An arc model discrimination method and device based on multi-feature fusion and a medium

Abstract

The application discloses a multi-feature fusion-based arc model discrimination method and device and a medium, relates to the field of fault arc detection, and aims to solve the problem of misjudgment in arc discrimination. The parameter region is based on an arc current signal, and dimensioned features and dimensionless features at different frequencies are extracted. The dimensioned features and the dimensionless features at different frequencies are respectively input into a convolutional neural network for training after feature fusion. The dimensioned features and the dimensionless features after fusion are input into the convolutional neural network for training after feature fusion, and a fusion result is obtained to determine whether an arc exists. The high-frequency features (dimensionless features and dimensioned features) at different frequencies are combined, feature fusion is performed in the convolutional neural network layer training process, different feature characteristics are analyzed and obtained from different angles, the diversity and comprehensiveness of the features are reasonably improved, the influence of low-frequency interference in the environment on discrimination is greatly reduced, and the misjudgment rate can be obviously improved.

Description

Technical Field

[0001] This application relates to the field of fault arc detection, and in particular to an arc model discrimination method, device and medium based on multi-feature fusion. Background Technology

[0002] With rapid societal development, the types of electrical equipment are constantly evolving, and electricity consumption is increasing daily. Once an arc fault occurs, it can easily trigger an electrical fire. Therefore, ensuring the safety of electrical equipment, users, and the safe and stable operation of the power grid is crucial. Arc data is difficult to distinguish directly from normal data in complex environments using only time-domain and low-frequency methods, posing a significant challenge to traditional circuit protection devices. This greatly increases the difficulty of detecting arc faults in circuits, hindering the effective elimination of safety hazards and potentially leading to fires.

[0003] Limited by traditional arc detection equipment and electronic technology, the selected arc signal sampling rate is low. Direct application of time-domain characteristics is easily affected by the actual environment, and the stability of arc characteristics needs to be improved. Although many models use frequency models, they are mainly for analyzing the low-frequency characteristics of arcs. However, low-frequency characteristics are often difficult to accurately describe the characteristics of arc changes under different loads. In the application process, misjudgment or even frequent false tripping can easily occur.

[0004] Therefore, how to solve the problem of misjudgment in arc detection is a technical problem that urgently needs to be solved by those in the field. Summary of the Invention

[0005] The purpose of this application is to provide a method, device, and medium for arc model discrimination based on multi-feature fusion.

[0006] To address the aforementioned technical problems, this application provides a method for arc model discrimination based on multi-feature fusion, comprising:

[0007] Acquire arc current signals;

[0008] Based on the arc current signal, dimensional and dimensionless features at different frequencies are extracted.

[0009] The dimensional features at different frequencies are fused and then input into a convolutional neural network for training.

[0010] The dimensionless features at different frequencies are fused and then input into a convolutional neural network for training.

[0011] The dimensional features and the dimensionless features are fused and then input into a convolutional neural network for training to obtain the fusion result.

[0012] The presence of an electric arc is determined based on the fusion result.

[0013] As a preferred embodiment, in the above-mentioned arc model discrimination method based on multi-feature fusion, the step of extracting dimensional and dimensionless features at different frequencies based on the arc current signal includes:

[0014] The arc current signal is processed to obtain characteristic signals at different frequencies;

[0015] Feature extraction is performed on the characteristic signals at different frequencies to obtain dimensional features;

[0016] Dimensional features are extracted from the dimensional features at different frequencies to obtain dimensionless features.

[0017] As a preferred embodiment, in the above-mentioned arc model discrimination method based on multi-feature fusion, the acquisition of the arc current signal includes:

[0018] The arc current signal is obtained by a sampling circuit connected to the live wire; wherein the sampling circuit includes a circuit instantaneous coil and a capacitor, and the circuit instantaneous coil and the capacitor are connected in parallel.

[0019] As a preferred embodiment, in the above-mentioned arc model discrimination method based on multi-feature fusion, the step of processing the arc current signal to obtain feature signals at different frequencies includes:

[0020] The arc current signal is converted into a digital signal using an ADC module.

[0021] The conversion results are sent to the hardware digital signal processing unit for noise reduction and gain processing.

[0022] The hardware processing results are sent to the software signal processing unit for Fourier transform and median filtering to obtain characteristic signals at different frequencies.

