An arc detection method, device, and medium based on continuous half-wave signals.
By performing high-frequency sampling and FFT transformation on continuous half-wave signals of power system lines, combined with neural network models and sliding window statistics, the problems of accuracy and anti-interference in arc detection were solved, achieving more efficient arc detection.
Patent Information
- Application Number
- CN202311518496.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-15
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2043-11-15
AI Technical Summary
Existing arc detection methods in power systems suffer from problems such as random arc location, difficulty in sensor installation covering all lines, high deployment costs, and misjudgment or missed detection in multi-load, multi-line environments.
An arc detection method based on continuous half-wave signals is adopted. By performing high-frequency sampling and FFT transformation on the current signal of the circuit under test, a neural network model is used to identify arc feature information. Combined with sliding window statistics and multi-channel frequency band analysis, the accuracy of arc detection is improved.
It effectively reduces the impact of external interference, improves the accuracy and anti-interference ability of arc detection, adapts to complex power environments, and reduces the false judgment rate.
Smart Images

Figure CN117572167B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of fault arc detection, and in particular to an arc detection method, device and medium based on continuous half-wave signals. Background Technology
[0002] A fault arc is a momentary spark phenomenon caused by the breakdown of insulation by line current. The generation of arcs can affect the reliability and safety of power system operation, and may even cause electrical fires and unpredictable hazards. Therefore, accurate and timely detection of fault arcs is a crucial aspect of power system operation.
[0003] Currently, the most common and practical methods for detecting electric arcs are to use electric arc detection equipment or sensors to detect faulty electric arcs.
[0004] Among them, the solution of detecting fault arcs by sensors has problems such as random location of fault arc generation, difficulty in determining the installation location of sensors, difficulty in covering all lines that may have fault arcs in practical applications, high deployment cost, and difficulty in implementation.
[0005] Arc detection equipment, limited by the level of electronic technology, mostly identifies arcs by using low-frequency current thresholds. However, due to the complexity of actual electrical circuits, with multiple loads and lines, it is difficult to calculate arc current uniformly, leading to problems of missed or false detection of faulty arcs.
[0006] Therefore, those skilled in the art urgently need an arc detection method based on continuous half-wave signals to solve the problem that it is difficult to accurately detect arc faults in power systems during operation. Summary of the Invention
[0007] The purpose of this application is to provide an arc detection method, device, and medium based on continuous half-wave signals, so as to solve the problem that it is difficult to accurately detect arcs caused by line faults during the operation of power systems.
[0008] To address the aforementioned technical problems, this application provides an arc detection method based on a continuous half-wave signal, comprising:
[0009] The continuous half-wave current signal of the line under test is sampled at high frequency and transformed by FFT to obtain multiple sets of arc feature information corresponding to different half-waves. Each set of arc feature information includes sub-feature information corresponding to different frequency bands.
[0010] Each set of arc feature information is input into a pre-established prediction model to determine the arc discrimination result of each consecutive half-wave; wherein, the prediction model is a neural network model trained in advance based on historical data of arc feature information;
[0011] Based on the arc discrimination results of each consecutive half-wave, the arc detection results of the circuit under test are determined.
[0012] On the other hand, based on the arc discrimination results of each consecutive half-wave, the arc detection results of the circuit under test are determined to include:
[0013] Using a first preset number of half-waves as the sliding window size and a second preset number of half-waves as the moving step size, the arc discrimination results of each consecutive half-wave are statistically analyzed using a sliding window. If all arc discrimination results within the current sliding window are arcs, the statistical value is incremented by one.
[0014] If the ratio of the statistical value to the total number of sliding windows is greater than or equal to the preset discrimination ratio, the result of the arc detection is determined to be an arc; otherwise, it is determined to be a non-arc.
[0015] On the other hand, the arc characteristic information of each group is input into a pre-established prediction model to determine the arc discrimination results of each consecutive half-wave, including:
[0016] The sub-feature information of each frequency band is input into the prediction model to obtain the arc discrimination result corresponding to each sub-feature information;
[0017] Based on the arc discrimination results of each sub-feature information, the arc discrimination result of the current arc feature information is determined.
[0018] On the other hand, based on the arc discrimination results of each sub-feature information, the arc discrimination results of the current arc feature information are determined as follows:
[0019] When the value of the arc discrimination result is 0, it represents a non-arc; when it is 1, it represents an arc. The values of the arc discrimination results of each sub-feature information are accumulated.
