Circuit breaker contact erosion amount evaluation method and device based on multi-dimensional information fusion

Abstract

The application discloses a circuit breaker contact ablation amount evaluation method and device based on multi-dimensional information fusion, in the method, each dimension information in the action process of the circuit breaker is collected, and each dimension information is pretreated, wherein the multi-dimensional information is subjected to time scale alignment, sampling rate normalization and signal strength normalization; the ablation amount of the contact is evaluated based on a feature fusion algorithm based on a feature parameter, wherein time-frequency domain energy features are extracted based on a vibration signal to be converted into a feature signal, a feature parameter is obtained based on the feature signal, and the ablation amount of the contact is obtained by evaluating the measured data of the circuit breaker based on the feature fusion algorithm.

Description

Technical Field

[0001] This invention belongs to the field of GIS high-voltage circuit breaker contact electrical life assessment technology, and in particular, a method and device for assessing the erosion of circuit breaker contacts based on multi-dimensional information fusion. Background Technology

[0002] For GIS high-voltage circuit breakers, the arc extinguishing process is completed by the contact system during the breaking process. The contact system is divided into main contacts and arc contacts. The main contacts are responsible for carrying current and are made of materials with good conductivity to reduce current carrying losses. The arc contacts are responsible for carrying the electric arc during breaking and closing and are made of materials resistant to ablation. During opening, the main contacts break before the arc contacts, and the arc contacts raise the arc, which is then extinguished in SF6 gas. During closing, the arc contacts make contact before the main contacts, and the voltage at the breaking point is expected to break through the arc contacts. After the arc contacts make contact, the main contacts make further contact and assume the current carrying function.

[0003] After repeated interruptions, the arc contact wears down and gradually shortens due to erosion. When the arc contact is eroded to a certain extent, it can no longer effectively bear the arc. At this point, the arc during the interruption process may cause the main contact to erode, thereby damaging the function of the entire contact system. Therefore, assessing the degree of arc contact erosion in the contact system is of great practical significance.

[0004] Arc energy is a key factor causing arc contact erosion, but it is difficult to measure. Currently, breaking current is commonly used to assess electrical lifetime, mainly including the cumulative breaking current weighted method and the arc time weighted assessment method. b Life curve method, etc. The dynamic resistance method measures the change in resistance during the opening and closing of the circuit breaker, which can intuitively reflect the burning condition of the arc contacts in the arc-extinguishing chamber of the circuit breaker. However, it can only be measured when the circuit breaker is not in operation and cannot achieve online monitoring of electrical life.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of the present invention, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention proposes a method and device for evaluating the erosion of circuit breaker contacts based on multi-dimensional information fusion, thereby achieving high-precision online monitoring of the electrical life of arc contacts.

[0007] The objective of this invention is achieved through the following technical solution: a method for evaluating the erosion of circuit breaker contacts based on multi-dimensional information fusion includes:

[0008] Step S100: Collect information in various dimensions during the operation of the circuit breaker, including vibration signal, angular displacement signal, opening and closing coil current signal, circuit break signal and circuit dynamic resistance signal.

[0009] Step S200: Preprocess the information of each dimension, including time-stamp alignment, sampling rate normalization and signal strength normalization of the information of each dimension, digital filtering of the vibration signal, and time-frequency domain transformation to convert the time domain signal to the time-frequency domain.

[0010] Step S300: Based on the preprocessed information of each dimension, feature extraction and feature value calibration are performed. Specifically, the circuit break signal is extracted to obtain the contact point of the circuit breaker arc contact and the corresponding main shaft angular displacement. The circuit dynamic resistance signal is extracted to obtain the main shaft angular displacement corresponding to the contact point of the circuit breaker main contact. The main shaft angular displacement corresponding to the main contact is labeled. The main shaft angular displacement is converted to the time domain. The angular displacements of the arc contact and the main contact are converted to the vibration signal.

[0011] Step S400: Evaluate the contact ablation amount using a feature fusion algorithm based on feature parameters. This involves extracting time-frequency domain energy features from the vibration signal and converting them into feature signals, obtaining feature parameters based on these feature signals, and then evaluating the circuit breaker's measured data using the feature fusion algorithm to obtain the contact ablation amount. Specifically, this involves extracting time-frequency domain energy features from the vibration signal and converting them into feature signals.

[0012]

[0013] In the formula, F(t) is the characteristic signal of the vibration signal, and E f (t) represents the energy of the vibration signal at a specific frequency band at time t, E total (t) represents the total energy of the vibration signal at time t. Based on the acquired characteristic signal, the characteristic parameters are obtained as follows:

[0014]

[0015] In the formula, F(t) is the characteristic signal of the vibration signal. Let be the energy of the vibration signal at a specific frequency band at time t. These are the characteristic parameters obtained by combining two sets of signals along the time axis.