[0023] As a preferred embodiment, in the above-mentioned arc model discrimination method based on multi-feature fusion, the step of extracting features from the feature signals at different frequencies to obtain dimensional features includes:

[0024] The characteristic signals at different frequencies are input into the high-frequency channel, and feature extraction is performed on the characteristic signals at different frequencies in half-wave units to obtain dimensional features.

[0025] As a preferred embodiment, in the above-mentioned arc model discrimination method based on multi-feature fusion, the step of extracting dimensionless features from the dimensional features at different frequencies includes:

[0026] Based on the first formula and the dimensional characteristics at different frequencies, the dimensionless characteristics are obtained.

[0027] The first formula is:

[0028]

[0029] T mean T is the average frequency of a single high-frequency channel within half-wave. max T is the maximum value within a single high-frequency channel. min Where is the minimum value within a single high-frequency channel, M is the number of frequency points within a half-wave that are higher than the average of the sum of the maximum and minimum values ​​within the half-wave, N is the total number of frequency points within a single high-frequency channel within the half-wave, Fea is one of the dimensionless features, i is the number of high-frequency channels, and T i It is the frequency point value that is higher than the sum of the maximum and minimum values ​​within half a wave and then takes the average.

[0030] As a preferred embodiment, in the above-mentioned arc model discrimination method based on multi-feature fusion, the dimensionless features further include: coefficient of variation, maximum / mean value, and minimum / mean value.

[0031] To address the aforementioned technical problems, this application also provides an arc model discrimination device based on multi-feature fusion, comprising:

[0032] The acquisition module is used to acquire arc current signals;

[0033] The extraction module is used to extract dimensional and dimensionless features at different frequencies based on the arc current signal.

[0034] The first fusion module is used to fuse the dimensional features at different frequencies and then input them into the convolutional neural network for training.

[0035] The second fusion module is used to fuse the dimensionless features at different frequencies and then input them into the convolutional neural network for training.

[0036] The third fusion module is used to fuse the dimensional features and the dimensionless features after fusion and input them into the convolutional neural network for training to obtain the fusion result.

[0037] The judgment module is used to determine whether an electric arc exists based on the fusion result.

[0038] To address the aforementioned technical problems, this application also provides an arc model discrimination device based on multi-feature fusion, comprising:

[0039] Memory, used to store computer programs;

[0040] A processor is used to implement the steps of the above-described arc model discrimination method based on multi-feature fusion when executing the computer program.

[0041] To address the aforementioned technical problems, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described arc model discrimination method based on multi-feature fusion.

[0042] The arc model discrimination method based on multi-feature fusion provided in this application includes: acquiring arc current signals; extracting dimensional and dimensionless features at different frequencies based on the arc current signals; fusing the dimensional features at different frequencies and inputting them into a convolutional neural network for training; fusing the dimensionless features at different frequencies and inputting them into a convolutional neural network for training; fusing the fused dimensional and dimensionless features and inputting them into the convolutional neural network for training to obtain the fusion result; and determining the existence of an arc based on the fusion result. By combining high-frequency features (dimensional and dimensionless features) of different frequencies and fusing them during the training process of the convolutional neural network layer, different feature characteristics are analyzed and obtained from different perspectives, reasonably improving the diversity and comprehensiveness of features, greatly reducing the impact of low-frequency interference in the environment on discrimination, and significantly improving the misclassification rate.

[0043] In addition, this application also provides an apparatus and a medium that correspond to the above method and have the same effect. Attached Figure Description

[0044] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 A flowchart illustrating an arc model discrimination method based on multi-feature fusion provided in this application embodiment;

[0046] Figure 2 A schematic diagram illustrating a scenario for an arc model discrimination method based on multi-feature fusion, provided in an embodiment of this application;

[0047] Figure 3 A structural diagram of an arc model discrimination device based on multi-feature fusion provided in an embodiment of this application;

[0048] Figure 4 This is a structural diagram of another arc model discrimination device based on multi-feature fusion provided in an embodiment of this application. Detailed Implementation

[0049] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0050] The core of this application is to provide a method, device, and medium for arc model discrimination based on multi-feature fusion.