[0020] If the accumulated value is greater than or equal to the preset discrimination threshold, the current arc feature information is determined to be an arc; otherwise, it is determined to be a non-arc. The discrimination threshold is less than or equal to the number of sub-feature information of different frequency bands included in the arc feature information.
[0021] On the other hand, based on the arc discrimination results of each sub-feature information, the arc discrimination results of the current arc feature information are determined as follows:
[0022] When the value of the arc discrimination result is 0, it represents a non-arc; when it is 1, it represents an arc. An XOR calculation is performed between 1 and the value of the arc discrimination result of each sub-feature information.
[0023] If the XOR result is 1, then the arc discrimination result of the current arc feature information is determined to be an arc;
[0024] If the XOR result is 0, then the arc discrimination result of the current arc feature information is determined to be non-arc.
[0025] On the other hand, building a predictive model includes:
[0026] The training set data is obtained by concatenating all the sub-feature information of each half-wave arc feature information;
[0027] The training data is input into the neural network to obtain the prediction model.
[0028] On the other hand, neural networks are fully connected networks, and fully connected networks have four layers, with the number of neurons in each layer being n, x, y, and 1, respectively.
[0029] Where n is the number of sub-arc feature information of different frequency bands included in the arc feature information; x and y are values less than n and greater than 1, respectively.
[0030] To address the aforementioned technical problems, this application also provides an arc detection device based on a continuous half-wave signal, comprising:
[0031] The sampling module is used to perform high-frequency sampling and FFT transformation on the continuous half-wave current signal of the line under test to obtain multiple sets of arc feature information corresponding to different half-waves. Each set of arc feature information includes sub-feature information corresponding to different frequency bands.
[0032] The prediction module is used to input the arc feature information of each group into the pre-established prediction model to determine the arc discrimination result of each consecutive half wave; wherein, the prediction model is a neural network model trained in advance based on historical data of arc feature information;
[0033] The discrimination module is used to determine the arc detection result of the circuit under test based on the arc discrimination results of each consecutive half-wave.
[0034] To address the aforementioned technical problems, this application also provides an arc detection device based on a continuous half-wave signal, comprising:
[0035] Memory, used to store computer programs;
[0036] A processor is used to execute computer programs to implement the steps of the arc detection method based on continuous half-wave signals as described above.
[0037] 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 arc detection method based on a continuous half-wave signal as described above.
[0038] This application provides an arc detection method based on continuous half-wave signals. First, the current signal of the circuit under test is sampled at high frequency and transformed using FFT to convert it from a single signal in the time domain into a multi-channel high-frequency signal corresponding to different frequency bands in the frequency domain. This time-to-frequency conversion helps reduce external interference. Furthermore, since external interference is unlikely to simultaneously interfere with signals in different frequency bands across multiple channels, the multi-channel current signal can reflect more characteristics of the current signal compared to a single channel. Therefore, comprehensive arc identification using multi-frequency signals can further improve the anti-interference capability and accuracy of the current signal. Moreover, in complex actual power environments, the current signal may experience sudden changes in its normal stable waveform, resembling an arc, due to strong interference, which can easily lead to misjudgment. Therefore, this solution specifically detects continuous half-wave current signals to determine whether an arc has occurred in the circuit under test. This can identify interference occurring in one or more half-wave current signals, thereby achieving more accurate arc detection.
[0039] The above improvements address the current signal aspect, enhancing the accuracy of arc detection. Furthermore, this solution utilizes a pre-trained neural network prediction model to detect arcing in the circuit under test, thus improving the accuracy of arc detection from the perspective of arc discrimination. By combining these two improvements to traditional arc detection schemes, this solution better meets the needs of accurate arc detection in electrical circuits during practical production and daily life.
[0040] The arc detection device and computer-readable storage medium based on continuous half-wave signals provided in this application correspond to the above-mentioned method and have the same effect. Attached Figure Description
[0041] 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.
[0042] Figure 1 A flowchart of an arc detection method based on a continuous half-wave signal provided by the present invention;
[0043] Figure 2 A flowchart of another arc detection method based on continuous half-wave signals provided by the present invention;
[0044] Figure 3 A schematic diagram of the inner sliding window of the continuous half-wave discrimination strategy provided by the present invention;
[0045] Figure 4A structural diagram of an arc detection device based on a continuous half-wave signal provided by the present invention;
[0046] Figure 5 This is a structural diagram of another arc detection device based on a continuous half-wave signal provided by the present invention. Detailed Implementation
[0047] 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.