[0016] In the circuit breaker contact erosion assessment method based on multi-dimensional information fusion, in step S200, the sampling rate matching adopts the interpolation method, which performs linear interpolation on the low sampling signals in each dimension of information by integer multiples, so as to match the high sampling rate signals of each dimension of information.

[0017] In the method for evaluating the erosion of circuit breaker contacts based on multi-dimensional information fusion, the time-frequency domain transformation method in step S200 includes S-transform, short-time Fourier transform, and wavelet transform.

[0018] In the aforementioned method for evaluating the erosion of circuit breaker contacts based on multi-dimensional information fusion, the S-transformation,

[0019] The continuous S-transform of the signal r(t) to be measured is:

[0020]

[0021] In the formula, t is the signal sampling time, f is the signal frequency, τ is the translation factor, r(t) is the signal to be measured, π is pi, e is the natural constant, j is the imaginary unit, and S is the result of the continuous S-transform.

[0022] For the signal r(t) to be measured, its discrete S-transform is:

[0023]

[0024]

[0025] Where x is the time sampling point number at a certain time interval, y is the frequency sampling point number at a certain frequency interval, T is the sampling period, N is the number of sampling points, R(k) is the discrete Fourier transform of the time series r(t); k is the frequency domain sampling point of the discrete Fourier transform; m is the intermediate variable in the calculation process, π is pi, e is the natural constant, and is the imaginary unit, and S is the result of the discrete S-transform.

[0026] In the aforementioned method for evaluating the erosion of circuit breaker contacts based on multi-dimensional information fusion, the feature fusion algorithm includes ANN algorithm, SVM algorithm, KNN algorithm, RVM algorithm, deep learning neural network algorithm, and Bayesian network algorithm.

[0027] In the circuit breaker contact erosion assessment method based on multi-dimensional information fusion, the feature vector of the training samples in the RVM algorithm is: The target vector is The classification function model of RVM is:

[0028]

[0029] In the formula, x is the input training sample set, x i Let be the i-th training sample; n be the number of samples; ω be the weight vector, ω = [ω0, ω1, ..., ω2]. n ] T ω i Let ω be the i-th component of ω; ω0 be the initial weights; and y be the classifier output value; K(x, x)i Let ω be the kernel function. Introducing the S-function into the classification model, and assuming P(t|ω) follows a Bernoulli distribution, the likelihood estimate is:

[0030]

[0031] Here, the probability function σ(y) of the Bernoulli distribution P(t|ω) is the probability distribution function of the S-function. Within the Bayesian framework, weights are obtained through the maximum likelihood method, using a Gaussian prior distribution. Constraint parameters were defined for each weight to achieve a smooth model.

[0032]

[0033] In the formula, η=[η0, η1,..., η n ] T For each weight, a hyperparameter is introduced as an n+1 dimensional hyperparameter, thus generating a sparse probabilistic model.

[0034] Initialize the hyperparameter η;

[0035] A Gaussian approximation is established on the posterior probability based on the likelihood estimation probability, resulting in an approximation of the marginal likelihood. The marginal likelihood function of the approximation is maximized, leading to a re-estimation of the hyperparameter vector values. This process is repeated until the hyperparameter vector values ​​converge.

[0036] The data to be predicted is predicted using the classification function model of RVM to obtain the prediction results.

[0037] An apparatus for implementing the circuit breaker contact erosion assessment method based on multi-dimensional information fusion includes,

[0038] The shock-absorbing base is supported on the ground;

[0039] A hydraulic spring operating mechanism, which is supported on the shock-absorbing base;

[0040] The opening and closing coils are mounted on the hydraulic spring operating mechanism;

[0041] Circuit breaker, which is connected to the opening and closing coil;

[0042] A simulated load is applied to the circuit breaker in a controllable manner;

[0043] The sensor interface module includes,

[0044] A voltage sensor, connected to the opening and closing coils, measures the circuit break signal.

[0045] An angular displacement sensor is mounted on the main shaft of the circuit breaker to measure angular displacement signals.

[0046] A vibration sensor is installed at the opening and closing coil to measure vibration signals.

[0047] A current sensor is connected to the opening and closing coil to measure the current signal of the opening and closing coil.

[0048] A loop dynamic resistance measurement unit is connected to the opening and closing coils and the circuit breaker to generate a loop dynamic resistance signal.

