit of the opening and closing coil, and short circuit between turns of the opening and closing coil.

[0024] Optionally, it also includes the generation of diagnostic reports, the specific steps of which are as follows:

[0025] A diagnostic report is generated based on the fault category. The diagnostic report includes the fault type, fault location, and maintenance recommendations.

[0026] A dynamic fault identification device for a repulsion valve circuit breaker under noisy background includes:

[0027] The data acquisition and preprocessing system is used to synchronously acquire and preprocess multi-source signals to obtain displacement time signals, gas composition concentration signals in the arc extinguishing chamber, and opening and closing coil current signals.

[0028] The feature extraction and standardization system (6) is used to perform feature extraction, standardization, data alignment and feature fusion operations on the displacement time signal, the gas composition concentration signal in the arc extinguishing chamber and the opening and closing coil current signal to generate fused features;

[0029] The fault identification system is used to input the fused features into a trained robust support vector machine model for anomaly identification and output the fault category.

[0030] Optionally, the data acquisition and preprocessing system includes a displacement detection module (2), an SF6 gas detection module (3), and a repulsion valve coil detection module (4);

[0031] The displacement detection module (2) includes a magnetic induction intensity detection unit (2-2) and a permanent magnet (2-1) adsorbed on the main shaft (1-1) of the transmission mechanism being monitored. It is used to detect the change data of magnetic induction intensity between the permanent magnet (2-1) and the magnetic induction intensity detection unit (2-2) in real time, and perform preprocessing to generate the displacement time signal.

[0032] The SF6 gas detection module (3) includes an electrically connected intake valve and an electrochemical detection unit, which are used to detect the component concentration of various gases in the arc extinguishing chamber and convert the component concentration of various gases into corresponding voltage signals. After preprocessing the voltage signals, the gas component concentration signal in the arc extinguishing chamber is obtained.

[0033] The repulsion valve coil detection module (4) includes an electrically connected Rogowski coil and a signal processing unit (4-1). The Rogowski coil is passed through by the repulsion valve drive coil (5). The module is used to sense the current of the repulsion valve drive coil (5) and convert it into a corresponding voltage signal. The voltage signal is preprocessed to generate the opening and closing coil current signal.

[0034] Optionally, the feature extraction and standardization system (6) includes a feature extraction unit, a standardization unit, and a data alignment and fusion unit;

[0035] The feature extraction unit is used to extract p-dimensional displacement features based on the displacement-time signal, and to jointly extract mp-dimensional gas-current features based on the gas composition concentration signal in the arc extinguishing chamber and the current signal of the opening and closing coil, thus forming an m-dimensional original feature vector.

[0036] A standardization unit is used to standardize the displacement characteristics and the gas-current characteristics to obtain standard displacement characteristics and standard gas-current characteristics.

[0037] The data alignment and fusion unit is used to take the standard displacement features as the first set of variables and the standard gas-current features as the second set of variables, calculate the covariance matrix of the two sets of variables through canonical correlation analysis, and solve the generalized eigenvalue problem to obtain the typical weight vector; and generate fused features based on the typical weight vector.

[0038] Optionally, a diagnostic report generation system is also included for generating a diagnostic report based on the fault category, the diagnostic report including the fault type, location, and maintenance recommendations.

[0039] As can be seen from the above technical solution, the present invention discloses a method and device for dynamic fault identification of repulsion valve circuit breakers under noisy backgrounds, which has the following advantages compared with the prior art:

[0040] 1. By integrating the displacement detection module, SF6 gas detection module, and repulsion valve coil detection module, the synchronous acquisition and analysis of mechanical displacement, gas composition, and drive coil current signals are realized, breaking through the limitations of traditional single-parameter monitoring and significantly improving the comprehensiveness of condition assessment.

[0041] 2. A robust support vector machine model under noisy conditions is constructed, and the model parameters are optimized using an adaptive genetic algorithm. This model can accurately identify complex faults such as contact wear and arc-extinguishing chamber insulation degradation in noisy environments, effectively avoiding misjudgments and missed judgments.

