CN-116304975-B - Multi-feature fusion-based arc model judging method, device and medium

CN116304975BCN 116304975 BCN116304975 BCN 116304975BCN-116304975-B

Abstract

The application discloses an arc model judging method, device and medium based on multi-feature fusion, relates to the field of fault arc detection, and aims to solve the problem that arc judgment is easy to misjudge, and a parameter area extracts dimensional features and dimensionless features under different frequencies based on an arc current signal; and respectively carrying out feature fusion on the dimensionality features and the dimensionality features under different frequencies, inputting the feature fusion into a convolutional neural network for training, carrying out feature fusion on the fused dimensionality features and dimensionality features, inputting the feature fusion into the convolutional neural network for training, and obtaining a fusion result to judge whether an arc exists. The method combines the high-frequency characteristics (dimensionless characteristics and dimensionless characteristics) of different frequencies, performs characteristic fusion in the convolutional neural network layer training process, analyzes and acquires different characteristic characteristics from different angles, reasonably improves the diversity and comprehensiveness of the characteristics, greatly reduces the influence of low-frequency interference in the environment on discrimination, and can obviously improve the misjudgement rate.

Inventors

WANG ZHICHAO
WANG JIANHUA
ZHAO LIBO
MA YUE
WANG HUARONG
Wei shengteng

Assignees

青岛鼎信通讯股份有限公司
青岛鼎信通讯科技有限公司

Dates

Publication Date: 20260508
Application Date: 20230222

Claims (8)

1. The arc model distinguishing method based on multi-feature fusion is characterized by comprising the following steps of: Collecting an arc current signal; Based on the arc current signals, extracting dimensional features and dimensionless features under different frequencies; based on the arc current signal, the method for extracting the dimensional characteristics and the dimensionless characteristics under different frequencies comprises the following steps: processing the arc current signal to obtain characteristic signals under different frequencies; extracting the characteristics of the characteristic signals under different frequencies to obtain dimensional characteristics; extracting the feature of the dimensional features under different frequencies to obtain dimensionless features; The feature extraction of the dimensional features under different frequencies to obtain dimensionless features comprises the following steps: According to a first formula and the dimensional characteristics under different frequencies, obtaining dimensionless characteristics; wherein, the first formula is: ; T mean is the average value of frequency points of a single high-frequency channel in a half wave, T max is the maximum value in the single high-frequency channel, T min is the minimum value in the single high-frequency channel, M is the number of frequency point values in the half wave, which are higher than the sum average value of the maximum value and the minimum value of the half wave, N is the total number of frequency points of the single high-frequency channel in the half wave, fea is one of the dimensionless characteristics, i is the number of the high-frequency channels, and T i is the frequency point value, which is higher than the sum average value of the maximum value and the minimum value of the half wave, in the half wave; feature fusion is carried out on the dimensional features under different frequencies, and then the feature fusion is input into a convolutional neural network for training; feature fusion is carried out on the dimensionless features under different frequencies, and then the feature fusion is input to the convolutional neural network for training; feature fusion is carried out on the fused dimensional features and the dimensionless features, and the feature fusion is input to the convolutional neural network for training, so that a fusion result is obtained; And judging whether an electric arc exists or not according to the fusion result.
2. The method for determining an arc model based on multi-feature fusion according to claim 1, wherein the acquiring an arc current signal comprises: the method comprises the steps of obtaining an arc current signal through a sampling circuit connected with a live wire, wherein the sampling circuit comprises a circuit snap coil and a capacitor, and the circuit snap coil is connected with the capacitor in parallel.
3. The method for discriminating an arc model based on multi-feature fusion according to claim 2, wherein said processing the arc current signal to obtain the feature signal at different frequencies includes: Converting the electric arc current signal into a digital signal through an ADC module; transmitting the conversion result to a hardware digital signal processing unit for noise reduction and gain processing; and sending the hardware processing result to a software signal processing unit for Fourier transformation and median filtering processing to obtain characteristic signals under different frequencies.
4. The method for discriminating an arc model based on multi-feature fusion according to claim 3 wherein said feature extraction of said feature signals at different frequencies to obtain dimensional features comprises: And respectively inputting the characteristic signals under different frequencies into a high-frequency channel, and carrying out characteristic extraction on the characteristic signals under different frequencies by taking a half wave as a unit to obtain dimensional characteristics.
5. The method for discriminating between electric arc models based on multi-feature fusion according to claim 4 wherein said dimensionless features further include a coefficient of variation, a maximum value/mean value, and a minimum value/mean value.
6. An arc model discriminating apparatus based on multi-feature fusion for performing the arc model discriminating method according to any one of claims 1 to 5, characterized in that the arc model discriminating apparatus includes: The acquisition module is used for acquiring arc current signals; The extraction module is used for extracting dimensional characteristics and dimensionless characteristics under different frequencies based on the arc current signals; The first fusion module is used for carrying out feature fusion on the dimensional features under different frequencies and inputting the feature fusion to the convolutional neural network for training; the second fusion module is used for carrying out feature fusion on the dimensionless features under different frequencies and inputting the feature fusion to the convolutional neural network for training; the third fusion module is used for carrying out feature fusion on the fused dimensional features and the dimensionless features, and inputting the feature fusion to a convolutional neural network for training to obtain a fusion result; and the judging module is used for judging whether the electric arc exists or not according to the fusion result.
7. An arc model discriminating apparatus based on multi-feature fusion for the arc model discriminating method according to any one of claims 1 to 5, characterized in that the arc model discriminating apparatus includes: A memory for storing a computer program; A processor for implementing the steps of the arc model discrimination method based on multi-feature fusion according to any one of claims 1 to 5 when executing the computer program.
8. A computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, which when executed by a processor, implements the steps of the arc model discriminating method based on multi-feature fusion according to any one of claims 1 to 5.