[0049] An embedded acquisition module is connected to the sensor interface module to acquire and preprocess vibration signals, angular displacement signals, opening and closing coil current signals, circuit break signals, and circuit dynamic resistance signals. The embedded acquisition module includes an analog-to-digital converter, a digital isolator, a microcontroller for sampling control, a communication unit for signal transmission and storage, and a storage unit.

[0050] A cloud computing module, which is communicatively connected to the communication unit, the cloud computing module comprising...

[0051] The feature extraction and feature value labeling unit extracts and labels features based on the preprocessed information from each dimension.

[0052] An evaluation algorithm deployment platform is connected to the feature extraction and feature value calibration unit to evaluate the ablation amount of the contact using a feature fusion algorithm.

[0053] Compared with existing technologies, this invention has the following advantages: The circuit breaker contact erosion assessment method based on multi-dimensional information fusion, combined with a machine learning-based diagnostic identification algorithm, can assess the degree of erosion of arc contacts. First, key information related to the circuit breaker contacts is acquired, including vibration signals during opening and closing, angular displacement of the circuit breaker main shaft, current signals in the opening and closing coil circuit, dynamic resistance of the main circuit, and voltage signals at the break point. Then, features of various signals at different times are extracted, and the features of the contact segment are integrated and summarized to obtain feature parameters. Finally, a classification algorithm is used to obtain the degree of erosion of the arc contacts. Furthermore, this invention innovatively uses vibration signals, angular displacement signals, and other multi-dimensional information for fusion to achieve the assessment of contact erosion. The detection device of this invention has the advantages of non-invasiveness and online monitoring, and the assessment method is flexible, meeting the needs of online monitoring of the electrical life of high-voltage switchgear. Attached Figure Description

[0054] Various other advantages and benefits of the present invention will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiments below. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort. Furthermore, the same reference numerals denote the same parts throughout the drawings.

[0055] In the attached diagram:

[0056] Figure 1 This is a flowchart of a circuit breaker contact erosion assessment method based on multi-dimensional information fusion according to an embodiment of the present invention;

[0057] Figure 2 This is a schematic diagram of a circuit breaker contact erosion assessment device based on multi-dimensional information fusion according to an embodiment of the present invention.

[0058] Figure 3 This is a schematic diagram of multi-dimensional information acquisition for a circuit breaker contact erosion assessment method based on multi-dimensional information fusion according to an embodiment of the present invention.

[0059] Figure 4 This is a schematic diagram of the morphology of signals in each dimension after collecting and integrating multi-dimensional information in a circuit breaker contact erosion assessment method based on multi-dimensional information fusion according to an embodiment of the present invention.

[0060] Figure 5 This is a flowchart illustrating the preprocessing and data conversion of information from various dimensions in a circuit breaker contact erosion assessment method based on multi-dimensional information fusion according to an embodiment of the present invention.

[0061] Figure 6 This is a time-frequency diagram obtained after time-frequency domain transformation of a circuit breaker contact erosion assessment method based on multi-dimensional information fusion according to an embodiment of the present invention.

[0062] Figure 7 This is a flowchart illustrating the extraction of information features from various dimensions in a circuit breaker contact ablation assessment method based on multi-dimensional information fusion according to an embodiment of the present invention.

[0063] Figure 8 This is a flowchart illustrating the integrated feature parameters and feature algorithm deployment of a circuit breaker contact erosion assessment method based on multi-dimensional information fusion according to an embodiment of the present invention.

[0064] Figure 9 This is a schematic diagram of characteristic signals of a circuit breaker contact erosion assessment method based on multi-dimensional information fusion according to an embodiment of the present invention.

[0065] The present invention will be further explained below with reference to the accompanying drawings and embodiments. Detailed Implementation

[0066] The following will refer to the appendix. Figures 1 to 9 Specific embodiments of the invention will be described in more detail below. While specific embodiments of the invention are shown in the accompanying drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the invention and to fully convey the scope of the invention to those skilled in the art.

[0067] It should be noted that certain terms are used in the specification and claims to refer to specific components. Those skilled in the art will understand that different terms may be used to refer to the same component. This specification and claims do not distinguish components based on differences in terminology, but rather on differences in function. The terms "comprising" or "including" used throughout the specification and claims are open-ended and should be interpreted as "comprising but not limited to." The following descriptions are preferred embodiments for carrying out the invention; however, these descriptions are for the purpose of understanding the general principles of the specification and are not intended to limit the scope of the invention. The scope of protection of this invention is determined by the appended claims.

