Fault arc feature extraction method and device based on self-convolution, and electronic equipment

Abstract

The present invention discloses a method and device for extracting fault arc features based on self-convolution, and an electronic device. The method comprises: cutting the fault arc waveform to obtain first data, which contains fault arc information; processing the first data to obtain second data and using it as a dedicated convolution kernel for detecting this type of fault arc, wherein the processing process includes adjusting the mean; collecting and pre-processing the power grid signal obtained by the sensor as third data; performing convolution calculation on the second data and the third data to obtain fourth data, wherein the convolution calculation separates the fault arc signal from the normal power grid signal in amplitude, thereby achieving the function of extracting the fault arc features; performing feature recognition processing on the fourth data to obtain fifth data, wherein the fifth data contains information on whether a fault arc has occurred in the electrical equipment. The present invention solves the problems of poor feature extraction effect and excessive parameters caused by the universal use of wavelet transform convolution kernels in traditional fault arc feature extraction methods, such as conventional wavelet transform convolution kernels.

Description

Technical Field

[0001] The present application relates to the field of feature extraction methods, and in particular to a fault arc feature extraction method and device based on self-convolution, and electronic equipment. Background Art

[0002] In recent years, electrical fires have become an increasingly serious threat to life and property. Among the factors that cause electrical fires, the harm caused by arc faults cannot be ignored. Arc faults occur in a very short time, and accumulated arc faults can cause fires.

[0003] Traditional arc fault feature extraction methods, such as the universal wavelet transform convolution kernel, result in poor feature extraction and too many parameters, making arc fault detection a difficult problem.

[0004] The above problems exist in the related art, and no effective solutions have been proposed yet. Summary of the Invention

[0005] The purpose of the embodiments of the present application is to provide a fault arc feature extraction method and device based on self-convolution, and an electronic device to solve the problem that fault arcs are very harmful but the existing fault arc feature extraction methods have poor feature extraction effects, a large number of parameters, and occupy a lot of hardware resources.

[0006] According to a first aspect of an embodiment of the present application, a method for extracting arc fault features based on self-convolution is provided, comprising:

[0007] Cutting the fault arc waveform to obtain first data, wherein the first data contains fault arc information;

[0008] processing the first data to obtain second data;

[0009] The power grid signal collected by the sensor and pre-processed is used as the third data;

[0010] Performing a convolution calculation on the second data and the third data to obtain fourth data;

[0011] The fourth data is subjected to feature recognition processing to obtain fifth data, wherein the fifth data contains information on whether an arc fault occurs in the electrical equipment.

[0012] Furthermore, the arc fault waveform is selected from:

[0013] The parallel or series electrical signal waveform collected by the sensor when a fault arc occurs in a household appliance;

[0014] Physical signal waveform when a fault arc occurs in household appliances;

[0015] Arc fault signal waveforms open sourced on the Internet;

[0016] The arc fault information is selected from:

[0017] The area where the arc fault signal waveform suddenly changes;

[0018] The area that matches the arc fault signal waveform when household appliances are used as loads;

[0019] When there is no fault arc, the signal waveform is the slow-changing harmonics of a normal power grid. When there is no fault arc, the signal waveform is the slow-changing harmonics of a normal power grid with household appliances as loads.

[0020] Furthermore, shearing the fault arc waveform to obtain first data includes:

[0021] The complete fault arc waveform or the left half or the right half of the complete fault arc waveform is cut to obtain first data.

[0022] Furthermore, processing the first data to obtain second data includes:

[0023] If the clipped fault arc waveform is a complete fault arc waveform, adjusting the mean of the first data to be 0, thereby obtaining second data;

[0024] If the cut fault arc waveform is the left half or right half of the complete fault arc waveform, the processing includes center-symmetric replication and adjustment of the mean to obtain second data, wherein the center-symmetry is to replicate the left half or right half of the fault arc waveform as the right half or left half of the fault arc waveform and splice it with the left half or right half of the original fault arc waveform to construct a complete fault arc waveform, wherein the mean is adjusted so that the mean of the constructed complete fault arc waveform is 0.

[0025] Furthermore, the power grid signal collected and pre-processed by the sensor is used as the third data, including:

[0026] The sensor collects the power grid signal to obtain an analog electrical signal, and converts the analog electrical signal into a digital electrical signal through an analog-to-digital converter;

[0027] The digital electrical signal is subjected to amplitude adjustment and filtering to obtain third data.