[0042] 3. Employing non-invasive magnetic displacement detection technology (permanent magnet adsorption installation) and fiber optic isolated transmission, the system requires no modification to the circuit breaker's mechanical structure. The SF6 gas detection module achieves in-situ sampling within the arc-extinguishing chamber via an intake valve, reducing downtime compared to traditional offline detection methods. The system is easy to install and has low maintenance costs. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0044] Figure 1 This is a schematic diagram of the system structure provided by the present invention;

[0045] Figure 2 This is a schematic diagram of the displacement detection module provided by the present invention;

[0046] Figure 3 This is a schematic diagram of the SF6 gas detection module of the present invention;

[0047] Figure 4 This is a schematic diagram of the method flow provided by the present invention;

[0048] Figure 5 A flowchart illustrating the adaptive genetic algorithm provided by this invention;

[0049] Figure 6 A schematic diagram of fault categories provided by the present invention;

[0050] In the diagram, 1 represents a 550kV repulsion valve type high-voltage circuit breaker; 1-1 is the main shaft of the transmission mechanism; 1-2 is the arc-extinguishing chamber; 2 is the displacement detection module; 2-1 is the permanent magnet; 2-2 is the magnetic induction intensity detection unit; 2-3 is the first communication unit; 3 is the SF6 gas detection module; 4 is the repulsion valve coil detection module; 4-1 is the signal processing unit; 5 is the repulsion valve drive coil; 6 is the feature extraction and standardization system; and 7 is the optical fiber. Detailed Implementation

[0051] 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.

[0052] This embodiment provides a dynamic fault identification device for a repulsion valve type circuit breaker under noisy conditions, such as... Figure 1 As shown, it includes:

[0053] The data acquisition and preprocessing system is used to synchronously acquire and preprocess multi-source signals to obtain displacement time signals, gas composition concentration signals in the arc extinguishing chamber, and opening and closing coil current signals.

[0054] The feature extraction and standardization system 6 is used to perform feature extraction, standardization, data alignment, and feature fusion operations on displacement time signals, gas composition concentration signals in arc extinguishing chambers, and current signals of opening and closing coils to generate fused features.

[0055] The fault identification system is used to identify anomalies by inputting fused features into a trained robust support vector machine model and outputting the fault category.

[0056] In one embodiment, the data acquisition and preprocessing system includes a displacement detection module 2, an SF6 gas detection module 3, and a repulsion valve coil detection module 4;

[0057] The displacement detection module 2 includes a magnetic induction intensity detection unit 2-2 and a permanent magnet 2-1 adsorbed on the main shaft 1-1 of the transmission mechanism being monitored. It is used to detect the change data of magnetic induction intensity between the permanent magnet 2-1 and the magnetic induction intensity detection unit 2-2 in real time, and to perform preprocessing to generate a displacement time signal.

[0058] The SF6 gas detection module 3 includes an electrically connected intake valve and an electrochemical detection unit, which are used to detect the component concentration of various gases in the arc-extinguishing chamber and convert the component concentration of various gases into corresponding voltage signals. After preprocessing the voltage signals, the gas component concentration signal in the arc-extinguishing chamber is obtained.

[0059] The repulsion valve coil detection module 4 includes an electrically connected Rogowski coil and a signal processing unit 4-1. The Rogowski coil is passed through by the repulsion valve drive coil 5. It is used to sense the current of the repulsion valve drive coil 5 and convert it into a corresponding voltage signal. After preprocessing the voltage signal, it generates the opening and closing coil current signal.

[0060] Furthermore, the displacement detection module 2, as... Figure 2It includes: a magnetic induction intensity detection unit 2-2, a permanent magnet 2-1, and a first communication unit 2-3; the permanent magnet 2-1 is attracted to the transmission shaft of the monitored transmission mechanism (i.e., the main shaft 1-1 of the monitored transmission mechanism), and the permanent magnet 2-1 moves with the monitored transmission mechanism; the magnetic induction intensity detection unit 2-2 is electrically connected to the first communication unit 2-3, and the magnetic induction intensity detection unit 2-2 is used to detect the distance between the permanent magnet 2-1 and the magnetic induction intensity detection unit 2-2. The greater the distance between the two, the weaker the magnetic induction intensity detected by the magnetic induction intensity detection unit 2-2; the first communication unit 2-3 is used to transmit the magnetic induction intensity signal (i.e., magnetic induction intensity change data) to the feature extraction and standardization system 6.