[0068] To facilitate understanding of the embodiments of the present invention, further explanations and descriptions will be provided below with reference to the accompanying drawings and specific embodiments. The accompanying drawings do not constitute a limitation on the embodiments of the present invention.

[0069] To better understand, such as Figures 1 to 9 As shown, the method for assessing the contact erosion of circuit breakers based on multi-dimensional information fusion includes:

[0070] Step S100: Collect information in various dimensions during the operation of the circuit breaker, including vibration signal, angular displacement signal, opening and closing coil current signal, circuit break signal and circuit dynamic resistance signal.

[0071] Step S200: Preprocess the information of each dimension, including time-stamp alignment, sampling rate normalization and signal strength normalization of the information of each dimension, digital filtering of the vibration signal, and time-frequency domain transformation to convert the time domain signal to the time-frequency domain.

[0072] Step S300: Based on the preprocessed information of each dimension, feature extraction and feature value calibration are performed. Specifically, the circuit break signal is extracted to obtain the contact point of the circuit breaker arc contact and the corresponding main shaft angular displacement. The circuit dynamic resistance signal is extracted to obtain the main shaft angular displacement corresponding to the contact point of the circuit breaker main contact. The main shaft angular displacement corresponding to the main contact is labeled. The main shaft angular displacement is converted to the time domain. The angular displacements of the arc contact and the main contact are converted to the vibration signal.

[0073] Step S400: The ablation amount of the contact is evaluated based on the feature parameters using a feature fusion algorithm. Specifically, time-frequency domain energy features are extracted from the vibration signal and converted into feature signals. Feature parameters are obtained based on the feature signals. The measured data of the circuit breaker are evaluated based on the feature fusion algorithm to obtain the ablation amount of the contact.

[0074] In a preferred embodiment of the circuit breaker contact erosion assessment method based on multi-dimensional information fusion, in step S200, the sampling rate matching adopts the interpolation method, which performs linear interpolation on the low sampling signals in each dimension of information by integer multiples, so as to match the high sampling rate signals of each dimension of information.

[0075] In a preferred embodiment of the circuit breaker contact erosion assessment method based on multi-dimensional information fusion, the time-frequency domain transformation method in step S200 includes S-transform, short-time Fourier transform, and wavelet transform.

[0076] In a preferred embodiment of the circuit breaker contact erosion assessment method based on multi-dimensional information fusion, in the S-transform,

[0077] The continuous S-transform of the signal r(t) to be measured is:

[0078]

[0079] In the formula, t is the signal sampling time, f is the specific frequency of the signal, τ is the translation factor, r(t) is the signal to be measured, π is pi, e is the natural constant, j is the imaginary unit, and S is the result of the continuous S-transform.

[0080] For the signal r(t) to be measured, its discrete S-transform is:

[0081]

[0082]

[0083] Where x is the time sampling point number at a certain time interval, y is the frequency sampling point number at a certain frequency interval, T is the sampling period, N is the number of sampling points, R(k) is the discrete Fourier transform of the time series r(t), k is the frequency domain sampling point of the discrete Fourier transform, m is the intermediate variable in the calculation process, π is pi, e is the natural constant, j is the imaginary unit, and S is the result of the discrete S-transform.

[0084] In a preferred embodiment of the circuit breaker contact erosion assessment method based on multi-dimensional information fusion, the feature fusion algorithm includes ANN algorithm, SVM algorithm, KNN algorithm, RVM algorithm, deep learning neural network algorithm, and Bayesian network algorithm.

[0085] In a preferred embodiment of the circuit breaker contact erosion assessment method based on multi-dimensional information fusion, the feature vector of the training samples in the RVM algorithm is: The target vector is The classification function model of RVM is:

[0086]

[0087] In the formula, x is the input training sample set, x i Let be the i-th training sample; n be the number of samples; ω be the weight vector, ω = [ω0, ω1, ..., ω2]. n ] T ω i Let ω be the i-th component of ω; ω0 be the initial weights; and y be the classifier output value; K(x, x) i Let ω be the kernel function. Introducing the S-function into the classification model, and assuming P(t|ω) follows a Bernoulli distribution, the likelihood estimate is:

[0088]

[0089] Where P(t|ω) is the probability function of the Bernoulli distribution, and σ(y) is the probability distribution function of the S-function. Within the Bayesian framework, weights are obtained through the maximum likelihood method, using a Gaussian prior distribution. Constraint parameters were defined for each weight to achieve a smooth model.

[0090]

[0091] In the formula, η=[η0, η1,..., η n ] T For each weight, a sparse probabilistic model is generated by introducing a hyperparameter in n+1 dimensions.