[0028] Further, performing convolution calculation on the second data and the third data to obtain fourth data includes:

[0029] Real-time convolution calculation: The second data is stored in advance. When the third data is input, the second data is convolved with the third data to obtain the fourth data.

[0030] Non-real-time convolution calculation: the second data and the third data are stored in advance, and the second data and the third data are convolved to obtain the fourth data.

[0031] Furthermore, feature recognition processing is performed on the fourth data to obtain fifth data, including:

[0032] comparing the amplitude of the fourth data with a set threshold, counting the number of times the fourth data exceeds the set threshold, and outputting fifth data indicating that an arc fault has been detected if the counted number exceeds the limit;

[0033] Comparing the variation range of the fourth data with a threshold value, counting the number of times the variation range of the fourth data exceeds the set threshold value, and outputting fifth data indicating that a fault arc is detected if the counted number of times exceeds the limit value;

[0034] Comparing the variance of the fourth data with a threshold, counting the number of times the variance of the fourth data exceeds the set threshold, and outputting fifth data indicating that a fault arc is detected if the counted number exceeds the limit;

[0035] The graph converted from the fourth data is compared with a preset graph, and the similarity between the graph converted from the fourth data and the preset graph is calculated. If the similarity exceeds a limit, the fifth data outputted indicates that a fault arc is detected.

[0036] According to a second aspect of an embodiment of the present application, a device for extracting arc fault features based on self-convolution is provided, comprising:

[0037] a cutting module, configured to cut the arc fault waveform to obtain first data, wherein the first data contains arc fault information;

[0038] A first processing module, configured to process the first data to obtain second data;

[0039] A second processing module is used to use the power grid signal collected by the sensor and pre-processed as third data;

[0040] a convolution calculation module, configured to perform a convolution calculation on the second data and the third data to obtain fourth data;

[0041] The identification module is used to perform feature identification processing on the fourth data to obtain fifth data, wherein the fifth data contains information on whether a fault arc occurs in the electrical equipment.

[0042] According to a third aspect of the embodiments of the present application, there is provided an electronic device, including:

[0043] one or more processors;

[0044] a memory for storing one or more programs;

[0045] When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in the first aspect.

[0046] According to a fourth aspect of an embodiment of the present application, a computer-readable storage medium is provided, on which computer instructions are stored. When the instructions are executed by a processor, the steps of the method described in the first aspect are implemented.

[0047] The technical solutions provided by the embodiments of the present application may have the following beneficial effects:

[0048] It can be seen from the above embodiments that the present application is based on a self-convolution fault arc feature extraction method, device, and electronic device. Because the cut first data originates from the fault arc waveform, the first data contains the fault arc feature; because the first data containing the fault arc feature is processed to obtain the second data, the second data is a dedicated convolution kernel containing the fault arc feature of the first data; because the second data containing the fault arc feature is convolved with the third data to be detected, the problem of poor performance of traditional feature extraction algorithms in extracting fault arc features in the third data is overcome, and better feature extraction performance is achieved by using the actual fault arc as a dedicated convolution kernel.

[0049] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0051] Figure 1 The figure is a flow chart showing a method for extracting arc fault features based on self-convolution according to an exemplary embodiment.

[0052] Figure 2 FIG. 1 is a current waveform of a series AC grid with a resistor as a load and containing a fault arc according to an exemplary embodiment.

[0053] Figure 3 FIG. 1 shows a roughly located arc fault according to an exemplary embodiment.

[0054] Figure 4 is the first data in step S1 according to an exemplary embodiment;

[0055] Figure 5 , which is a result of the center-symmetrical replication processing of the first data in step S2 according to an exemplary embodiment.

[0056] Figure 6is the second data in step S2 according to an exemplary embodiment.

[0057] Figure 7 is the fourth data in step S3 according to an exemplary embodiment.

[0058] Figure 8 A fault arc feature extraction method based on self-convolution according to an exemplary embodiment is compared with a traditional feature extraction method based on discrete wavelet transform.

[0059] Figure 9 ] are absolute values ​​of feature extraction results of two algorithms according to an exemplary embodiment.