[0048] The core of this application is to provide an arc detection method, device, and medium based on continuous half-wave signals.
[0049] 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.
[0050] Currently, the main methods for detecting fault arcs are as follows:
[0051] The first method primarily uses sensors to detect characteristics such as arc light, arc sound, and electromagnetic radiation in the circuit under test. However, because the location of the fault arc is random, the placement of the sensors is difficult to determine, and completely covering all lines in the power system with sensors is also costly, making it difficult to apply in practice.
[0052] Therefore, the second approach described below is currently the most commonly used method for arc detection in cables. This involves collecting the current in the circuit under test using an arc detection device and analyzing the current ratio, root mean square difference, kurtosis, waveform factor, etc. Time-frequency domain features are extracted using Empirical Mode Decomposition (EMD), Kalman filtering, Fourier transform, wavelet transform, and harmonic analysis, and threshold values are set for comparison. However, this method is limited by complex circuit environments with multiple loads and multiple lines, and frequently suffers from misjudgments.
[0053] Therefore, to address the problems of easy misjudgment and missed detection, and poor detection accuracy of traditional arc detection schemes, this application provides an arc detection method based on continuous half-wave signals, such as... Figure 1 As shown, it includes:
[0054] S11: Perform high-frequency sampling and FFT transformation on the continuous half-wave current signal of the line under test to obtain multiple sets of arc feature information corresponding to different half-waves. Each set of arc feature information includes sub-feature information corresponding to different frequency bands.
[0055] S12: Input the arc characteristic information of each group into the pre-established prediction model to determine the arc discrimination result of each continuous half wave.
[0056] The prediction model is a neural network model trained in advance based on historical data of electric arc characteristics.
[0057] S13: Determine the arc detection result of the circuit under test based on the arc discrimination results of each continuous half-wave.
[0058] For step S11, it is readily known that there are already many mature technical solutions for high-frequency sampling and Fast Fourier Transform (FFT) of the current signal to extract its feature information. Therefore, this embodiment will not elaborate further on the specific implementation of step S11. The purpose of step S11 is to sample the features of the current signal of the line under test through high-frequency sampling and FFT transformation, convert the time-domain signal into a high-frequency-domain signal, and obtain multiple frequency-domain signals divided into multiple channels based on different frequency bands to obtain arc feature information that can further reflect whether an arc has occurred in the line under test.
[0059] It should also be noted that step S11 should be a continuous process. By continuously sampling the current signal of the circuit under test at high frequency and performing FFT transformation, multiple sets of arc feature information of multiple consecutive half waves are obtained. Each set of arc feature information includes multiple sub-feature information corresponding to different frequency bands. Then, subsequent steps can be used to determine whether an arc has occurred in the circuit under test.
[0060] For step S12, the pre-established prediction model is a machine learning model trained on a neural network, used to determine whether an arc phenomenon has occurred in the current test line based on the input arc feature information. Based on the characteristic that machine learning models continuously improve their prediction accuracy as they learn (with the continuous increase of training data), arc detection with more guaranteed prediction accuracy is achieved.
[0061] It's easy to understand that the amount of historical data used to train the prediction model can be determined based on the required accuracy of actual arc detection. Generally, the higher the required prediction accuracy, the larger the amount of historical data is needed. Furthermore, as the prediction model is used in practice, the data acquired during use can be used as new historical data to update and train the prediction model, thereby further improving the accuracy of this method in arc detection.
[0062] It should also be noted that the output of the training model is usually a probability value, that is, the probability that an electric arc may occur in the current circuit under test. As mentioned above, the arc detection result can be represented in a simple form as arc and non-arc (i.e., normal), reducing computational difficulty. Therefore, this embodiment also provides the following scheme for converting the probability value obtained from the prediction model into the two arc detection results:
[0063]
[0064] Where i = 1, 2, ..., N, represents the i-th half-wave signal; o(i) represents the i-th predicted probability value obtained by the neural network; P i This indicates the result of the electric arc detection, where 1 represents an electric arc and 0 represents a non-electric arc.
[0065] Regarding step S13, it should be noted that this method comprehensively determines the final arc detection result of the circuit under test based on the arc discrimination results of continuous half-wave signals. That is, the arc discrimination result is an intermediate result in the entire process, characterizing whether a certain half-wave signal exhibits arc characteristics. The arc detection result, on the other hand, is the final result of the entire process, characterizing whether an arc phenomenon currently exists in the circuit under test. Furthermore, the arc discrimination result and the arc detection result can be simply divided into two types: arc and non-arc.