Description

Multi-feature fusion-based arc model judging method, device and medium Technical Field The application relates to the field of fault arc detection, in particular to an arc model judging method, device and medium based on multi-feature fusion. Background With the rapid development of society, the type of electric equipment is endless, and the power consumption is increased day by day, once the fault arc appears, easily causes electric fire accident, therefore can ensure that electric equipment, user's personal safety and electric wire netting safe and stable operation are vital. Arc data and normal data are difficult to directly distinguish through time domain and low frequency band under complex environment, so that great challenges are brought to the traditional circuit protection device, the difficulty of finding arc faults in a circuit is greatly increased, potential safety hazards cannot be well eliminated, and fire disasters are further caused. The method is limited by traditional arc detection equipment and electronic technology, the sampling rate of the selected arc signal is low, the characteristics of the directly applied time domain are easily influenced by the actual environment to change, the stability of the arc characteristics is to be improved, a plurality of models are mainly analyzed aiming at the low-frequency characteristics of the arc although the frequency models are used, the characteristics of the arc change of different loads are difficult to accurately describe by the low-frequency characteristics, and misjudgment and even frequent false tripping are easy to occur in the application process. Therefore, how to solve the problem that the arc discrimination is easy to misjudge is a technical problem to be solved urgently by the person skilled in the art. Disclosure of Invention The application aims to provide an arc model judging method, device and medium based on multi-feature fusion. In order to solve the technical problems, the application provides an arc model judging method based on multi-feature fusion, which comprises the following steps: Collecting an arc current signal; Based on the arc current signals, extracting dimensional features and dimensionless features under different frequencies; feature fusion is carried out on the dimensional features under different frequencies, and then the feature fusion is input into a convolutional neural network for training; feature fusion is carried out on the dimensionless features under different frequencies, and then the feature fusion is input into a convolutional neural network for training; Feature fusion is carried out on the dimensionless features and the dimensionless features after fusion, and the feature fusion is input into a convolutional neural network for training, so that a fusion result is obtained; And judging whether an electric arc exists or not according to the fusion result. As a preferred solution, in the above arc model determining method based on multi-feature fusion, the extracting the dimensional features and the dimensionless features under different frequencies based on the arc current signal includes: processing the arc current signal to obtain characteristic signals under different frequencies; extracting the characteristics of the characteristic signals under different frequencies to obtain dimensional characteristics; and extracting the feature of the dimensional features under different frequencies to obtain dimensionless features. As a preferred solution, in the above arc model discriminating method based on multi-feature fusion, the collecting the arc current signal includes: the method comprises the steps of obtaining an arc current signal through a sampling circuit connected with a live wire, wherein the sampling circuit comprises a circuit snap coil and a capacitor, and the circuit snap coil is connected with the capacitor in parallel. As a preferred solution, in the above method for determining an arc model based on multi-feature fusion, the processing the arc current signal to obtain the feature signals under different frequencies includes: Converting the electric arc current signal into a digital signal through an ADC module; transmitting the conversion result to a hardware digital signal processing unit for noise reduction and gain processing; and sending the hardware processing result to a software signal processing unit for Fourier transformation and median filtering processing to obtain characteristic signals under different frequencies. As a preferred solution, in the above method for determining an arc model based on multi-feature fusion, the feature extraction is performed on the feature signals under different frequencies to obtain dimensional features, including: And respectively inputting the characteristic signals under different frequencies into a high-frequency channel, and carrying out characteristic extraction on the characteristic signals under different frequencies by taking a half