[0060] Figure 10 The absolute values ​​of the feature extraction results of the two algorithms shown in accordance with an exemplary embodiment are sorted from large to small.

[0061] Figure 11 FIG1 is a diagram illustrating an electric drill self-convolution kernel constructed according to an exemplary embodiment.

[0062] Figure 12 The figure shows a computer self-convolution kernel constructed according to an exemplary embodiment.

[0063] Figure 13 FIG. 4 is a fan self-convolution kernel constructed according to an exemplary embodiment.

[0064] Figure 14 FIG. 4 is a fluorescent lamp self-convolution kernel constructed according to an exemplary embodiment.

[0065] Figure 15 FIG1 is a self-convolution kernel for an electric kettle constructed according to an exemplary embodiment.

[0066] Figure 16 The present invention is a block diagram showing a device for extracting arc fault features based on self-convolution according to an exemplary embodiment. DETAILED DESCRIPTION

[0067] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0068] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. As used in this application and the appended claims, the singular forms "a," "an," "the," and "the" are intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0069] It should be understood that although the terms first, second, third, etc. may be used in this application to describe various information, such information should not be limited to these terms. These terms are merely used to distinguish information of the same type from one another. For example, first information may also be referred to as second information, and similarly, second information may also be referred to as first information, without departing from the scope of this application. Depending on the context, the term "if" as used herein may be interpreted as "when," "when," or "in response to determining."

[0070] Figure 1 FIG. 1 is a flow chart of a method for extracting arc fault features based on self-convolution according to an embodiment of the present invention. Figure 1 As shown, the invention includes the following steps:

[0071] S1: Cutting the arc fault waveform to obtain first data, wherein the first data contains arc fault information.

[0072] Specifically, based on the characteristic that a fault arc causes high-frequency fluctuations in the power grid, the cutting position and length are adjusted to ensure that the first data sequence obtained by cutting contains key fault arc information.

[0073] The arc fault waveform is selected from: a parallel or series electrical signal waveform collected by a sensor when an arc fault occurs in a household appliance; a physical signal waveform when an arc fault occurs in a household appliance; an open source arc fault signal waveform on the Internet;

[0074] The arc fault information is selected from: an area where the arc fault signal waveform suddenly changes; an area where the arc fault signal waveform corresponds to a household appliance as a load;

[0075] When there is no fault arc, the signal waveform is the slow-changing harmonics of a normal power grid. When there is no fault arc, the signal waveform is the slow-changing harmonics of a normal power grid with household appliances as loads.

[0076] The complete fault arc waveform or the left half or the right half of the complete fault arc waveform is cut to obtain first data.

[0077] Based on the characteristic that the fault arc causes high-frequency fluctuations in the power grid, the cutting position and length are adjusted to ensure that the first data sequence obtained by cutting contains key fault arc information.

[0078] The fault arc waveform is cut to obtain the first data. The fault arc waveform includes: electrical signal waveforms such as parallel or series current signals or voltage signals collected by sensors when a fault arc occurs in various types of household appliances such as resistive, inductive, capacitive, and switching power supplies; physical signal waveforms such as sound, light, and heat when a fault arc occurs in various types of household appliances such as resistive, inductive, capacitive, and switching power supplies; and open source sound, light, electricity, heat, and other fault arc signal waveforms on the Internet.

[0079] Methods for cutting the fragment sequence include: cutting manually; designing a script program for automatic cutting; and using an open source fault arc fragment sequence method on the Internet.

[0080] Cut the fault arc waveform to obtain first data, where the first data contains fault arc information. The fault arc information includes: the first data sequence must contain the characteristics of a sudden change in the fault arc signal waveform; the first data sequence must conform to the characteristics of the fault arc signal waveform with various household appliances as loads; on the contrary, when there is no fault arc, the signal waveform is a slow-changing harmonic of a normal power grid; when there is no fault arc, the signal waveform is a slow-changing harmonic of a normal power grid with various household appliances as loads.

[0081] S2: Process the first data to obtain second data.