[0061] Furthermore, the chip used in the magnetic induction intensity detection unit 2-2 is the RM3100.

[0062] In one embodiment, the permanent magnet 2-1 is made of neodymium iron boron material.

[0063] SF6 gas detection module 3, as shown Figure 3 The device shown includes: an intake valve, an electrochemical detection unit, and a communication unit; the intake valve is fixedly connected to the outlet of the arc-extinguishing chamber 1-2 of the 550kV repulsion valve type high-voltage circuit breaker 1 under test.

[0064] The intake valve is used to draw gas from the arc-extinguishing chamber of the 550kV repulsion valve type high-voltage circuit breaker 1 to be tested, and send the drawn gas into the electrochemical detection unit.

[0065] The electrochemical detection unit includes: an electrochemical sensor and a signal acquisition and processing subunit; the electrochemical sensor comprises a working electrode, an electrolyte, a diaphragm, and a packaging shell; the electrode, electrolyte, and diaphragm are all housed inside the packaging shell; the electrolyte can be acidic, alkaline, or neutral, adapted to the reaction characteristics of the gas being measured; the diaphragm separates the electrodes and allows ion conduction; the signal acquisition and processing subunit acquires the signal output from the electrochemical sensor and amplifies, filters, calibrates, and digitizes the signal;

[0066] The communication unit transmits the signal output from the signal acquisition and processing subunit to the feature extraction and standardization system 6 via optical fiber 7.

[0067] The repulsion valve coil detection module 4 includes: a Rogowski coil and a signal processing unit 4-1; the Rogowski coil is electrically connected to the signal processing unit 4-1;

[0068] The Rogowski coil is passed through the repulsion valve drive coil 5 of the 550kV repulsion valve type high voltage circuit breaker 1, and is used to sense the current flowing through the repulsion valve drive coil 5 and convert it into a corresponding voltage signal.

[0069] The signal processing unit 4-1 filters and amplifies the voltage signal induced by the Rogowski coil, and then transmits it to the feature extraction and standardization system 6 via a wired connection.

[0070] In one embodiment, the feature extraction and standardization system 6 includes a feature extraction unit, a standardization unit, and a data alignment and fusion unit;

[0071] The feature extraction unit is used to extract p-dimensional displacement features based on displacement-time signals and mp-dimensional gas-current features based on the gas composition concentration signal in the arc extinguishing chamber and the current signal of the opening and closing coils, thus forming an m-dimensional original feature vector.

[0072] The standardization unit is used to standardize displacement characteristics and gas-current characteristics to obtain standard displacement characteristics and standard gas-current characteristics.

[0073] The data alignment and fusion unit is used to take standard displacement features as the first set of variables and standard gas-current features as the second set of variables, calculate the covariance matrix of the two sets of variables through canonical correlation analysis, and solve the generalized eigenvalue problem to obtain the typical weight vector; and generate fused features based on the typical weight vector.

[0074] In one embodiment, a diagnostic report generation system is also included for generating a diagnostic report based on the fault category, the diagnostic report including the fault type, location, and maintenance recommendations.

[0075] Another method of this invention discloses a dynamic fault identification method for repulsion valve circuit breakers under noisy backgrounds, such as... Figure 4 The specific steps are as shown:

[0076] Step 1: Synchronously acquire and preprocess the multi-source signals to obtain the displacement time signal, the gas composition concentration signal in the arc extinguishing chamber, and the opening and closing coil current signal;

[0077] Step 2: Perform feature extraction, standardization, data alignment, and feature fusion operations on the displacement time signal, the gas composition concentration signal in the arc extinguishing chamber, and the opening and closing coil current signal to generate fused features;

[0078] Step 3: Input the fused features into the trained robust support vector machine model for anomaly identification and output the fault category.

[0079] In one embodiment, step 1, the acquisition and preprocessing of multi-source signals includes:

[0080] The specific steps for acquiring the displacement-time signal are as follows:

[0081] Based on the real-time detection of magnetic induction intensity change data by displacement detection module 2, the magnetic induction intensity signal is converted into a displacement-time curve, calibrated and filtered to eliminate environmental magnetic field interference, and the displacement-time signal is obtained.