[0092] In one embodiment, the method includes,

[0093] Step S100: Collect information from various dimensions during the circuit breaker's operation;

[0094] Step S200: Preprocess the information in each dimension, align and transform the data;

[0095] Step S300: Data feature extraction and feature value calibration;

[0096] Step S400: Integrate feature parameters and deploy feature algorithms.

[0097] Furthermore, step S200 includes:

[0098] Step S201: Perform time-stamp alignment, sampling rate normalization, and signal strength normalization on each data item;

[0099] Step S202: Perform digital filtering on the vibration signal and use time-frequency domain transformation to convert the time-domain signal to the time-frequency domain.

[0100] Furthermore, step S300 includes:

[0101] Step S301: Extract the break signal to obtain the contact point of the circuit breaker arc contact and the corresponding spindle angular displacement;

[0102] Step S302: Extract the dynamic loop resistance to obtain the main shaft angular displacement corresponding to the contact point of the circuit breaker main contact;

[0103] Step S303: Convert the principal axis angular displacement of the key node to the time domain and label and integrate it on other signals.

[0104] Furthermore, step S400 includes:

[0105] Step S401: Design features and calculate feature parameters;

[0106] Step S402: Using known features and feature parameters, select a diagnostic algorithm to diagnose the measured data.

[0107] In a specific implementation of step S100, multiple data points need to be acquired, including: vibration signals, angular displacement signals, opening and closing coil current signals, circuit break signals, and circuit dynamic resistance signals. The acquisition device in this invention has multi-channel acquisition capabilities, and its hardware structure is as follows: Figure 2As shown in the diagram, the device consists of a multi-dimensional information acquisition unit and an evaluation algorithm deployment platform. The multi-dimensional information acquisition unit can acquire current signals, vibration signals, and angular displacement signals, and includes an analog-to-digital converter and a digital isolator. Sampling control is performed by a microcontroller, and data transmission and storage are handled by a communication unit and a storage unit. The evaluation algorithm deployment platform receives the acquired multi-dimensional information, as well as some external data, integrates and summarizes it, extracts data features, compares the obtained feature data with empirical parameters, and obtains the evaluation result. A schematic diagram of the multi-dimensional information acquisition is shown below. Figure 3 As shown, after acquisition and integration, the forms of signals in each dimension are as follows: Figure 4 As shown.

[0108] In a specific implementation of step S200, the specific process of preprocessing and data transformation of information in each dimension is as follows: Figure 5 As shown, it specifically includes:

[0109] Step S201: Perform time-stamp alignment, sampling rate normalization, and signal strength normalization on each data item:

[0110] Since multidimensional information may be collected by different acquisition devices, and the sampling timescale and sampling rate may vary, it is first necessary to align the sampling rate and timescale of data from different sources. Sampling rate alignment uses interpolation, performing linear interpolation on low-sampled signals at integer multiples to adapt to high-sampled-rate signals. This aligns the signals from multiple devices at the same time.

[0111] Step S202: Perform digital filtering on the vibration signal and convert the time-domain signal to the time-frequency domain using a time-frequency domain transformation method. In practical applications, the transformation method can include, but is not limited to, S-transform, short-time Fourier transform, wavelet transform, etc. In this embodiment, the S-transform is used as an example for explanation:

[0112] The continuous S-transform of the signal r(t) to be measured is:

[0113]

[0114] In the formula, t is the signal sampling time, f is the specific frequency of the signal, τ is the translation factor, r(t) is the signal to be measured, π is pi, e is the natural constant, j is the imaginary unit, and S is the result of the continuous S-transform.

[0115] For the signal r(t) to be measured, its discrete S-transform is:

[0116]

[0117]

[0118] Where x is the time sampling point number at a certain time interval, y is the frequency sampling point number at a certain frequency interval, T is the sampling period, N is the number of sampling points, R(k) is the discrete Fourier transform of the time series r(t), k is the frequency domain sampling point of the discrete Fourier transform, m is the intermediate variable in the calculation process, π is pi, e is the natural constant, j is the imaginary unit, and S is the result of the discrete S-transform.

[0119] The time-frequency diagram obtained after the S-transform is as follows: Figure 6 As shown.

[0120] In a specific implementation of step S300, the process of extracting information features from each dimension is as follows: Figure 7 As shown, it includes:

[0121] Step S301: Extract the break signal to obtain the contact point of the circuit breaker arc contact and the corresponding angular displacement of the main shaft. The break signal is an approximate step signal; the moment of its inflection is taken as the electrical contact moment, and the angular displacement at that moment is obtained.