[0082] Specifically, the processing process is divided into two processing processes based on whether the second data is complete. If the clipped fault arc waveform is a complete fault arc waveform, the mean of the first data is adjusted to 0, thereby obtaining the second data. If the clipped fault arc waveform is the left half or right half of the complete fault arc waveform, the processing includes center-symmetrical replication and mean adjustment, thereby obtaining the second data, wherein the center-symmetrical replication is to replicate the left half or right half of the fault arc waveform as the right half or left half of the fault arc waveform and splice it with the left half or right half of the original fault arc waveform to construct a complete fault arc waveform, wherein the mean is adjusted so that the mean of the constructed complete fault arc waveform is 0.

[0083] The above processing steps may also be normalized. Normalization is to control the interval distribution of the second data and prevent overflow and errors in subsequent data calculations. Normalization may not be performed within a controllable range.

[0084] The second data is used as a special convolution kernel for extracting fault arc features, including: a convolution kernel specifically used to extract physical fault arc features such as sound, light, and heat when a fault arc occurs in various types of household appliances such as resistive, inductive, capacitive, and switching power supplies; and a special convolution kernel for extracting fault arc features such as sound, light, electricity, and heat that is open source on the Internet.

[0085] S3: The power grid signal collected by the sensor and pre-processed is used as the third data.

[0086] Specifically, the sensor converts the signal type into an analog electrical signal, and the analog-to-digital converter converts the analog electrical signal into a digital electrical signal.

[0087] More specifically, acquisition includes: using various sound, light, electricity, heat and other sensors to collect; after collection, it can be quantified into digital signals; it is collected directly or through connections such as the Internet of Things and the Internet.

[0088] Specifically, the preprocessing includes adjusting the amplitude range and filtering.

[0089] Furthermore, the preprocessing also includes: extracting the collected signals to reduce the data volume, standardizing the data to make it in a unified format, and other preprocessing operations.

[0090] The third data includes: physical signals such as sound, light, and heat when various types of household appliances such as resistive, inductive, capacitive, and switching power supplies are running in the power grid; the signal may contain a fault arc or may not contain a fault arc, and the main purpose is to detect a fault arc.

[0091] S4: Perform convolution calculation on the second data and the third data to obtain fourth data.

[0092] Specifically, convolution calculations are divided into real-time and non-real-time calculations. Real-time convolution calculations: The second data is stored in advance. When the third data is input, the second data is convolved with the third data to obtain the fourth data. Non-real-time convolution calculations: The second and third data are stored in advance and the second data is convolved with the third data to obtain the fourth data.

[0093] Convolution calculation, the hardware carrier includes: using ASIC chips, MCU, FPGA, PC and cloud servers and other computing devices for calculation; the timeliness of convolution calculation is real-time or non-real-time calculation; the convolution calculation is implemented by performing sliding window convolution on the constructed second data and the collected third data. When the third data is serially input and updated, the convolution calculation can be implemented by fixing the second data, multiplier and adder.

[0094] Specifically, the relationship between the fourth data and the third data is: if the third data contains a fault arc, the amplitude of the location where the fault arc occurs in the fourth data is larger; if the third data contains a fault arc, the overall amplitude of the fourth data is lower and stable.

[0095] S5: Perform feature recognition processing on the fourth data to obtain fifth data, wherein the fifth data contains information on whether an arc fault occurs in the electrical equipment.

[0096] Specifically, the amplitude of the fourth data is compared with the set threshold, and the number of times the fourth data exceeds the set threshold is counted. If the counted number exceeds the limit, the fifth data output indicates that a fault arc is detected; the variation range of the fourth data is compared with the threshold, and the number of times the variation range of the fourth data exceeds the set threshold is counted. If the counted number exceeds the limit, the fifth data output indicates that a fault arc is detected; the variance of the fourth data is compared with the threshold, and the number of times the variance of the fourth data exceeds the set threshold is counted. If the counted number exceeds the limit, the fifth data output indicates that a fault arc is detected; the graph converted from the fourth data is compared with the preset graph, and the degree of similarity between the graph converted from the fourth data and the set graph is calculated. If the degree of similarity exceeds the limit, the fifth data output indicates that a fault arc is detected.

[0097] The feature processing methods here also include: filtering, denoising, variance calculation, pattern recognition and other feature processing algorithms.

[0098] The above method is applied to a specific embodiment below to demonstrate the specific implementation process and technical effects of the present invention.