[0082] The specific steps for acquiring the gas composition concentration signal in the arc extinguishing chamber are as follows:

[0083] The gas intake valve draws a gas sample from arc-extinguishing chamber 1-2 and delivers it to the electrochemical detection unit. The electrochemical sensor (working electrode + electrolyte) detects the concentrations of SO2, H2S, CO, and CO2 in the gas. The signal acquisition and processing subunit amplifies, filters, and digitizes the output data from the electrochemical sensor to obtain the gas component concentration signal within the arc-extinguishing chamber. Different electrochemical sensors are used to detect the concentrations of various gases.

[0084] The acquisition of the opening and closing coil current signal is as follows:

[0085] The opening and closing current is acquired and converted into a voltage signal. After being filtered and amplified by the signal processing unit 4-1, the current waveform characteristics (such as peak value, rise time, and duration) are extracted to generate a current-time curve, thus obtaining the opening and closing coil current signal.

[0086] Furthermore, during the acquisition of the gas composition concentration signal in the arc extinguishing chamber, the electrochemical sensor converts the gas concentration signal into a voltage signal. The signal acquisition and processing subunit amplifies the voltage signal, converting it into a voltage signal of 0-3.3V. Due to the presence of interference noise in the signal, filtering is required. The filtered signal is then converted into a digital signal by the ADC module, thus obtaining the gas composition concentration signal in the arc extinguishing chamber.

[0087] The steps for generating the current-time curve are as follows: First, the Rogowski coil induces a current in the driving coil, outputting a weak voltage signal. Then, a low-pass filter is used to remove high-frequency noise from the weak voltage signal, and an operational amplifier is used to amplify the weak signal output by the Rogowski coil. Next, the amplified voltage signal is read through the processor's ADC pin in the repulsion valve coil detection module 4. The processor's timer interrupt is used for ADC sampling, with consistent sampling intervals. An array is allocated in the processor's memory to store the ADC sampled values; after each ADC conversion, the result is stored in the array. Finally, the ADC sampled values ​​are converted into actual current values, and current waveform characteristics (such as peak value, rise time, and duration) are extracted. A time axis is generated based on the sampling rate and the number of sampling points, and the current values ​​are correlated with the time axis to generate the current-time curve.

[0088] In one embodiment, the step of obtaining the fused features is as follows:

[0089] p-dimensional displacement features are extracted based on displacement-time signals, and mp-dimensional gas-current features are extracted by combining gas composition concentration signals in the arc extinguishing chamber and current signals of the opening and closing coils, thus forming an m-dimensional original feature vector.

[0090] The displacement characteristics and gas-current characteristics are standardized to obtain standard displacement characteristics and standard gas-current characteristics.

[0091] Standard displacement characteristics are used as the first set of variables, and standard gas-current characteristics are used as the second set of variables. The covariance matrix of the two sets of variables is calculated by canonical correlation analysis, and the generalized eigenvalue problem is solved to obtain the typical weight vector. The fusion feature is generated based on the typical weight vector.

[0092] Furthermore, in the specific steps of acquiring fused features, the alignment of the three heterogeneous data sources—displacement, gas, and current—includes time alignment and semantic alignment. Time alignment involves using the same pulse generator to trigger acquisition, eliminating acquisition time errors.

[0093] Before fusion, a displacement-gas-current multidimensional feature vector x∈R is first constructed based on the acquired signals. m And it is standardized, and the standardized expression is:

[0094]

[0095] In the formula, Med i IQR represents the median of feature i across all samples. i Let i be the interquartile range of feature i.

[0096] Considering the different feature vector dimensions of displacement, gas, and current signals, canonical correlation analysis (CCA) is used to achieve semantic alignment and feature fusion of multi-source heterogeneous data on the standardized data. Specifically:

[0097] Based on the standardized feature matrix, the displacement signal (p-dimensional features, such as displacement / acceleration), gas + current signal (mp-dimensional features, such as H2S concentration / current rise rate, etc.). Calculate the covariance matrix Σ of X1 and X2, solve the generalized eigenvalue problem, obtain the typical weight vectors a and b, and thus obtain the feature with the greatest correlation.