[0066] Step S13 uses a continuous half-wave signal to identify the arc in order to improve the accuracy of arc detection. This is because in the actual operation of a power system, the power environment of the line under test is complex, with many loads and branches. When multiple loads are mixed, it is difficult to determine the resistivity, capacitance, and inductive properties of the circuit, and it is even more difficult to eliminate the influence of external interference on arc detection.
[0067] For example, in actual detection processes, strong interference signals may cause a normally stable current signal to exhibit arc-like abrupt changes within one or more half-cycles. In such cases, using a single half-wave period or a discrete half-wave statistical discrimination strategy may lead to misjudgment of the arc, hindering the improvement of arc detection accuracy.
[0068] The "false" arc characteristics caused by strong interference signals are usually sudden and sporadic, and will not cause the current signal to exhibit arc characteristics over a long period of time. Therefore, the arc detection method provided in this application, by sampling continuous half-wave current signals, predicting arc results for each continuous half-wave current signal, and finally determining whether an arc has occurred in the circuit under test based on the arc prediction results of the continuous half-wave current signals, can effectively avoid the influence of strong interference on arc discrimination, thereby improving the accuracy of arc detection.
[0069] As described above, this application provides an arc detection method based on continuous half-wave signals, which improves the accuracy of arc detection mainly in the following aspects: First, by performing high-frequency sampling and FFT transformation on the current signal, the signal is converted from the time domain to the frequency domain, which reduces the impact of external interference on the detection results to a certain extent; Second, by converting a single-channel signal into a multi-frequency channel signal, the representativeness and anti-interference ability of the signal are improved, further enhancing the accuracy of arc detection; Third, the prediction model established using a neural network model is used to determine whether an arc has occurred. The accuracy of machine learning models is generally more guaranteed, and the accuracy can be improved with continuous learning, further ensuring the accuracy of arc detection by this method; Fourth, considering that in the complex operating environment of actual power systems, sudden strong interference may cause the current signal to exhibit characteristics similar to an arc, which can easily lead to misjudgment, this method uses multiple continuous half-wave signals for arc detection, which can effectively avoid a series of short-term, discrete interference factors, including sudden strong interference, further improving the accuracy of this method in detecting whether an arc has occurred on the line.
[0070] The arc detection method provided in this application improves the accuracy of arc detection by addressing both signal and arc discrimination aspects. Therefore, the arc detection accuracy achieved by this solution is significantly improved compared to traditional solutions, making it adaptable to the complex power environment of actual power system operation and thus meeting the arc detection needs in practical applications.
[0071] On the other hand, as can be seen from the above, this application does not limit the specific discrimination logic for determining whether an arc has occurred in the circuit under test based on multiple consecutive half-wave signals. The simplest discrimination logic can be:
[0072] When the number of consecutive half-wave signals that the prediction model identifies as an electric arc exceeds a preset threshold, it can be determined that an electric arc has occurred in the circuit under test; otherwise, it is determined to be a non-electric arc.
[0073] However, this embodiment further provides a preferred implementation scheme as a new continuous discrimination strategy for the discrimination logic of determining whether an arc has occurred in the line under test based on continuous half-wave signals. The above-mentioned step S13 specifically includes:
[0074] S131: Using the first preset number of half-waves as the sliding window size and the second preset number of half-waves as the moving step size, perform sliding window statistics on the arc discrimination results of each consecutive half-wave. If all arc discrimination results in the current sliding window are arcs, then increment the statistical value by one.
[0075] S132: If the ratio of the statistical value to the total number of sliding windows is greater than or equal to the preset discrimination ratio, then the arc detection result is determined to be an arc; otherwise, it is determined to be a non-arc.
[0076] In other words, this embodiment provides a statistical analysis of the discrimination results in a dual sliding window mode, and determines the discrimination logic for the arc detection results based on the statistical results. It is easy to understand that the first preset quantity, the second preset quantity, and the preset discrimination ratio mentioned above are not limited in their specific values in this embodiment, and appropriate values can be freely selected according to actual needs.
[0077] Specifically, the preferred solution provided in this embodiment will be further explained below with reference to examples:
[0078] like Figure 2 As shown, assuming there are 10 consecutive half-wave signals (represented by numbers 1 to 10), the arc discrimination result for each consecutive half-wave signal is determined through steps S11 and S12. A value of 1 represents an arc, and 0 represents a non-arc signal. The arc discrimination results for the 10 consecutive half-wave signals are 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, and so on. Therefore, the following applies: Figure 2 The results are shown in the diagram.