[0051] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0052] In recent years, fires caused by electrical faults have remained high, causing enormous losses to the lives and property of the nation and its people. The primary cause of electrical fires is fault arcing in low-voltage power distribution lines. Fault arcing is characterized by its concealment, randomness, and complexity, posing a serious electrical fire hazard to low-voltage power distribution lines. Aging insulation or poor contact in the lines easily leads to series fault arcing, which is very common in power distribution systems. The current amplitude is not easily distinguishable from parallel fault arcing, increasing the detection difficulty of traditional protection devices and representing one of the loopholes in the existing electrical protection system. Therefore, research on methods for detecting series fault arcing is of great significance.

[0053] Methods for detecting series-connected fault arcs mainly focus on three aspects: the mathematical model of the fault arc, its physical characteristics, and the changes in current / voltage waveforms. The mathematical model of the arc primarily describes its dynamic changes through mathematical equations. Given the randomness and complexity of the arc, the influence of various environmental parameters needs to be considered. However, the dynamic nature of these environmental parameters leads to different generated arc models. Utilizing the mathematical model of the arc for fault arc detection is mainly concentrated in the theoretical research and experimental simulation stages. Using physical characteristics for arc fault detection requires sensors with fixed positions, which contradicts the randomness of the fault arc. The fault location can easily exceed the sensor's detection range, thus affecting the detection results.

[0054] Methods based on current / voltage waveform changes mainly focus on three aspects: time domain, frequency domain, and time-frequency domain features, which are also the main research directions for arc fault detection. Common time-domain feature extraction methods for arc current include integral values ​​and coefficients of variation, while frequency-domain feature extraction is mainly achieved through Fourier transform and wavelet transform. For these extracted features, two approaches are taken: first, comparing them with preset thresholds to determine if an arc fault has occurred; second, establishing an arc fault identification model, extracting feature vectors again through signal analysis and processing methods, and inputting them into classifiers such as support vector machines and neural networks to complete classification and obtain a judgment result. However, this method has high computational complexity and high hardware performance requirements. Furthermore, the diversity and representativeness of the features applied by many methods need improvement, making it difficult to guarantee the representativeness of the pre-extracted series arc fault features.

[0055] Limited by traditional arc detection equipment and electronic technology, the selected arc signal sampling rate is low, and direct application of time-domain features is greatly affected by time variations, resulting in insufficient representativeness of arc features. Furthermore, the low-frequency features extracted based on this are also difficult to accurately describe the characteristics of arc changes. In application, they are often affected by complex environments, easily leading to misjudgments or even frequent false tripping. To address the problem of arc discrimination easily causing misjudgments in real-world environments, this application provides an arc model discrimination method based on multi-feature fusion, such as... Figure 1 As shown, it includes:

[0056] S11: Acquire arc current signal;

[0057] Arc current signals are acquired through a sampling circuit. This embodiment does not limit the specific structure of the sampling circuit or the sampling frequency; it can be set according to actual needs.

[0058] S12: Based on the arc current signal, extract dimensional and dimensionless features at different frequencies;

[0059] Dimensional eigenvalues ​​often have intuitive physical meanings and are the most commonly used characteristic indicators. Dimensional eigenvalues ​​mainly include: maximum value, minimum value, peak-to-peak value, mean, variance, standard deviation, root mean square value, root mean square value / effective value (RMS), mean square error (MSE), root mean square error (RMSE), and root mean square magnitude.

[0060] While dimensional metrics are sensitive to signal characteristics, they also vary with operating conditions (such as load) and are highly susceptible to environmental interference, resulting in unstable performance. In contrast, dimensionless metrics can eliminate the influence of these disturbances. Dimensionless metrics mainly include peak factor (the ratio of signal peak value to RMS value), impulse factor (the ratio of signal peak value to rectified average value (the average of absolute values)), margin factor (the ratio of signal peak value to RMS amplitude), kurtosis factor, waveform factor (the ratio of RMS value to rectified average value), and skewness factor, etc.

[0061] This embodiment extracts dimensional and dimensionless features of arc current signals at different frequencies. There are no specific restrictions on which features to extract; they can be set according to specific needs.

[0062] S13: After fusing the dimensional features at different frequencies, the data are input into a convolutional neural network for training.