[0098] X1′=a T X1,X2′=b T X2.

[0099] In one embodiment, the steps for constructing a robust support vector machine model are as follows:

[0100] Acquire multiple sample signals and calculate the mean and variance of the multiple sample signals;

[0101] A set of probability distributions for noisy information is defined based on mean and variance;

[0102] Construct an initial robust chance-constrained support vector machine model based on a set of probability distributions;

[0103] An adaptive genetic algorithm is used to optimize the parameters of the initial robust chance-constrained support vector machine model, thereby generating a robust support vector machine model.

[0104] Furthermore, the steps for constructing a robust support vector machine model with adaptive parameter optimization under noisy conditions are as follows:

[0105] Statistically collected N sample signals x = (x 1 ,...,x N The statistical measures of Σ are: mean μ and variance Σ, and their expressions are:

[0106]

[0107] Let P:={x:E[x]=μ,cov[x]=Σ} denote the set containing the probability distribution of noisy information.

[0108] Given the penalty parameter C, find the optimal robust support vector machine model f(x) under noisy conditions.<w,x> +b enables the sample to identify faults with a high probability at a given confidence level.

[0109]

[0110] Since chance constraints are non-convex, optimization problems with chance constraints are difficult to solve even when the exact probability distribution is known, let alone robust chance constraint problems with unknown distributions.

[0111] Therefore, (RSVM) is transformed into a polynomial-time solvable second-order cone programming problem, and the transformed expression is:

[0112]

[0113] in,

[0114] In one embodiment, an adaptive genetic algorithm is used to optimize the penalty parameter C, and the optimization process is as follows: Figure 5 As shown, the adaptive genetic algorithm needs to adaptively adjust the crossover probability and mutation probability. The formula for calculating the crossover probability is:

[0115]

[0116] The mutation probability is calculated as follows:

[0117]

[0118] In the formula, These represent the upper and lower bounds of the crossover probability, respectively; f max f represents the maximum fitness value in the population; avg This represents the average fitness value of the population. This indicates the fitness value of the parent with the largest fitness value among the parent generations participating in the crossover; These represent the upper and lower bounds of the crossover probability, respectively; f m This indicates the fitness value of the parent with the highest fitness value among the parents participating in the crossover.

[0119] In one embodiment, the fault categories include jamming during opening and closing, insufficient travel of the mechanism, abnormal opening and closing speed, contact wear, poor contact between moving and stationary contacts, deterioration of the insulation of the arc-extinguishing chamber, contact erosion, aging and increased resistance of the opening and closing coil, open circuit of the opening and closing coil, and short circuit between turns of the opening and closing coil.

[0120] Furthermore, such as Figure 6 As shown, the fault categories include: mechanical faults of the operating mechanism: jamming of opening and closing, insufficient travel of the mechanism, and abnormal speed of opening and closing movement; faults of the coil and drive mechanism: aging of the opening and closing coil with increased resistance, open circuit of the opening and closing coil, and short circuit between turns of the opening and closing coil; faults of the arc-extinguishing chamber: severe damage to the contacts (such as contact wear and contact erosion), failure of insulation material (i.e., deterioration of arc-extinguishing chamber insulation), and a large amount of partial discharge (i.e., poor contact between moving and stationary contacts).

[0121] In one embodiment, the method further includes generating a diagnostic report, the specific steps of which are as follows:

[0122] A diagnostic report is generated based on the fault category. The diagnostic report includes the fault type, fault location, and maintenance recommendations.

[0123] Furthermore, this embodiment uses the example of fault diagnosis and maintenance suggestions for excessive SO2 concentration in arc-extinguishing chamber 1-2 of a 550kV SF6 circuit breaker:

[0124] The multi-source status information monitoring device of the SF6 circuit breaker in a 550kV substation detected that the SO2 concentration in arc-extinguishing chamber 1-2 exceeded the standard. The standard value is ≤1ppm, while the measured value is 5ppm.