[0079] Furthermore, in one possible implementation scenario, the size of the sliding window is 5, meaning each sliding window includes 5 consecutive half-wave signals; the movement step size of the sliding window is 1, meaning each time the sliding window moves, it slides over one half-wave signal; for example, Figure 2 The 10 consecutive half-wave signals shown can be moved a total of 5 times by the sliding window. Each half-wave signal contained within the sliding window at different times is considered a base unit. Figure 2 There are a total of 5 base units, and each base unit includes 5 consecutive half-wave signals.
[0080] Furthermore, according to step S11, the statistical value is incremented by 1 only when the arc discrimination result of all half-wave signals within the sliding window is 1. Figure 2 It can be seen that only the arc discrimination result of the 5th half-wave signal is 0. At this point, only the base units containing half-wave signals 6-10 can have their statistical value incremented by 1. Therefore, for... Figure 2 In one scenario shown, the final statistical value is 1, and the total number of sliding windows = the total number of base cells = 6.
[0081] Furthermore, if the preset discrimination ratio is 2 / 3, assuming that the discrimination as in step S132 is performed every 30 slides (i.e., when 30 base units are generated), if the statistical value is greater than or equal to 20, then the arc detection result of the current circuit under test is determined to be an arc. Conversely, if the statistical value is less than 20, then the arc detection result of the current circuit under test is determined to be a non-arc.
[0082] Similarly, the statistical and discrimination principles of steps S131 and S132 above (taking 30 base units as the discrimination trigger condition for step S132, with a discrimination ratio of 2 / 3) can be expressed by the following formula:
[0083]
[0084]
[0085] Where final represents the arc detection result, Unit(i) represents the discrimination result of the i-th base unit, 1 represents an arc and 0 represents a non-arc; con_X represents the prediction result of X consecutive half waves, and X consecutive half waves are used as one base unit, denoted as 1 Unit; ∑(con_X(i)) represents the calculation result of the i-th base unit.
[0086] As described above, the arc detection results determined by the dual sliding window statistical mode provided in this embodiment can further avoid the impact of long-term interference on the arc detection results. Compared with the simple discrimination logic that judges the final detection result as an arc simply by the continuous occurrence of the discrimination result as an arc, this embodiment can still avoid misjudgment even when the interference caused by the interference is long and covers the smallest discrimination unit (i.e., the continuous occurrence of a certain number of half-wave signals can be used as a basis to determine the detection result as an arc), thus further improving the accuracy of arc detection.
[0087] On the other hand, considering that in the actual operation of power systems, converting the current signal into arc feature information of a single FFT feature channel using only FFT transformation results in weak arc features that can be represented, which to some extent affects the accuracy of subsequent arc detection and discrimination. Therefore, in step S11 above, converting the time-domain signal of a single channel into a multi-channel signal corresponding to different frequency bands allows the arc feature information to characterize more features of the current signal.
[0088] Furthermore, in step S12, when performing arc discrimination on each half-wave signal, it is necessary to consider a set of sub-feature information from different frequency bands of multiple channels corresponding to that half-wave signal for comprehensive arc discrimination. Based on this, this embodiment also provides a possible implementation scheme for the specific implementation of step S12 above, which further includes:
[0089] S121: Input the sub-feature information of each frequency band into the prediction model to obtain the arc discrimination result corresponding to each sub-feature information;
[0090] S122: Determine the arc discrimination result of the current arc feature information based on the arc discrimination results of each sub-feature information.
[0091] In other words, the preferred solution provided in this embodiment converts the current signal into N FFT feature channels based on different frequency bands in the frequency domain when sampling the current signal, thus obtaining N sub-feature information. Each sub-feature information can represent the current characteristics of the current signal in a certain frequency band. Compared with a single FFT feature channel (i.e., only one arc feature information), the feature representation dimension of each half-wave signal is increased by N times.
[0092] Furthermore, when a fault arc exists in the line, there will be obvious peak-shaped fluctuations and continuous fluctuations in the N FFT feature channels; while under normal circumstances, most interference will not affect all N frequency bands; therefore, the multi-channel discrimination strategy provided in this embodiment can further reduce the probability of arc misjudgment.
[0093] Furthermore, regarding how to determine the arc discrimination result of the corresponding half-wave signal based on the N FFT feature channels in step S122 above, this embodiment provides a possible implementation scheme:
[0094] When the value of the arc discrimination result is 0, it represents a non-arc; when it is 1, it represents an arc. The values of the arc discrimination results of each sub-feature information are accumulated.
