Series arc fault detection method and system

Abstract

The application belongs to the technical field of arc fault detection, and particularly relates to a series arc fault detection method and system, which comprises the following steps: acquiring a bus current signal; analyzing the acquired current signal, calculating and judging time domain characteristic quantities of the current signal when an arc fault occurs, and obtaining screened time domain characteristic quantities; performing energy decomposition on the screened time domain characteristic quantities, obtaining fuzzy entropy of a variational mode component, and combining the obtained fuzzy entropy to obtain characteristic evaluation weights; diagnosing the arc fault according to a support vector machine and the obtained characteristic evaluation weights, and realizing series arc fault detection.

Description

Technical Field

[0001] This invention belongs to the field of arc fault detection technology, specifically relating to a method and system for detecting series arc faults. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Low-voltage DC power supply boasts advantages such as high efficiency, ease of integration with energy storage devices, and high reliability, making it widely used in photovoltaic power generation, electric vehicles, large data centers, and other systems. It also holds broad application prospects in multi-electric aircraft, electrified ships, aerospace, and DC power distribution networks. In DC power supply systems, DC arcing faults can occur due to insulation failure, loose metal joints, component aging, or animal bites. Unlike AC arcing, where the current crosses zero and extinguishes spontaneously, DC arcing does not. If it occurs and is not detected promptly, the fault can spread to adjacent circuits, damaging photovoltaic modules, transmission lines, and control systems. In severe cases, the continued burning of the arc can even lead to a fire.

[0004] According to the inventors, current methods for detecting arc faults often involve selecting appropriate feature quantities as inputs into neural networks. However, these feature selections typically use single time-domain features as inputs to the neural network algorithm, resulting in a single detection index, low accuracy, instability, and insufficient fault information. Integrated energy systems contain numerous devices, and electrical sparks frequently occur in power lines. Once an arc fault occurs, it can cause significant damage to power lines, seriously threatening the safety and stability of the power system. Previously, arc fault detection and diagnosis often relied on single feature quantities as indicators, easily leading to erroneous judgments. Furthermore, the complexity of power lines makes it difficult to accurately pinpoint the location of the arc fault, resulting in a significant waste of manpower and resources through piecemeal investigations, making arc fault detection extremely inefficient. Summary of the Invention

[0005] To address the aforementioned issues, this invention proposes a series arc fault detection method and system. This method performs multi-dimensional extraction and analysis of arc fault characteristics, avoiding diagnostic errors caused by single-feature extraction due to limited indicators. This improves the accuracy of arc fault diagnosis, precisely diagnoses series arc faults occurring in integrated energy systems, and enhances the safety of integrated energy systems.

[0006] According to some embodiments, the first aspect of the present invention provides a method for detecting series arc faults, which adopts the following technical solution:

[0007] A method for detecting series arc faults, comprising:

[0008] Obtain the bus current signal;

[0009] Analyze the acquired current signal, calculate and determine the time-domain characteristics of the current signal when an arc fault occurs, and obtain the filtered time-domain characteristics.

[0010] The selected time-domain features are decomposed into energy to obtain the fuzzy entropy of the variational mode components. The feature evaluation weights are obtained by combining the obtained fuzzy entropy.

[0011] By using support vector machines and the obtained feature evaluation weights, arc faults can be diagnosed, thus realizing the fault detection of series arcs.

[0012] As a further technical limitation, during the analysis of the acquired current signal, the time-domain characteristics of the current signal are calculated and stored; the time-domain characteristics include peak-to-peak value, variance, standard deviation, mean, RMS value, kurtosis, waveform index, peak factor, impulse factor, and skewness.

[0013] Furthermore, the time-domain characteristics of the obtained current signal are monitored. When the time-domain characteristics exceed a set threshold, it is determined that there may be an arc fault, and the time-domain characteristics are selected for fault detection.

[0014] As a further technical limitation, an improved variational mode decomposition is adopted in the process of obtaining the fuzzy entropy of the variational mode components, namely, the variational mode decomposition is optimized by using the Australian Wild Dogs algorithm.

[0015] As a further technical constraint, the decomposition of the time-domain features after screening is used to obtain the fuzzy entropy of each variational mode component. The three variational mode components with the largest fuzzy entropy are selected, and the current signals are merged and reconstructed to obtain new signal components. The current time-domain features of the new signal components are calculated, and the feature evaluation weights of the current time-domain features of the obtained new signal components are calculated using the Relief-F feature selection algorithm.

[0016] As a further technical limitation, in the process of diagnosing arc faults based on support vector machines and the obtained feature evaluation weights, comprehensive time-domain features are constructed based on the screened time-domain features and feature evaluation weights. The constructed comprehensive time-domain features and the preset fault diagnosis model are combined to perform arc fault diagnosis. The obtained fault detection results are then used for secondary fault diagnosis through an improved support vector machine to realize the fault detection of series arcs.

[0017] Furthermore, the fault diagnosis model employs a random forest algorithm and an improved support vector machine; the improved support vector machine uses the Nordic Dogs Algorithm (DOA algorithm) to optimize the penalty factor and weight vector.