[0099] This embodiment performs data acquisition based on the AC series current signal of the resistive load and uses the invented self-convolution algorithm to extract the fault arc characteristics.

[0100] 1. Cut the fault arc waveform to obtain the first data.

[0101] like Figure 2 The figure shows the current waveform of a series AC grid with a resistor as the load within a grid cycle of 20ms, which contains a fault arc. The self-convolution kernel is constructed based on the self-convolution algorithm of the invention.

[0102] like Figure 3 As shown in Figure 1, the time position of the fault arc segment is approximately between sampling points 4899 and 7179. However, the clipped sequence needs to be further adjusted.

[0103] like Figure 4 As shown, the final cutting results in a sequence of length 6: [44740 34003 34729 35164 3472934438] as the first data, which contains key fault arc information.

[0104] 2. Process the first data to obtain second data.

[0105] In order to make the first data be used as a convolution kernel to extract features for the fault arc occurring in two directions, that is, the fault arc position of the power grid waveform may be in the two directions where the first-order derivative of the power grid waveform is greater than 0 and less than 0, the extracted sequence is centrally symmetrically replicated to obtain a sequence with a length of 11 [34438 3472935164 34729 34003 44740 34003 34729 35164 34729 34438]. Figure 5 shown.

[0106] In order to extract only the fault arc signal features during the convolution process of the convolution kernel and the power grid signal without generating large distortion to the power grid signal, it is necessary to perform integral zeroing processing on the convolution kernel. That is, to ensure that during the convolution process of the power grid signal and the convolution kernel, the convolution kernel slides within the time domain range of the power grid signal, and the convolution sum of the power grid signal and the convolution kernel is 0. Only at the fault arc does the convolution sum suddenly increase, so as to achieve the function of extracting the fault arc features.

[0107] If the convolution kernel integral is not processed to 0, the feature extraction effect cannot be achieved. The extracted features cannot be distinguished from the distortion of the normal signal and are buried in the distortion of the convolution between the power grid signal and the convolution kernel.

[0108] The convolution kernel of the discrete Fourier transform is a series of standard sine waves with different frequencies, and the mean of each sine wave convolution kernel is 0. The convolution kernel of the discrete wavelet transform is a wavelet such as db4, and the mean of the wavelet convolution kernel is also 0.

[0109] Therefore, in order to make the integral of the convolution kernel equal to 0, each element of the convolution kernel needs to be reduced by a value equal to the mean of the convolution kernel.

[0110] After the previous step, the integral of the convolution kernel is 0, achieving the following effect: when the convolution kernel is convolved with the normal power grid signal, the convolution sum is 0; when the convolution kernel is convolved with the fault arc signal, the convolution sum is significantly increased. This achieves the function of fault arc feature extraction.

[0111] like Figure 6 As shown, the constructed convolution kernel is: [-1.0953 -0.8043 -0.3693 -0.8043 -1.53039.2067 -1.5303 -0.8043 -0.3693 -0.8043 -1.0953]×10 -3 .

[0112] The above convolution kernel sequence is the second data.

[0113] 3. The power grid signal collected and pre-processed by the sensor is used as the third data, and the second data and the third data are used to perform convolution calculation to obtain the fourth data.

[0114] The extracted, processed and constructed convolution kernel is convolved with the power grid signal.

[0115] like Figure 7 As shown in the figure, after the grid signal containing the fault arc is processed by the self-convolution algorithm, the convolution output value at the fault arc position is much larger than the convolution output value at the normal position. This means that the fault arc feature extraction function is realized.

[0116] The sequence length of the convolution kernel for fault arc feature extraction is 11, which has a very small number of parameters and is conducive to hardware implementation. It has advantages in realizing low-cost fault arc detection equipment.

[0117] At the same time, thanks to the simple algorithm structure, the algorithm output extension is extremely low, the feature extraction is fast, and the detection speed is fast.

[0118] At the same time, according to the experimental results and the comparison with the traditional algorithm below, a fault arc feature extraction method based on self-convolution is invented, which has good feature extraction effect and high accuracy of fault arc detection.

[0119] In summary, the invented fault arc feature extraction method based on self-convolution can achieve good fault arc feature extraction effect, fast extraction speed and low extraction hardware cost.

[0120] 4. Perform feature recognition processing on the fourth data to obtain fifth data, wherein the fifth data contains information on whether a fault arc occurs in the electrical equipment.