[0063] S14: The dimensionless features at different frequencies are fused and then input into a convolutional neural network for training;

[0064] S15: The dimensional features and the dimensionless features are fused and then input into a convolutional neural network for training to obtain the fusion result;

[0065] Feature fusion refers to processing multiple feature information to obtain a fused feature that references all the feature information. There are various ways to perform feature fusion, such as adding, multiplying, or weighting the individual feature information.

[0066] Convolutional Neural Networks (CNNs) are a typical deep learning model. For signal or image data, they can learn feature representation mechanisms at different levels of abstraction, and are trainable, multi-level feedforward artificial neural networks. A CNN consists of layers arranged sequentially, with each layer using a differentiable function to pass activation data from one layer to another. CNNs primarily consist of three types of layers: convolutional layers, pooling layers, and fully connected layers (fully connected layers are the same as in conventional neural networks). By stacking these layers, a complete CNN can be constructed.

[0067] In this embodiment, a convolutional neural network is used to train and fuse dimensional features and dimensionless features at different frequencies. The fused dimensional features and dimensionless features are then fused together and input into the convolutional neural network for training to obtain the fusion result.

[0068] S16: Determine whether an electric arc exists based on the fusion result.

[0069] The presence of an electric arc is determined based on the fusion result. This embodiment does not limit the specific judgment conditions and can be set according to the corresponding environment.

[0070] The arc model discrimination method based on multi-feature fusion provided in this embodiment includes: acquiring arc current signals; extracting dimensional and dimensionless features at different frequencies based on the arc current signals; fusing the dimensional features at different frequencies and inputting them into a convolutional neural network for training; fusing the dimensionless features at different frequencies and inputting them into a convolutional neural network for training; fusing the fused dimensional and dimensionless features and inputting them into the convolutional neural network for training to obtain the fusion result; and determining whether an arc exists based on the fusion result. By combining high-frequency features (dimensional and dimensionless features) at different frequencies and fusing them during the training process of the convolutional neural network layer, different feature characteristics are analyzed and obtained from different perspectives, reasonably improving the diversity and comprehensiveness of features, greatly reducing the impact of low-frequency interference in the environment on discrimination, and significantly improving the misclassification rate.

[0071] Based on the above embodiments, this embodiment provides a preferred solution: an arc model discrimination method based on multi-feature fusion, wherein the step of extracting dimensional and dimensionless features at different frequencies based on the arc current signal includes:

[0072] The arc current signal is processed to obtain characteristic signals at different frequencies;

[0073] Feature extraction is performed on the characteristic signals at different frequencies to obtain dimensional features;

[0074] Dimensional features are extracted from the dimensional features at different frequencies to obtain dimensionless features.

[0075] In this embodiment, processing the arc current signal refers to converting the analog arc current signal into a digital signal. Furthermore, to improve the accuracy of feature extraction, signal processing such as noise reduction and gain adjustment can be applied to the analog signal. Preferably, processing the arc current signal to obtain feature signals at different frequencies includes:

[0076] The arc current signal is converted into a digital signal using an analog-to-digital converter (ADC) module.

[0077] The conversion result is sent to the hardware digital signal processing unit for noise reduction and gain processing.

[0078] The hardware processing results are sent to the software signal processing unit for Fourier transform and median filtering to obtain characteristic signals at different frequencies.

[0079] The analog signal of the arc current signal is converted into a digital signal by an analog-to-digital converter, and the digital signal is subjected to noise reduction and gain processing to reduce the noise floor, further improve the signal stability and enhance the signal quality.

[0080] In this embodiment, there can be various methods for noise reduction and gain processing, and no specific restrictions are placed on the methods of noise reduction and gain processing.

[0081] Furthermore, the processing results are subjected to Fourier transform, converting the time-domain signal into a frequency-domain signal that is easier to analyze. High-frequency signals at different frequencies are extracted, and noise signals are further filtered out through median filtering. Preferably, feature extraction is performed based on 2-50MHz.

[0082] After cyclic processing of the arc current signal, feature extraction is performed on the characteristic signal at different frequencies to obtain dimensional features. Further feature extraction is performed on the dimensional features at different frequencies to obtain dimensionless features.