[0122] Step S302: Extract the dynamic loop resistance to obtain the main shaft angular displacement corresponding to the main contact point of the circuit breaker, and mark the main shaft angular displacement corresponding to the main contact point.

[0123] Step S303: Convert the main shaft angular displacement of the key node to the time domain and mark and integrate it on other signals; convert the angular displacement of the arc contact point and the main contact point to the vibration signal.

[0124] In a specific implementation of step S400, the integration of feature parameters and the specific implementation process of feature algorithm deployment are as follows: Figure 8 As shown, it specifically includes:

[0125] Step S401: Extract time-frequency domain energy features from the vibration signal, convert the original signal into a feature signal, and use the feature signal to reflect the key nodes of the contact, such as... Figure 9 As shown. The specific process of extracting time-frequency domain energy features from vibration signals and converting them into feature signals is as follows:

[0126]

[0127] In the formula, F(t) is the characteristic signal of the vibration signal, and E f (t) represents the energy of a specific frequency band of the vibration signal at time t. This characteristic frequency band is formed by the difference in the high-frequency portion after Fourier transform of the contact segment signal. In this embodiment, the characteristic frequency band is [7kHz, 12kHz]. E total (t) represents the total energy of the vibration signal at time t. Based on the acquired characteristic signal, the characteristic parameters are obtained as follows:

[0128]

[0129] In the formula, F(t) is the characteristic signal of the vibration signal. Let be the energy of the vibration signal at a specific frequency band at time t. This is the feature parameter vector obtained by combining two sets of signals along the time axis.

[0130] Step S402: Based on the feature parameters obtained from the feature signals, various feature fusion algorithms can be selected to evaluate the measured data and obtain the ablation amount of the contact, such as: ANN, SVM, KNN, RVM, deep learning neural networks, Bayesian networks, etc. In this embodiment, the RVM algorithm is used as an example for illustration, but it is not limited to this method in actual applications. The calculation process is implemented on the evaluation algorithm deployment platform. The evaluation algorithm deployment platform mainly consists of a computing unit that can deploy multiple evaluation algorithms and a communication unit that summarizes the feature parameter data. In this embodiment, the EAIDK-610 edge computing platform is used. The EAIDK-610 platform uses RK3399 as the main chip, the CPU uses a quad-core Cortex-A53 + dual-core Cortex-A72, the GPU uses a quad-core image processor Mali-T860, the CPU frequency is 1.8GHz, and the running memory uses dual-channel LPDDR3 (64-bit) 4GB. In actual applications, it is not limited to this device.

[0131] The RVM algorithm is a sparse probabilistic model based on a global Bayesian framework, assuming that the feature vectors of the training samples are... The target vector is The classification function model of RVM is defined as follows:

[0132]

[0133] In the formula, x is the input training sample set, x i Let be the i-th training sample; n be the number of samples; ω be the weight vector, ω = [ω0, ω1, ..., ω2]. n ] T ω i Let ω be the i-th component of ω; ω0 be the initial weights; and y be the classifier output value; K(x, x) i The kernel function is σ. The logistic sigmoid function is introduced into the classification model, assuming that P(t|ω) follows a Bernoulli distribution. The classification problem here does not include the noise variable σ. 2 The likelihood estimate is:

[0134]

[0135] Where P(t|ω) is the probability function of the Bernoulli distribution, σ(y) = 1 / (1+e^(-ω / ω)). -y Let be the probability distribution function of the S-function. Within the Bayesian framework, the weights can be obtained through the maximum likelihood method. To prevent overlearning, RVM uses a Gaussian prior distribution. Constraint parameters were defined for each weight to achieve a smooth model.

[0136]

[0137] In the formula, η=[η0, η1,..., η n ] T For each weight, a sparse probabilistic model is generated by introducing a hyperparameter in n+1 dimensions.

[0138] The learning and prediction process for evaluating circuit breaker measured data using RVM is as follows:

[0139] (1) Initialize the hyperparameter vector η;

[0140] (2) Based on equation (4), establish a Gaussian approximation for the posterior probability to obtain an approximation for the marginal likelihood;

[0141] (3) Maximize the approximate marginal likelihood function and introduce a re-estimation of the hyperparameter vector values;

[0142] (4) Repeat steps (2) and (3) until the hyperparameter vector values ​​converge.

[0143] (5) The data to be predicted is predicted according to equation (3) to obtain the prediction result.

[0144] The results of the contact ablation assessment based on the measured data are shown in Table 1.