[0121] The fifth data includes the arc fault detection conclusion, i.e., whether there is an arc fault or not. The fifth data may be in the form of a high or low level, an electronic message, or a physical phenomenon such as whether a light is on or a horn is sounded.

[0122] In summary, thanks to the invented self-convolution-based fault arc feature extraction method, fault arc detection can achieve high accuracy, fast detection speed and low detection hardware cost.

[0123] like Figure 8 As shown, the top one is the third data, the middle one is the result of the traditional feature extraction method based on discrete wavelet transform, and the bottom one is the fourth data, which is the result of the fault arc feature extraction method based on self-convolution. It can be seen that the fault arc feature extraction method based on self-convolution has excellent extraction effect.

[0124] like Figure 9As shown, after absolute value processing, the left figure a is the absolute value of the processing result of the invented self-convolution algorithm, and the right figure b is the absolute value of the processing result of the classic discrete wavelet transform algorithm.

[0125] The processing results of the two algorithms are rearranged from most significant to least significant. The left figure shows the invented self-convolution algorithm, and the right figure shows the classic discrete wavelet transform algorithm. Note that the invented self-convolution algorithm transforms the arc fault signal to 400 million, far greater than the 20,000 obtained by the discrete wavelet transform algorithm on the right. In terms of gain for the arc fault signal, the invented self-convolution algorithm's response is calculated to be 19,600 times that of the traditional discrete wavelet transform algorithm. Meanwhile, the invented self-convolution algorithm's response to normal grid signals is far less than 19,600 times that of the traditional discrete wavelet transform algorithm.

[0126] like Figure 10 As shown, the sequences processed by the two feature processing algorithms are sorted from large to small, and the top 1,000 largest sequences are plotted. Figure 10 The a in it is the invented self-convolution algorithm, Figure 10 The b in is the classic discrete wavelet transform algorithm. Figure 10 The relative size relationship between the amplitude of the front number and the amplitude of the back number in the two figures can be seen. Figure 10 The larger the front number, the better. Figure 10 The smaller the amplitude of the last number, the better. The self-convolution algorithm's gain for fault arcs and suppression of normal grid signals are both better than the classic discrete wavelet transform algorithm. The invented self-convolution algorithm in the left figure also has an advantage in preventing misjudgments. The fault arc response of the self-convolution algorithm has large spikes and a small number of spikes, and it drops significantly after less than 100 spikes. The discrete wavelet transform response in the right figure has low spike height and an excessive number of spikes, and the response only drops significantly after 100 spikes. Given that the number of fault arcs in the known test waveform is 3, the invented self-convolution algorithm also has an advantage in preventing normal grid signals from being misjudged as fault arcs.

[0127] The following quantitative comparison is made between the invented fault arc feature extraction method based on self-convolution and the traditional fault arc feature extraction method based on discrete wavelet transform.

[0128] In order to quantitatively compare the effects of discrete wavelet transform algorithm and invented self-convolution algorithm in extracting fault arc features, a quantitative method for extracting fault arc features is proposed.

[0129]

[0130] In order to facilitate the calculation of the above equation, a discrete fault arc feature extraction and quantification method is proposed.

[0131]

[0132] In the above equation, n is the maximum length of the grid sequence containing a fault arc, x is the value after the discrete wavelet transform algorithm or the self-convolution algorithm, arc1, arc2, arc3, ​​arc4... are the first, second, third, fourth... fault arc occurrence locations, and k is the number of fault arc occurrences.

[0133] According to the proposed method for quantifying feature extraction, the grid waveform contains three fault arcs. Therefore, the absolute values ​​of the three largest values ​​from the two algorithms are summed to represent the feature extraction method's gain for the fault arcs. The absolute values ​​of all values ​​from the two algorithms are also summed to represent the feature extraction method's gain for the normal signal without fault arcs. According to the proposed quantitative evaluation method for feature extraction, the ratio of fault arc gain to normal signal gain without fault arcs yields a quantified feature extraction evaluation value of 92.8647 for the invented self-convolution algorithm, compared to 86.3501 for the classic discrete wavelet transform algorithm. The proposed self-convolution method, which constructs a specialized convolution kernel, outperforms the traditional discrete wavelet transform method.