[0083] This embodiment provides a preferred solution, wherein the step of extracting features from the feature signals at different frequencies to obtain dimensional features includes:

[0084] The characteristic signals at different frequencies are input into the high-frequency channel, and feature extraction is performed on the characteristic signals at different frequencies in half-wave units to obtain dimensional features.

[0085] In this embodiment, the feature signals at different frequencies are extracted in half-wave units to obtain dimensional features. For example, a half-wave unit of 10ms is selected, and data at different frequencies within one half-wave period are extracted by using the sampling frequency and the number of sampling points within one period.

[0086] Furthermore, this embodiment provides a preferred solution, wherein the step of extracting dimensionless features from the dimensional features at different frequencies includes:

[0087] Based on the first formula and the dimensional characteristics at different frequencies, the dimensionless characteristics are obtained.

[0088] The first formula is:

[0089]

[0090] T mean T is the average frequency of a single high-frequency channel within half-wave. max T is the maximum value within a single high-frequency channel.min Where is the minimum value within a single high-frequency channel, M is the number of frequency points within a half-wave that are higher than the average of the sum of the maximum and minimum values ​​within the half-wave, N is the total number of frequency points within a single high-frequency channel within the half-wave, Fea is one of the dimensionless features, i is the number of high-frequency channels, and T i It is the frequency point value that is higher than the sum of the maximum and minimum values ​​within half a wave and then takes the average.

[0091] This embodiment proposes and constructs a new arc feature, Fea, in the above formula, based on the characteristics of AC arc signals, using half-wave as the unit. This feature can effectively describe the characteristics of current frequency domain signals with significantly discrete data, and serves as one of the dimensionless features after the second feature processing. The high-frequency channels here correspond to the number of feature signals at different frequencies mentioned above.

[0092] Preferably, the dimensionless features further include: coefficient of variation, maximum / mean value, and minimum / mean value.

[0093] Among them, the coefficient of variation is the half-wave coefficient of variation, the maximum value / mean is the half-wave maximum value / half-wave mean, and the minimum value / mean is the half-wave minimum value / half-wave mean.

[0094] The feature Fea is then fused with features such as half-wave variation coefficient, half-wave maximum / half-wave mean, and half-wave minimum / half-wave mean. The dimensionless quadratic feature after fusion has a dimension size of (P, Q), where P is the number of selected channels (i.e., high-frequency channels) and Q is the number of selected features. The dimensionless first-order feature has a dimension size of (P, O), where O is the number of frequency points in each half-wave.

[0095] After completing the feature classification described above, the dimensional features are normalized, resulting in a feature Fea range of 0 to 1 (similar to the dimensionless features), which facilitates the generalization of the statistical distribution of samples. Dimensionless features, however, do not require normalization; Fea > 0, preserving their inherent properties. The two types of processed features are then fed into a convolutional neural network, processed through different input layers, convolutional layers, and pooling layers. The dimension of the dimensionless features becomes (P1, Q1), and the dimension of the dimensional features becomes (P1, Q2). The two types of features are then fused before the fully connected layer, resulting in a fused feature dimension of (P1 + P2, Q2). The predicted probability (weight vector and bias vector) Prei is then calculated through the fully connected layer.

[0096]

[0097] At this point, it can be determined whether an electric arc occurs in the half-wave (an arc is determined when the probability value exceeds 0.5 and marked as 1; otherwise, it is considered normal and marked as 0). It should be noted that this is only an illustrative judgment data, and other values ​​can be set according to actual needs.

[0098] Then, the discrimination result of each half-wave is sent to the classification decision unit model in real time. The model classifies whether arcing occurs at this time by using the set half-wave strategy, and outputs whether arcing occurs or normal, so as to facilitate the circuit control of whether the trip unit should take action.

[0099] Dimensional feature normalization can eliminate the attribute relationship between fixed features to a certain extent. Fault features are extracted based on different scales. Dimensional features and purely dimensionless features are respectively substituted into the convolutional neural network for processing. The two are fused in the middle layer of the convolutional neural network, which gives different perspectives on feature analysis and increases feature diversity.

[0100] To better acquire arc current signals, this embodiment provides a preferred solution, wherein acquiring the arc current signal includes:

[0101] The arc current signal is obtained by a sampling circuit connected to the live wire; wherein the sampling circuit includes a circuit instantaneous coil and a capacitor, and the circuit instantaneous coil and the capacitor are connected in parallel.