[0145] Table 1

[0146]

[0147] The contact erosion assessment device based on multidimensional information fusion described in this invention effectively achieves online assessment of arc contact erosion through multidimensional information fusion, feature extraction, feature parameter tuning, and a high-voltage GIS contact erosion diagnosis model based on deep learning. Compared with other methods in the same field, this invention has the advantages of online monitoring and non-invasiveness, laying a solid foundation for fault diagnosis of high-voltage power equipment.

[0148] An apparatus for implementing the circuit breaker contact erosion assessment method based on multi-dimensional information fusion includes,

[0149] The shock-absorbing base is supported on the ground;

[0150] A hydraulic spring operating mechanism, which is supported on the shock-absorbing base;

[0151] The opening and closing coils are mounted on the hydraulic spring operating mechanism;

[0152] Circuit breaker, which is connected to the opening and closing coil;

[0153] A simulated load is applied to the circuit breaker in a controllable manner;

[0154] The sensor interface module includes,

[0155] A voltage sensor, connected to the opening and closing coils, measures the circuit break signal.

[0156] An angular displacement sensor is mounted on the main shaft of the circuit breaker to measure angular displacement signals.

[0157] A vibration sensor is installed at the opening and closing coil to measure vibration signals.

[0158] A current sensor is connected to the opening and closing coil to measure the current signal of the opening and closing coil.

[0159] A loop dynamic resistance measurement unit is connected to the opening and closing coils and the circuit breaker to generate a loop dynamic resistance signal.

[0160] An embedded acquisition module is connected to the sensor interface module to acquire and preprocess vibration signals, angular displacement signals, opening and closing coil current signals, circuit break signals, and circuit dynamic resistance signals. The embedded acquisition module includes an analog-to-digital converter, a digital isolator, a microcontroller for sampling control, a communication unit for signal transmission and storage, and a storage unit.

[0161] A cloud computing module, which is communicatively connected to the communication unit, the cloud computing module comprising...

[0162] The feature extraction and feature value labeling unit extracts and labels features based on the preprocessed information from each dimension.

[0163] An evaluation algorithm deployment platform is connected to the feature extraction and feature value calibration unit to evaluate the ablation amount of the contact using a feature fusion algorithm.

[0164] Although embodiments of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the specific embodiments and application fields described above. The specific embodiments described above are merely illustrative and instructive, and not restrictive. Those skilled in the art can make many other forms based on the guidance of this specification and without departing from the scope of protection of the claims of the present invention, and all of these are within the scope of protection of the present invention.

Claims

1. A method for evaluating the erosion of circuit breaker contacts based on multi-dimensional information fusion, characterized in that, It includes the following steps, Step S100: Collect information in various dimensions during the operation of the circuit breaker, including vibration signal, angular displacement signal, opening and closing coil current signal, circuit break signal and circuit dynamic resistance signal. Step S200: Preprocess the information of each dimension, including time-stamp alignment, sampling rate normalization and signal strength normalization of the information of each dimension, digital filtering of the vibration signal, and time-frequency domain transformation to convert the time domain signal to the time-frequency domain. Step S300: Based on the preprocessed information of each dimension, feature extraction and feature value calibration are performed. Specifically, the circuit break signal is extracted to obtain the contact point of the circuit breaker arc contact and the corresponding main shaft angular displacement. The circuit dynamic resistance signal is extracted to obtain the main shaft angular displacement corresponding to the contact point of the circuit breaker main contact. The main shaft angular displacement corresponding to the main contact is labeled. The main shaft angular displacement is converted to the time domain. The angular displacements of the arc contact point and the main contact point are converted to the vibration signal. Step S400: Evaluate the contact ablation amount using a feature fusion algorithm based on feature parameters. This involves extracting time-frequency domain energy features from the vibration signal and converting them into feature signals, obtaining feature parameters based on these feature signals, and then evaluating the circuit breaker's measured data using the feature fusion algorithm to obtain the contact ablation amount. Specifically, this involves extracting time-frequency domain energy features from the vibration signal and converting them into feature signals. In the formula, F(t) is the characteristic signal of the vibration signal, and E f (t) represents the energy of the vibration signal at a specific frequency band at time t, E total (t) represents the total energy of the vibration signal at time t. Based on the acquired characteristic signal, the characteristic parameters are obtained as follows: In the formula, F(t) is the characteristic signal of the vibration signal. Let be the energy of the vibration signal at a specific frequency band at time t. These are the characteristic parameters obtained by combining two sets of signals along the time axis.

2. The method for evaluating the erosion of circuit breaker contacts based on multi-dimensional information fusion according to claim 1, wherein, Preferably, in step S200, the sampling rate matching adopts the interpolation method, which performs linear interpolation on the low sampling signals in each dimension of information by multiples of integers, so as to match the high sampling rate signals of each dimension of information.