[0125] The monitoring device's main screen displays the following fault: contact erosion or impurities in arc-extinguishing chambers 1-2;

[0126] Maintenance strategy and steps: (1) Recover the SF6 gas in arc-extinguishing chamber 1-2 and flush arc-extinguishing chamber 1-2 with high-purity nitrogen. (2) Replace the ablated contacts and clean the contact surfaces. (3) Clean the inner wall and components of arc-extinguishing chamber 1-2 with anhydrous ethanol. (4) Refill with new SF6 gas and check that the SO2 concentration drops to 0.5ppm and the humidity is ≤150ppm, which meets the operating standards. (5) After maintenance, the circuit breaker is closed and operates normally, and the SO2 concentration is stable within the standard range. (6) Fault classification judgment and maintenance decision output are completed.

[0127] Excessive SO2 concentration is a significant indicator of SF6 circuit breaker arc-extinguishing chamber malfunction, typically related to contact erosion or poor contact. Such problems can be effectively resolved through maintenance measures such as gas detection, contact replacement, and cleaning of arc-extinguishing chambers 1-2, ensuring safe equipment operation.

[0128] Furthermore, this embodiment uses the fault diagnosis and maintenance suggestions for abnormal current in the tripping drive coil of a 550kV SF6 circuit breaker as an example:

[0129] The multi-source status information monitoring device detected that the drive coil current was less than 10% of the standard value, which is 5A, while the measured value was 0.5A.

[0130] The monitoring device's main screen displays the following faults: coil aging, open circuit, or loose or poor connection between the coil and the external circuit.

[0131] Maintenance strategy and steps: (1) Check if the drive power supply voltage is normal and ensure that the power supply line is fault-free; (2) Check if the connection between the coil and the external circuit is firm, clean the contact surface and tighten it again; (3) Use a multimeter to measure the coil resistance. The standard value is 1Ω and the actual measured value is 10Ω; (4) If the resistance is too high, confirm the coil fault and replace it with a new drive coil of the same model; (5) After replacing the coil, perform a circuit breaker operation test, measure the drive coil current, and ensure that it returns to the standard value range.

[0132] By checking the power supply, measuring the coil resistance, and replacing the faulty coil, such problems can be effectively solved, ensuring the reliability of the circuit breaker's tripping operation.

[0133] In one embodiment, historical fault cases are compared periodically to optimize fault classification rules and threshold settings, thereby improving diagnostic accuracy.

[0134] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0135] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A repulsion valve type circuit breaker fault dynamic identification method in a noisy background, characterized in that, Specific steps are as follows: Synchronous acquisition and preprocessing of multi-source signals are performed to obtain displacement-time signals, gas composition concentration signals in the arc extinguishing chamber and split-charging coil current signals; Feature extraction, standardization, data alignment and feature fusion operations are performed on the displacement-time signals, the gas composition concentration signals in the arc extinguishing chamber and the split-charging coil current signals to generate fusion features; The fusion features are input into a trained robust support vector machine model for abnormality identification, and a fault category is output; The construction steps of the robust support vector machine model are as follows: A plurality of sample signals are obtained, and the mean and variance of the plurality of sample signals are calculated; A probability distribution set of noisy information is defined based on the mean and variance; An initial robust chance-constrained support vector machine model is constructed based on the probability distribution set; The initial robust chance-constrained support vector machine model is parameter-optimized by using an adaptive genetic algorithm to generate the robust support vector machine model; The expression of the initial robust chance-constrained support vector machine model is as follows: ； The transformed expression is as follows: ； where w is a weight vector; b is a bias term; is a slack variable; C is a penalty coefficient; is a label of the kth sample; is a feature vector of the kth sample; m is the total number of samples; is the confidence level; is the mean vector of the kth sample; is a value calculated from the confidence level.

2. The repulsion valve type circuit breaker fault dynamic identification method in a noise-containing background according to claim 1, characterized in that, The acquisition steps of the fusion features are as follows: p-dimensional displacement features are extracted based on the displacement-time signals, and m-p-dimensional gas-current features are extracted based on the joint of the gas composition concentration signals in the arc extinguishing chamber and the split-charging coil current signals to form an m-dimensional original feature vector; Standardization processing is performed on the displacement features and the gas-current features to obtain standard displacement features and standard gas-current features; The standard displacement features are taken as a first group of variables, the standard gas-current features are taken as a second group of variables, a covariance matrix of the two groups of variables is calculated by canonical correlation analysis, and a generalized eigenvalue problem is solved to obtain a canonical weight vector; and the fusion features are generated based on the canonical weight vector.