[0018] According to some embodiments, a second aspect of the present invention provides a series arc fault detection system, which adopts the following technical solution:

[0019] A series arc fault detection system, comprising:

[0020] The acquisition module is configured to acquire the bus current signal;

[0021] The calculation module is configured to analyze the acquired current signal, calculate and determine the time-domain features of the current signal when an arc fault occurs, and obtain the filtered time-domain features; perform energy decomposition on the filtered time-domain features to obtain the fuzzy entropy of the variational mode components, and combine the obtained fuzzy entropy to obtain the feature evaluation weights.

[0022] The detection module is configured to diagnose arc faults based on support vector machines and the obtained feature evaluation weights, thereby realizing the fault detection of series arcs.

[0023] According to some embodiments, a third aspect of the present invention provides a computer-readable storage medium, employing the following technical solution:

[0024] A computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the series arc fault detection method as described in the first aspect of the present invention.

[0025] According to some embodiments, the fourth aspect of the present invention provides an electronic device, which adopts the following technical solution:

[0026] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the series arc fault detection method as described in the first aspect of the present invention.

[0027] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0028] This invention enables multi-dimensional extraction and analysis of arc fault features, avoiding diagnostic errors caused by single-feature extraction due to limited indicators, and improving the accuracy of arc fault diagnosis. It also performs secondary feature extraction and analysis on lines suspected of arc faults, resulting in more accurate fault information contained within the features, further improving the detection accuracy of fault features. Furthermore, it constructs comprehensive time-domain features for fault type identification and diagnosis, ensuring the accuracy of arc fault detection results and achieving precise judgment.

[0029] This invention can accurately diagnose series arc faults occurring in integrated energy systems, improving the safety of integrated energy systems and providing a new approach for the accurate diagnosis of arc faults in future integrated energy systems. By using a cloud platform for real-time monitoring and alarm prompts, it can clearly indicate which line has an arc fault, saving manpower and resources and effectively ensuring the stability of power system lines. Attached Figure Description

[0030] The accompanying drawings, which form part of this embodiment, are used to provide a further understanding of this embodiment. The illustrative embodiments and their descriptions are used to explain this embodiment and do not constitute an improper limitation of this embodiment.

[0031] Figure 1 This is a flowchart of the series arc fault detection method in Embodiment 1 of the present invention;

[0032] Figure 2 This is a schematic diagram illustrating the specific process of the series arc fault detection method in Embodiment 1 of the present invention;

[0033] Figure 3 This is a diagram showing the normal and fault states of the current acquisition in Embodiment 1 of the present invention;

[0034] Figure 4 This is a data acquisition architecture diagram in Embodiment 1 of the present invention;

[0035] Figure 5 This is a time-domain feature map from Embodiment 1 of the present invention;

[0036] Figure 6 This is a time-domain feature judgment diagram in Embodiment 1 of the present invention;

[0037] Figure 7 This is a schematic diagram of the VMD decomposition results and selected IMF components in Embodiment 1 of the present invention;

[0038] Figure 8 This is a schematic diagram of the random forest algorithm in Embodiment 1 of the present invention;

[0039] Figure 9 This is a schematic diagram of the improved support vector machine in Embodiment 1 of the present invention;

[0040] Figure 10 This is a schematic diagram of cloud platform monitoring in Embodiment 1 of the present invention;

[0041] Figure 11 This is a structural block diagram of the series arc fault detection system in Embodiment 2 of the present invention. Detailed Implementation

[0042] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0043] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0044] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0045] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0046] Example 1

[0047] Embodiment 1 of the present invention introduces a method for detecting series arc faults.

[0048] like Figure 1 and Figure 2 The method for detecting a series arc fault shown includes:

[0049] Obtain the bus current signal;

[0050] Analyze the acquired current signal, calculate and determine the time-domain characteristics of the current signal when an arc fault occurs, and obtain the filtered time-domain characteristics.

[0051] The selected time-domain features are decomposed into energy to obtain the fuzzy entropy of the variational mode components. The feature evaluation weights are obtained by combining the obtained fuzzy entropy.

[0052] By using support vector machines and the obtained feature evaluation weights, arc faults can be diagnosed, thus realizing the fault detection of series arcs.

[0053] As one or more implementation methods, in the process of acquiring bus current signals, a current data acquisition device is used to collect the bus current data of each equipment line in the integrated energy system in real time, and record the location of each collected bus current; that is, a data acquisition device is installed on each equipment power line in the integrated energy system to collect the bus current, and the bus current is sequentially marked with the location of the line, such as L1, L2, L3..., Ln, and each data is stored sequentially; when an arc fault occurs, the current changes as follows: Figure 3 As shown, the data acquisition architecture is as follows Figure 4 As shown.

[0054] As one or more implementation methods, during the analysis of the acquired current signal, calculations and storage are performed as follows: Figure 5 The time-domain characteristics of the current signal shown include peak-to-peak value, variance, standard deviation, mean, RMS value, kurtosis, waveform index, peak factor, impulse factor, and skewness.