[0095] If the accumulated value is greater than or equal to the preset discrimination threshold, the current arc feature information is determined to be an arc; otherwise, it is determined to be a non-arc. The discrimination threshold is less than or equal to the number of sub-feature information of different frequency bands included in the arc feature information.
[0096] It should be noted that the specific value of the discrimination threshold mentioned above is not strictly limited in this embodiment, and can be determined according to the actual accuracy requirements of arc detection. Generally, the discrimination threshold can be set to N. In this case, the accumulated value is at most equal to the discrimination threshold. That is, only when all N FFT feature channels are judged by the prediction model to have an arc will the arc discrimination result of the current half-wave signal be judged as 1 (i.e., an arc).
[0097] Similarly, the multi-channel discrimination strategy of the above embodiments can be expressed by the following mathematical formula:
[0098]
[0099] Where res represents the arc discrimination result of the current half-wave signal; P i This represents the arc discrimination result of a single FFT feature channel.
[0100] On the other hand, regarding how to determine the arc discrimination result of the corresponding half-wave signal based on the N FFT feature channels in the above step S122, this embodiment also provides another possible implementation scheme:
[0101] When the value of the arc discrimination result is 0, it represents a non-arc; when it is 1, it represents an arc. An XOR calculation is performed between 1 and the value of the arc discrimination result of each sub-feature information.
[0102] If the XOR result is 1, then the arc discrimination result of the current arc feature information is determined to be an arc;
[0103] If the XOR result is 0, then the arc discrimination result of the current arc feature information is determined to be non-arc.
[0104] In other words, unlike the previous embodiment which used summation as the standard to determine the arc discrimination result of the half-wave signal, this embodiment uses XOR logic operation. The advantage of this embodiment is that, when the above method is implemented in a practical circuit, the XOR logic calculation circuit is simpler and easier to implement than circuit structures such as adders. However, this embodiment also has the following disadvantages compared to the previous embodiment: it can only implement the discrimination logic that determines the arc discrimination result of the current half-wave signal as 1 when all N FFT feature channel result values are 1, i.e., it only corresponds to one possible implementation scenario of the above embodiment.
[0105] It should also be noted that an extra "1" is added when performing the XOR calculation on the arc discrimination results of each sub-feature information. The purpose is to avoid the situation where the XOR result is also "1" when all values are "0", that is, to avoid the situation where the XOR calculation results in an arc discrimination result (1) when all channels are determined to be non-arc (0). Thus, the arc discrimination result corresponding to the current half-wave signal is considered to be an arc (1) only when the frequency domain signals of all channels are determined to be arc (1) by the prediction model.
[0106] Similarly, adding multiple "1"s to the XOR calculation can also achieve the above purpose. Adding one "1" is only a preferred implementation.
[0107] Furthermore, as can be seen from the above embodiments, the current signal can be converted into N FFT feature channels through FFT transformation, thereby increasing the feature representation dimension of each half-wave signal by N times, providing a basis for more accurate arc detection.
[0108] It is easy to understand that even if the intermediate arc feature information is converted into sub-feature information of N FFT feature channels in the above embodiments, the same prediction model is used to predict the results for the sub-feature information of each FFT feature channel (it is also possible to use a separate prediction model for each frequency band, but the implementation cost is too high and it is generally not adopted). Therefore, there are higher requirements for the adaptability of the prediction model to different frequency bands.
[0109] Therefore, based on the above embodiments, this embodiment provides a further implementation scheme for training the prediction model, including:
[0110] The training set data is obtained by concatenating all the sub-feature information of each half-wave arc feature information;
[0111] The training data is input into the neural network to obtain the prediction model.
[0112] It should be noted that this embodiment does not limit the splicing method of each sub-feature information. The purpose is to send all the feature information representing each frequency band dimension of the half-wave signal into the neural network for training, that is, to provide a multi-channel fusion strategy so that the neural network can have good prediction accuracy for half-wave signals of each frequency band.
[0113] Furthermore, this embodiment also provides a possible implementation scheme for the neural network used to train the prediction model:
[0114] The neural network described above is a fully connected (FC) network, and the fully connected network has four layers, with the number of neurons in each layer being n, x, y, and 1, respectively.
[0115] Where n is the number of sub-arc feature information of different frequency bands included in the arc feature information; x and y are values less than n and greater than 1, respectively.