3. The repulsion valve type circuit breaker fault dynamic identification method in a noise-containing background according to claim 1, characterized in that, The fault category includes split-charging sticking, mechanism stroke deficiency, split-charging movement speed abnormality, contact wear, dynamic-static contact poor contact, arc extinguishing chamber insulation deterioration, contact ablation, split-charging coil aging, split-charging coil wire breakage and split-charging coil inter-turn short circuit.

4. The repulsion valve type circuit breaker fault dynamic identification method in a noisy background according to claim 1, characterized in that, A diagnostic report is generated based on the fault category, and the diagnostic report contains a fault type, a fault position and a maintenance suggestion. The application of the fault dynamic identification method for repulsion valve circuit breakers in a noisy background according to any one of claims 1-4 comprises:

5. A repulsion valve type circuit breaker fault dynamic recognition device in a noise-containing background, characterized in that, A data acquisition and preprocessing system is used for synchronous acquisition and preprocessing of multi-source signals to obtain displacement-time signals, gas composition concentration signals in the arc extinguishing chamber and split-charging coil current signals; A feature extraction and standardization system (6) is used for feature extraction, standardization, data alignment and feature fusion operations on the displacement-time signals, the gas composition concentration signals in the arc extinguishing chamber and the split-charging coil current signals to generate fusion features; A fault identification system is used for inputting the fusion features into a trained robust support vector machine model for abnormality identification, and outputting a fault category. The data acquisition and preprocessing system comprises a displacement detection module (2), an SF6 gas detection module (3) and a repulsion valve coil detection module (4).

6. A repulsion valve type circuit breaker fault dynamic identification device in a noisy background according to claim 5, characterized in that, ​ The displacement detection module (2) comprises a magnetic induction intensity detection unit (2-2) and a permanent magnet (2-1) adsorbed on the monitored transmission mechanism main shaft (1-1), which is used for detecting the magnetic induction intensity change data between the permanent magnet (2-1) and the magnetic induction intensity detection unit (2-2) in real time, and pre-processing to generate the displacement time signal; The SF6 gas detection module (3) comprises an adsorption valve and an electrochemical detection unit connected in an electrical manner, which is used for detecting the component concentration of various gases in the arc extinguishing chamber, and converting the component concentration of various gases into corresponding voltage signals, and obtaining the gas component concentration signal in the arc extinguishing chamber after pre-processing the voltage signals corresponding to the component concentration of various gases; The repulsive force valve coil detection module (4) comprises a Rogowski coil and a signal processing unit (4-1) connected in an electrical manner, and the Rogowski coil is penetrated by the repulsive force valve driving coil (5); which is used for inducting the current of the repulsive force valve driving coil (5), and converting it into corresponding voltage signals, and generating the closing and opening coil current signal after pre-processing the voltage signals corresponding to the current of the repulsive force valve driving coil (5).

7. A repulsion valve type circuit breaker fault dynamic recognition device in a noisy background according to claim 5, characterized in that, The feature extraction and standardization system (6) comprises a feature extraction unit, a standardization unit and a data alignment and fusion unit; The feature extraction unit is used for extracting p-dimensional displacement features based on the displacement time signal, and extracting m-p-dimensional gas-current features based on the arc extinguishing chamber gas component concentration signal and the closing and opening coil current signal, to form an m-dimensional original feature vector; The standardization unit is used for standardizing the displacement features and the gas-current features to obtain standard displacement features and standard gas-current features; The data alignment and fusion unit is used for taking the standard displacement features as a first group of variables, and taking the standard gas-current features as a second group of variables, calculating the covariance matrix of the two groups of variables through canonical correlation analysis, and solving the generalized eigenvalue problem to obtain a canonical weight vector; and generating a fusion feature based on the canonical weight vector.

8. The repulsive valve-type circuit breaker fault dynamic identification device in a noisy background according to claim 5, characterized in that, It also comprises a diagnosis report generation system, which is used for generating a diagnosis report according to the fault category, and the diagnosis report contains the fault type, location and maintenance suggestion.

Images