[0102] This embodiment uses a transient coil for signal sampling. Connecting the capacitor in parallel with the coil ensures the stability of the coil sampling. At the same time, the capacitor can filter the signal, reduce the influence of noise, and facilitate the extraction of the main features of the arc signal.

[0103] To enable those skilled in the art to better understand this solution, this embodiment provides a specific scenario implementation plan. Figure 2 This is a schematic diagram of a scenario provided in this embodiment, such as... Figure 2 As shown.

[0104] The first capacitor C1 is connected in parallel with the instantaneous coil L of the circuit. High-frequency sampling is performed directly at the live wire position. The hardware digital signal processing unit performs AD sampling, ADC conversion, noise reduction, and gain processing. The software signal processing unit performs Fourier transform and median filtering. In this embodiment, two feature signals at different frequencies are extracted from the arc current signal and input to high-frequency channel 1 and high-frequency channel 2 respectively. After one feature processing step, dimensional features are obtained. Then, after a second feature processing step, multiple dimensionless features (i.e.,...) are obtained. Figure 2 The dimensional and dimensionless features of different channels are fused and input into the convolutional neural network for training. The two types of features are trained through different network layers, and finally feature fusion and model training are performed in the same network layer to obtain the discrimination result of each half wave. The discrimination result of each half wave is sent to the classification decision unit model in real time. The model classifies whether arcing occurs at this time according to the set half wave strategy, and outputs whether arcing or normal, so as to facilitate the circuit control of whether the trip unit should act.

[0105] In the above embodiments, the arc model discrimination method based on multi-feature fusion has been described in detail. This application also provides embodiments of an arc model discrimination device based on multi-feature fusion. It should be noted that this application describes the embodiments of the device from two perspectives: one is based on functional modules, and the other is based on hardware.

[0106] From the perspective of functional modules Figure 3 A structural diagram of an arc model discrimination device based on multi-feature fusion provided in this application embodiment is shown below. Figure 3 As shown, the arc model discrimination device based on multi-feature fusion includes:

[0107] Acquisition module 31 is used to acquire arc current signals;

[0108] Extraction module 32 is used to extract dimensional and dimensionless features at different frequencies based on the arc current signal;

[0109] The first fusion module 33 is used to fuse dimensional features at different frequencies and then input them into the convolutional neural network for training.

[0110] The second fusion module 34 is used to fuse the dimensionless features at different frequencies and then input them into the convolutional neural network for training.

[0111] The third fusion module 35 is used to fuse the fused dimensional features with the dimensionless features and input them into the training of the convolutional neural network to obtain the fusion result.

[0112] The judgment module 36 is used to determine whether an electric arc exists based on the fusion result.

[0113] The arc model discrimination device based on multi-feature fusion provided in this embodiment includes: an acquisition module 31 that acquires arc current signals; an extraction module 32 that extracts dimensional and dimensionless features at different frequencies based on the arc current signals; a first fusion module 33 that fuses the dimensional features at different frequencies and inputs them into a convolutional neural network for training; a second fusion module 34 that fuses the dimensionless features at different frequencies and inputs them into the convolutional neural network for training; a third fusion module 35 that fuses the fused dimensional and dimensionless features and inputs them into the convolutional neural network for training to obtain a fusion result; and a judgment module 36 that determines whether an arc exists based on the fusion result. By combining high-frequency features (dimensional and dimensionless features) at different frequencies and performing feature fusion during the training of the convolutional neural network layers, different feature characteristics are analyzed and obtained from different perspectives, reasonably improving the diversity and comprehensiveness of features, greatly reducing the impact of low-frequency interference in the environment on discrimination, and significantly improving the misjudgment rate.

[0114] In addition, the device also includes:

[0115] The signal processing subunit is used to process the arc current signal to obtain characteristic signals at different frequencies;

[0116] The first extraction subunit is used to extract features from the feature signals at different frequencies to obtain dimensional features.

[0117] The second extraction subunit is used to extract the dimensional features at different frequencies to obtain dimensionless features.

[0118] The conversion subunit is used to convert the arc current signal into a digital signal through the ADC module;

[0119] The first processing subunit is used to send the conversion result to the hardware digital signal processing unit for noise reduction and gain processing.