[0116] It is easy to see that the number of layers in a fully connected network is not fixed, and the four layers mentioned above are only one possible implementation.
[0117] Furthermore, when the number of layers in a fully connected network is greater than 2, the number of neurons in the first and last layers is fixed, with the first layer having n neurons, where n is equal to the N value of the aforementioned N FFT feature channels. The number of neurons in the intermediate layers (i.e., non-first and non-last layers) is not fixed, but is generally less than the number of neurons in the first layer and greater than the number of neurons in the last layer, thereby improving convergence efficiency. This embodiment provides a possible value for the number of neurons in the intermediate layers, which is n / 6.
[0118] Furthermore, when there are multiple intermediate layers, the number of neurons in each intermediate layer can be the same or different, but generally taking the same value can result in better performance in the training convergence of the actual model. That is, a preferred solution is: x = y = n / 6.
[0119] Furthermore, based on the above embodiments, the overall process of the arc detection method provided in this application will be further explained:
[0120] like Figure 3As shown, the continuous half-wave current signal of the circuit under test is subjected to high-frequency sampling and FFT transformation, and each half-wave will output N FFT feature channels. For FFT feature channels 1 to N, which have the same feature dimension, a multi-channel fusion strategy is used to feed them into the neural network for training to obtain a prediction model. For the continuous half-wave signal to be predicted, after high-frequency sampling and FFT transformation, the trained prediction model will give predictions for N FFT feature channels for each half-wave signal. These N prediction results are processed by a multi-channel discrimination strategy to output the arc discrimination result for each half-wave signal. For the predictions of the 1st to Xth consecutive half-waves, the continuous half-wave discrimination strategy is used to statistically analyze the prediction results of the consecutive half-waves to obtain the final arc detection result as either an arc or normal (non-arc).
[0121] In the above embodiments, a method for arc detection based on continuous half-wave signals has been described in detail. This application also provides an embodiment of an arc detection device based on continuous half-wave signals. It should be noted that this application describes the device embodiment from two perspectives: one based on functional modules and the other based on hardware.
[0122] From the perspective of functional modules, such as Figure 4 The embodiment shown provides an arc detection device based on a continuous half-wave signal, comprising:
[0123] The sampling module 21 is used to perform high-frequency sampling and FFT transformation on the continuous half-wave current signal of the line under test to obtain multiple sets of arc feature information corresponding to different half-waves. Each set of arc feature information includes sub-feature information corresponding to different frequency bands.
[0124] The prediction module 22 is used to input the arc feature information of each group into the pre-established prediction model to determine the arc discrimination result of each continuous half wave; wherein, the prediction model is a neural network model trained in advance based on historical data of arc feature information;
[0125] The discrimination module 23 is used to determine the arc detection result of the circuit under test based on the arc discrimination results of each continuous half-wave.
[0126] 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.
[0127] Figure 5 A structural diagram of an arc detection device based on a continuous half-wave signal, as provided in another embodiment of this application, is shown below. Figure 5 As shown, an arc detection device based on a continuous half-wave signal includes: a memory 30 for storing a computer program;
[0128] The processor 31 is used to execute a computer program to implement the steps of an arc detection method based on a continuous half-wave signal as described in the above embodiment.
[0129] The arc detection device based on continuous half-wave signals provided in this embodiment may include, but is not limited to, arc fault protection devices (AFDD) and arc fault detectors (AFD).
[0130] The processor 31 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 31 may be implemented using at least one of the following hardware forms: Digital Signal Processor (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 31 may also include a main processor and a coprocessor. The main processor, also known as the 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 31 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 31 may also include an Artificial Intelligence (AI) processor, which is used to handle computational operations related to machine learning.
[0131] The memory 30 may include one or more computer-readable storage media, which may be non-transitory. The memory 30 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 30 is used to store at least the following computer program 301, which, after being loaded and executed by the processor 31, is capable of implementing the relevant steps of the arc detection method based on a continuous half-wave signal disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 30 may also include an operating system 302 and data 303, and the storage method may be temporary or permanent storage. The operating system 302 may include Windows, Unix, Linux, etc. The data 303 may include, but is not limited to, an arc detection method based on a continuous half-wave signal.
[0132] In some embodiments, an arc detection device based on a continuous half-wave signal may further include a display screen 32, an input / output interface 33, a communication interface 34, a power supply 35, and a communication bus 36.
[0133] Those skilled in the art will understand that Figure 5 The structure shown does not constitute a limitation on an electric arc detection device based on a continuous half-wave signal, and may include more or fewer components than shown.