[0134] According to the proposed quantitative evaluation method of feature extraction effect, a fault arc feature extraction method based on self-convolution is superior to the classic discrete wavelet transform algorithm.

[0135] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0136] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0137] Figures 11 to 16 This is a dedicated convolution kernel for arc fault detection according to an exemplary embodiment. Because this dedicated convolution kernel is derived from the arc fault itself, it is also called a "self-convolution kernel". Specifically: Figure 11FIG1 is a diagram illustrating an electric drill self-convolution kernel constructed according to an exemplary embodiment. Figure 12 The figure shows a computer self-convolution kernel constructed according to an exemplary embodiment. Figure 13 FIG. 4 is a fan self-convolution kernel constructed according to an exemplary embodiment. Figure 14 FIG. 4 is a fluorescent lamp self-convolution kernel constructed according to an exemplary embodiment. Figure 15 FIG1 is a self-convolution kernel for an electric kettle constructed according to an exemplary embodiment.

[0138] Corresponding to the aforementioned embodiment of the method for extracting arc fault features based on self-convolution, the present application also provides an embodiment of a device for extracting arc fault features based on self-convolution.

[0139] Figure 16 FIG1 is a block diagram of a fault arc feature extraction device based on self-convolution according to an exemplary embodiment. Figure 16 , the device comprises:

[0140] A cutting module 1 is used to cut the arc fault waveform to obtain first data, wherein the first data contains arc fault information;

[0141] A first processing module 2, configured to process the first data to obtain second data;

[0142] A second processing module 3 is used to use the power grid signal collected by the sensor and pre-processed as third data;

[0143] a convolution calculation module 4, configured to perform a convolution calculation on the second data and the third data to obtain fourth data;

[0144] The identification module 5 is used to perform feature identification processing on the fourth data to obtain fifth data, wherein the fifth data contains information on whether an arc fault occurs in the electrical equipment.

[0145] Regarding the apparatus in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0146] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present application scheme. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0147] Accordingly, the present application also provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the self-convolution-based fault arc feature extraction method as described above.

[0148] Correspondingly, the present application also provides a computer-readable storage medium having computer instructions stored thereon, characterized in that when the instructions are executed by a processor, the above-mentioned self-convolution-based fault arc feature extraction method is implemented.

[0149] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the contents disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present application that follow the general principles of the present application and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered merely as exemplary, and the true scope and spirit of the present application are indicated by the claims.

[0150] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A fault arc feature extraction method based on self-convolution, characterized in that: include: Cutting the fault arc waveform to obtain first data, wherein the first data contains fault arc information; processing the first data to obtain second data; The power grid signal collected by the sensor and pre-processed is used as the third data; Performing a convolution calculation on the second data and the third data to obtain fourth data; Performing feature recognition processing on the fourth data to obtain fifth data, wherein the fifth data contains information on whether an arc fault occurs in the electrical equipment; The step of shearing the fault arc waveform to obtain first data includes: Cutting the complete fault arc waveform or cutting the left half or the right half of the complete fault arc waveform to obtain first data; Processing the first data to obtain second data includes: If the clipped fault arc waveform is a complete fault arc waveform, adjusting the mean of the first data to be 0, thereby obtaining second data; If the clipped fault arc waveform is the left half or the right half of the complete fault arc waveform, the processing includes center-symmetric replication and adjustment of the mean, thereby obtaining second data, wherein the center-symmetry is to replicate the left half or the right half of the fault arc waveform as the right half or the left half of the fault arc waveform and to splice it with the left half or the right half of the original fault arc waveform to construct a complete fault arc waveform, wherein the mean is adjusted so that the mean of the constructed complete fault arc waveform is 0; Performing a convolution calculation on the second data and the third data to obtain fourth data includes: Real-time convolution calculation: The second data is stored in advance. When the third data is input, the second data is convolved with the third data to obtain the fourth data. Non-real-time convolution calculation: the second data and the third data are stored in advance, and the second data and the third data are convolved to obtain the fourth data; Performing feature recognition processing on the fourth data to obtain fifth data includes: comparing the amplitude of the fourth data with a set threshold, counting the number of times the fourth data exceeds the set threshold, and outputting fifth data indicating that an arc fault has been detected if the counted number exceeds the limit; Comparing the variation range of the fourth data with a threshold value, counting the number of times the variation range of the fourth data exceeds the set threshold value, and outputting fifth data indicating that a fault arc is detected if the counted number of times exceeds the limit value; Comparing the variance of the fourth data with a threshold, counting the number of times the variance of the fourth data exceeds the set threshold, and outputting fifth data indicating that a fault arc is detected if the counted number exceeds the limit; The graph converted from the fourth data is compared with a preset graph, and the similarity between the graph converted from the fourth data and the preset graph is calculated. If the similarity exceeds a limit, the fifth data outputted indicates that a fault arc is detected.