3. The method for evaluating the erosion of circuit breaker contacts based on multi-dimensional information fusion according to claim 1, wherein, In step S200, the time-frequency domain transformation methods include S-transform, short-time Fourier transform, and wavelet transform.

4. The method for evaluating the erosion of circuit breaker contacts based on multi-dimensional information fusion according to claim 3, wherein, In the S-transform The continuous S-transform of the signal r(t) to be measured is: In the formula, t is the signal sampling time, f is the signal frequency, τ is the translation factor, r(t) is the signal to be measured, π is pi, e is the natural constant, j is the imaginary unit, and S is the result of continuous S-transformation; For the signal r(t) to be measured, its discrete S-transform is: Where x is the time sampling point number at a certain time interval, y is the frequency sampling point number at a certain frequency interval, T is the sampling period, N is the number of sampling points, R(k) is the discrete Fourier transform of the time series r(t), k is the frequency domain sampling point of the discrete Fourier transform, m is the intermediate variable in the calculation process, π is pi, e is the natural constant, j is the imaginary unit, and S is the result of the discrete S-transform.

5. The method for evaluating the erosion of circuit breaker contacts based on multi-dimensional information fusion according to claim 1, wherein, The feature fusion algorithms include ANN algorithm, SVM algorithm, KNN algorithm, RVM algorithm, deep learning neural network algorithm, and Bayesian network algorithm.

6. The method for evaluating the erosion of circuit breaker contacts based on multi-dimensional information fusion according to claim 5, wherein, In the RVM algorithm, the feature vector of the training samples is The target vector is The classification function model of RVM is: In the formula, x is the input training sample set, x i Let be the i-th training sample; n be the number of samples; ω be the weight vector, ω = [ω0, ω1, ..., ω2]. n ] T ω i Let ω be the i-th component of ω; ω0 is the initial weight, and y is the classifier output value; K(x, x) i Let ω be the kernel function. Introducing the S-function into the classification model, and assuming P(t|ω) follows a Bernoulli distribution, the likelihood estimate is: Where P(t|ω) is the probability function of the Bernoulli distribution, and σ(y) is the probability distribution function of the S-function. Within the Bayesian framework, the weights are obtained through the maximum likelihood method, using a Gaussian prior distribution. Constraint parameters were defined for each weight to achieve a smooth model. In the formula, η = [η0, η1,..., η n ] T For each weight, a sparse probabilistic model is generated by introducing a hyperparameter in n+1 dimensions. Initialize the hyperparameter η; A Gaussian approximation is established on the posterior probability based on the likelihood estimation probability, resulting in an approximation of the marginal likelihood. The marginal likelihood function of the approximation is maximized, leading to a re-estimation of the hyperparameter vector values. This process is repeated until the hyperparameter vector values ​​converge. The data to be predicted is predicted using the classification function model of RVM to obtain the prediction results.

7. An apparatus for implementing the method for evaluating the erosion of circuit breaker contacts based on multi-dimensional information fusion according to any one of claims 1-6, characterized in that, It includes, The shock-absorbing base is supported on the ground; A hydraulic spring operating mechanism, which is supported on the shock-absorbing base; The opening and closing coils are mounted on the hydraulic spring operating mechanism; Circuit breaker, which is connected to the opening and closing coil; A simulated load is applied to the circuit breaker in a controllable manner; The sensor interface module includes, A voltage sensor, connected to the opening and closing coils, measures the circuit break signal. An angular displacement sensor is mounted on the main shaft of the circuit breaker to measure angular displacement signals. A vibration sensor is installed at the opening and closing coil to measure vibration signals. A current sensor is connected to the opening and closing coil to measure the current signal of the opening and closing coil. A loop dynamic resistance measurement unit is connected to the opening and closing coils and the circuit breaker to generate a loop dynamic resistance signal. An embedded acquisition module is connected to the sensor interface module to acquire and preprocess vibration signals, angular displacement signals, opening and closing coil current signals, circuit break signals, and circuit dynamic resistance signals. The embedded acquisition module includes an analog-to-digital converter, a digital isolator, a microcontroller for sampling control, a communication unit for signal transmission and storage, and a storage unit. A cloud computing module, which is communicatively connected to the communication unit, includes... The feature extraction and feature value labeling unit extracts and labels features based on the preprocessed information from each dimension. An evaluation algorithm deployment platform is connected to the feature extraction and feature value calibration unit to evaluate the ablation amount of the contact using a feature fusion algorithm.

Images