[0120] The second processing subunit is used to send the hardware processing results to the software signal processing unit for Fourier transform and median filtering to obtain characteristic signals at different frequencies.

[0121] Since the embodiments of the apparatus and the embodiments of the method correspond to each other, please refer to the description of the embodiments of the method for the embodiments of the apparatus, which will not be repeated here.

[0122] Figure 4 A structural diagram of another arc model discrimination device based on multi-feature fusion provided in this application embodiment is shown below. Figure 4 As shown, the arc model discrimination device based on multi-feature fusion includes: a memory 40 for storing computer programs;

[0123] The processor 41 is used to execute a computer program to implement the steps of the method for obtaining user operation habit information as described in the above embodiment (arc model discrimination method based on multi-feature fusion).

[0124] The arc model discrimination device based on multi-feature fusion provided in this embodiment can include, but is not limited to, smartphones, tablets, laptops, or desktop computers.

[0125] The processor 41 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 41 may be implemented using at least one hardware form selected from Digital Signal Processors (DSPs), Field-Programmable Gate Arrays (FPGAs), and Programmable Logic Arrays (PLAs). The processor 41 may also include a main processor and a coprocessor. The main processor, also known as a Central Processing Unit (CPU), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 41 may integrate a Graphics Processing Unit (GPU), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 41 may also include an Artificial Intelligence (AI) processor, which handles computational operations related to machine learning.

[0126] The memory 40 may include one or more computer-readable storage media, which may be non-transitory. The memory 40 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 40 is used to store at least the following computer program 401, which, after being loaded and executed by the processor 41, is capable of implementing the relevant steps of the arc model discrimination method based on multi-feature fusion disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 40 may also include an operating system 402 and data 403, and the storage method may be temporary storage or permanent storage. The operating system 402 may include Windows, Unix, Linux, etc. The data 403 may include, but is not limited to, the data involved in implementing the arc model discrimination method based on multi-feature fusion.

[0127] In some embodiments, the arc model discrimination device based on multi-feature fusion may further include a display screen 42, an input / output interface 43, a communication interface 44, a power supply 45, and a communication bus 46.

[0128] Those skilled in the art will understand that Figure 4 The structure shown does not constitute a limitation on the arc model discrimination device based on multi-feature fusion, and may include more or fewer components than shown.

[0129] The arc model discrimination device based on multi-feature fusion provided in this application includes a memory and a processor. When the processor executes the program stored in the memory, it can implement the following method: an arc model discrimination method based on multi-feature fusion, including: acquiring arc current signals; extracting dimensional and dimensionless features at different frequencies based on the arc current signals; fusing the dimensional features at different frequencies and inputting them into a convolutional neural network for training; fusing the dimensionless features at different frequencies and inputting them into a convolutional neural network for training; fusing the fused dimensional and dimensionless features and inputting them into the convolutional neural network for training to obtain a fusion result; and determining whether an arc exists based on the fusion result. By combining high-frequency features (dimensional and dimensionless features) of different frequencies and fusing them during the training process of the convolutional neural network layer, different feature characteristics are analyzed and obtained from different perspectives, reasonably improving the diversity and comprehensiveness of features, greatly reducing the impact of low-frequency interference in the environment on discrimination, and significantly improving the misclassification rate.

[0130] Finally, this application also provides an embodiment corresponding to a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps described in the above embodiment of the arc model discrimination method based on multi-feature fusion.

[0131] It is understood that if the methods in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0132] The computer-readable storage medium provided in this embodiment stores a computer program. When the processor executes the program, it can implement the following method: an arc model discrimination method based on multi-feature fusion, including: acquiring arc current signals; extracting dimensional and dimensionless features at different frequencies based on the arc current signals; fusing the dimensional features at different frequencies and inputting them into a convolutional neural network for training; fusing the dimensionless features at different frequencies and inputting them into a convolutional neural network for training; fusing the fused dimensional and dimensionless features and inputting them into the convolutional neural network for training to obtain a fusion result; and determining whether an arc exists based on the fusion result. By combining high-frequency features (dimensional and dimensionless features) at different frequencies and fusing them during the training of the convolutional neural network layer, different feature characteristics can be analyzed and obtained from different perspectives, reasonably improving the diversity and comprehensiveness of features, greatly reducing the impact of low-frequency interference in the environment on discrimination, and significantly improving the misclassification rate.