2. The method according to claim 1, characterized in that The arc fault waveform is selected from: The parallel or series electrical signal waveform collected by the sensor when a fault arc occurs in a household appliance; Physical signal waveform when a fault arc occurs in household appliances; Arc fault signal waveforms open sourced on the Internet; The arc fault information is selected from: The area where the arc fault signal waveform suddenly changes; The area that matches the arc fault signal waveform when household appliances are used as loads; When there is no fault arc, the signal waveform is the slow-changing harmonics of a normal power grid. When there is no fault arc, the signal waveform is the slow-changing harmonics of a normal power grid with household appliances as loads.

3. The method according to claim 1, characterized in that The power grid signal collected by the sensor and pre-processed is used as the third data, including: The sensor collects the power grid signal to obtain an analog electrical signal, and converts the analog electrical signal into a digital electrical signal through an analog-to-digital converter; The digital electrical signal is subjected to amplitude adjustment and filtering to obtain third data.

4. A fault arc feature extraction device based on self-convolution, characterized in that: include: a cutting module, configured to cut the arc fault waveform to obtain first data, wherein the first data contains arc fault information; A first processing module, configured to process the first data to obtain second data; A second processing module is used to use the power grid signal collected by the sensor and pre-processed as third data; a convolution calculation module, configured to perform a convolution calculation on the second data and the third data to obtain fourth data; an identification module, configured to perform feature identification processing on the fourth data to obtain fifth data, wherein the fifth data contains information on whether an arc fault occurs in the electrical equipment; The step of shearing the fault arc waveform to obtain first data includes: Cutting the complete fault arc waveform or cutting the left half or the right half of the complete fault arc waveform to obtain first data; Processing the first data to obtain second data includes: If the clipped fault arc waveform is a complete fault arc waveform, adjusting the mean of the first data to be 0, thereby obtaining second data; If the clipped fault arc waveform is the left half or the right half of the complete fault arc waveform, the processing includes center-symmetric replication and adjustment of the mean, thereby obtaining second data, wherein the center-symmetry is to replicate the left half or the right half of the fault arc waveform as the right half or the left half of the fault arc waveform and to splice it with the left half or the right half of the original fault arc waveform to construct a complete fault arc waveform, wherein the mean is adjusted so that the mean of the constructed complete fault arc waveform is 0; Performing a convolution calculation on the second data and the third data to obtain fourth data includes: Real-time convolution calculation: The second data is stored in advance. When the third data is input, the second data is convolved with the third data to obtain the fourth data. Non-real-time convolution calculation: the second data and the third data are stored in advance, and the second data and the third data are convolved to obtain the fourth data; Performing feature recognition processing on the fourth data to obtain fifth data includes: comparing the amplitude of the fourth data with a set threshold, counting the number of times the fourth data exceeds the set threshold, and outputting fifth data indicating that an arc fault has been detected if the counted number exceeds the limit; Comparing the variation range of the fourth data with a threshold value, counting the number of times the variation range of the fourth data exceeds the set threshold value, and outputting fifth data indicating that a fault arc is detected if the counted number of times exceeds the limit value; Comparing the variance of the fourth data with a threshold, counting the number of times the variance of the fourth data exceeds the set threshold, and outputting fifth data indicating that a fault arc is detected if the counted number exceeds the limit; The graph converted from the fourth data is compared with a preset graph, and the similarity between the graph converted from the fourth data and the preset graph is calculated. If the similarity exceeds a limit, the fifth data outputted indicates that a fault arc is detected.

5. An electronic device, characterized in that: include: one or more processors; a memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 3.

6. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the instruction is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images