[0134] This application provides an arc detection device based on a continuous half-wave signal, which includes a memory and a processor. When the processor executes the program stored in the memory, it can implement the following method: an arc detection method based on a continuous half-wave signal.
[0135] 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 method embodiments.
[0136] 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.
[0137] The foregoing has provided a detailed description of an arc detection method, apparatus, and medium based on a continuous half-wave signal 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.
[0138] 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. An arc detection method based on a continuous half-wave signal, characterized by, include: The continuous half-wave current signal of the line under test is sampled at high frequency and transformed by FFT to obtain multiple sets of arc feature information corresponding to different half-waves. Each set of arc feature information includes sub-feature information corresponding to different frequency bands. Each set of arc feature information is input into a pre-established prediction model to determine the arc discrimination result for each consecutive half-wave; wherein, the prediction model is a neural network model pre-trained based on historical data of the arc feature information; the sub-feature information of each frequency band is input into the prediction model respectively to obtain the arc discrimination result corresponding to each sub-feature information; based on the arc discrimination result of each sub-feature information, the arc discrimination result of the current arc feature information is determined; Based on the arc discrimination results of each consecutive half-wave, the arc detection result of the line under test is determined; wherein, a value of 0 for the arc discrimination result indicates a non-arc, and a value of 1 indicates an arc, an XOR calculation is performed between 1 and the value of the arc discrimination result of each sub-feature information; if the XOR result is 1, the arc discrimination result of the current arc feature information is determined to be an arc; if the XOR result is 0, the arc discrimination result of the current arc feature information is determined to be a non-arc. Using a first preset number of half-waves as the sliding window size and a second preset number of half-waves as the movement step size, the arc discrimination results of each consecutive half-wave are statistically analyzed using a sliding window. If all the arc discrimination results within the current sliding window are arcs, the statistical value is incremented by one. If the ratio of the statistical value to the total number of sliding windows is greater than or equal to a preset discrimination ratio, the arc detection result is determined to be an arc; otherwise, it is determined to be a non-arc.
2. The arc detection method based on a continuous half-wave signal according to claim 1, characterized in that, Based on the arc discrimination results of each of the sub-feature information, the arc discrimination result of the current arc feature information is determined as follows: When the value of the arc discrimination result is 0, it represents a non-arc; when it is 1, it represents an arc. The values of the arc discrimination results for each of the sub-feature information are accumulated. If the accumulated value is greater than or equal to the preset discrimination threshold, the current arc feature information is determined to be an arc; otherwise, it is determined to be a non-arc. The discrimination threshold is less than or equal to the number of sub-feature information in different frequency bands included in the arc feature information.
3. The arc detection method based on a continuous half-wave signal according to claim 1, characterized in that, Establishing the prediction model includes: The training set data is obtained by concatenating all the sub-feature information of the arc feature information of each half-wave. The training set data is input into the neural network to obtain the prediction model.
4. The arc detection method based on continuous half-wave signals according to claim 2, characterized in that, The neural network is a fully connected network, and the fully connected network has four layers, with the number of neurons in each layer being n, x, y, and 1, respectively. Wherein, n is the number of sub-arc feature information of different frequency bands included in the arc feature information; x and y are values less than n and greater than 1, respectively.
5. An arc detection device based on a continuous half-wave signal, applicable to the arc detection method based on a continuous half-wave signal as described in any one of claims 1 to 4, characterized in that, The arc detection device based on continuous half-wave signals includes: The sampling module is used to perform high-frequency sampling and FFT transformation on the continuous half-wave current signal of the line under test to obtain multiple sets of arc feature information corresponding to different half-waves. Each set of arc feature information includes sub-feature information corresponding to different frequency bands. The prediction module is used to input the arc feature information of each group into a pre-established prediction model to determine the arc discrimination result of each consecutive half-wave; wherein, the prediction model is a neural network model pre-trained based on historical data of the arc feature information; The discrimination module is used to determine the arc detection result of the circuit under test based on the arc discrimination result of each consecutive half-wave.
6. An arc detection device based on a continuous half-wave signal, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the arc detection method based on a continuous half-wave signal as described in any one of claims 1 to 4.
7. 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 detection method based on a continuous half-wave signal as described in any one of claims 1 to 4.
Citation Information
Patent Citations
Fault arc signal detection method adopting convolutional neural network
CN113049922A
Fault arc detection method and fault arc detector
CN115840118A