[0133] The foregoing has provided a detailed description of the arc model discrimination method, apparatus, and medium based on multi-feature fusion provided in this application. The various embodiments in the 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 it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

[0134] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A method for identifying electric arc models based on multi-feature fusion, characterized in that, include: Acquire arc current signals; Based on the arc current signal, dimensional and dimensionless features at different frequencies are extracted. The extraction of dimensional and dimensionless features at different frequencies based on the arc current signal includes: The arc current signal is processed to obtain characteristic signals at different frequencies; Feature extraction is performed on the characteristic signals at different frequencies to obtain dimensional features; Dimensionless features are obtained by extracting features from the dimensional features at different frequencies. The step of extracting dimensionless features from the dimensional features at different frequencies includes: Based on the first formula and the dimensional characteristics at different frequencies, the dimensionless characteristics are obtained. The first formula is: ； T mean T is the average frequency of a single high-frequency channel within half-wave. max T is the maximum value within a single high-frequency channel. min Where is the minimum value within a single high-frequency channel, M is the number of frequency points within a half-wave that are higher than the average of the sum of the maximum and minimum values ​​within the half-wave, N is the total number of frequency points within a single high-frequency channel within the half-wave, Fea is one of the dimensionless features, i is the number of high-frequency channels, and T i The frequency point value that is higher than the sum of the maximum and minimum values ​​within half a wave and then takes the average value; The dimensional features at different frequencies are fused and then input into a convolutional neural network for training. The dimensionless features at different frequencies are fused and then input into the convolutional neural network for training. The dimensional features and the dimensionless features are fused together and then input into the convolutional neural network for training to obtain the fusion result. The presence of an electric arc is determined based on the fusion result.

2. The arc model discrimination method based on multi-feature fusion according to claim 1, characterized in that, The acquisition of the arc current signal includes: The arc current signal is obtained by a sampling circuit connected to the live wire; wherein the sampling circuit includes a circuit instantaneous coil and a capacitor, and the circuit instantaneous coil and the capacitor are connected in parallel.

3. The arc model discrimination method based on multi-feature fusion according to claim 2, characterized in that, The process of processing the arc current signal to obtain characteristic signals at different frequencies includes: The arc current signal is converted into a digital signal using an ADC module. The conversion result is sent to the hardware digital signal processing unit for noise reduction and gain processing. The hardware processing results are sent to the software signal processing unit for Fourier transform and median filtering to obtain characteristic signals at different frequencies.

4. The arc model discrimination method based on multi-feature fusion according to claim 3, characterized in that, The step of extracting features from the feature signals at different frequencies to obtain dimensional features includes: The characteristic signals at different frequencies are input into the high-frequency channel, and feature extraction is performed on the characteristic signals at different frequencies in half-wave units to obtain dimensional features.

5. The arc model discrimination method based on multi-feature fusion according to claim 4, characterized in that, The dimensionless features also include: coefficient of variation, maximum / mean, minimum / mean.

6. An arc model discrimination device based on multi-feature fusion, used to execute the arc model discrimination method according to any one of claims 1 to 5, characterized in that, The arc model discrimination device includes: The acquisition module is used to acquire arc current signals; The extraction module is used to extract dimensional and dimensionless features at different frequencies based on the arc current signal. The first fusion module is used to fuse the dimensional features at different frequencies and then input them into the convolutional neural network for training. The second fusion module is used to fuse the dimensionless features at different frequencies and then input them into the convolutional neural network for training. The third fusion module is used to fuse the dimensional features and the dimensionless features after fusion and input them into the convolutional neural network for training to obtain the fusion result. The judgment module is used to determine whether an electric arc exists based on the fusion result.

7. An arc model discrimination device based on multi-feature fusion, used in the arc model discrimination method according to any one of claims 1 to 5, characterized in that, The arc model discrimination device includes: Memory, used to store computer programs; A processor, configured to implement the steps of the arc model discrimination method based on multi-feature fusion as described in any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the arc model discrimination method based on multi-feature fusion as described in any one of claims 1 to 5.

Images