[0055] The formulas for calculating the eigenvalues ​​in each time domain are as follows:

[0056] Peak-to-peak value: X P =max(|x i |)

[0057] variance:

[0058] Standard deviation:

[0059] Mean:

[0060] Valid values:

[0061] kurtosis:

[0062] Waveform Indicators:

[0063] Peak factor:

[0064] Pulse factor:

[0065] Skewness:

[0066] Where, x i is the current value at point i, and N is the number of sampling points.

[0067] The system monitors the time-domain characteristics of the acquired current signal. When a time-domain characteristic exceeds a set threshold, it indicates a potential arc fault, and this time-domain characteristic is selected for fault detection. Specifically, all calculated time-domain characteristic values ​​are monitored. Each time-domain characteristic has a specific threshold. If the threshold is exceeded, it indicates a potential arc fault in the line's current, and this set of current data is immediately marked as warning mode and input to the next module for further analysis. If the threshold is not exceeded, it indicates that the current data is operating normally without fluctuations and no arc fault has occurred, and it is marked as safe mode. In particular, if any set of time-domain characteristic values ​​exceeds the threshold, the line current is set to warning mode, requiring further judgment. A judgment diagram for a certain time-domain characteristic is shown below. Figure 6 As shown.

[0068] As one or more implementation methods, the following are adopted: Figure 8The Undead algorithm shown optimizes variational mode decomposition (VMD) of the current to obtain variational mode components (IMFs). It then calculates the fuzzy entropy of each IMF component, selects the three IMFs with the highest fuzzy entropy, and merges and reconstructs the signal into new signal components. Figure 7 As shown in Table 1, the time-domain feature calculation process of step two is repeated for the new current signal. The Relief-F feature selection algorithm is used to evaluate the weights of the time-domain features. Features with a weight value greater than 0.1 are retained, and the filtered feature calculation results are output.

[0069] Table 1. Weights of each feature

[0070]

[0071] Variational Mode Decomposition (VMD) is used to decompose current signals. However, the selection of K and α significantly affects the decomposition effect, and these parameters are often set manually based on experience. In practical applications of arc fault signals, excessively large K values ​​lead to over-decomposition, while insufficient K values ​​result in inadequate decomposition and mode aliasing. α represents the initial center constraint strength of each mode. Therefore, it is necessary to select appropriate K and α for VMD decomposition to accurately reflect arc fault information. This application uses the Australian Wild Dog algorithm to optimize the K and α parameters of VMD, and sets the optimal result as the default parameter for signal decomposition.

[0072] The principle of VMD decomposition is defined as follows:

[0073] (1) Constructing variational problems

[0074] Assuming the original signal can be decomposed into K IMF components, the formula for the k-th IMF component is as follows: u k (t)=A k (t)cos[φ k [(t)], k∈{1,…,K};where, phase φ k (t) is a non-decreasing function, and φ k A'(t)≥0; k (t) represents the envelope function, and A k (t)≥0.

[0075] The solution to the constrained variational problem is shown in the following equation: Where t is time; j1 is the imaginary unit; ω k Let t be the center frequency of the k-th IMF component, k = 1, 2, ..., K, where K is the maximum number of decompositions; δ(t) is the impulse function.

[0076] (2) Solving variational problems

[0077] Introducing α and λ(t) transforms the variational problem into an unconstrained variational problem, i.e.:

[0078]

[0079] Where λ(t) is the Lagrange multiplier; α is the penalty factor; and f(t) is the original signal to be decomposed.

[0080] Solving for the saddle point of an unconstrained variational problem is equivalent to solving the optimal value of a constrained variational problem. The specific steps are as follows:

[0081] ① n is initialized;

[0082] ② Execute the loop, n = n + 1;

[0083] ③ For all ω>0, update

[0084] Get

[0085] ④ Update ω k ;

[0086] Get

[0087] ⑤ Update λ;

[0088] Get

[0089] Repeat steps ② to ⑤ above, and stop iterating when the following condition is met. Where ε is the convergence accuracy, and ε>0.

[0090] The dingo algorithm is defined as follows:

[0091] First, the population is initialized; that is... Among them, lb i and ub i Each represents an individual The lower and upper bounds of rand i It is a random number uniformly generated between [0,1].

[0092] (1) Group Attack

[0093] Wild dogs hunt large prey in packs, locating and surrounding the prey; this is defined as: in, This is the new location of the search agent, where na is a random integer generated in reverse order of [2, SizePop / 2], and SizePop is the size of the wild dog population. It is a subset of the search agent (the wild dog that will attack), where X is a randomly generated wild dog population. It is the current search agent. It is the best search agent found in the previous iteration, and β1 is a uniformly generated random number in [-2,2].

[0094] (2) Persecution

[0095] Wild dogs typically hunt small prey until they catch it alone, defined as: in, This is the new location for search agents. It is the current search agent. It is the best search agent found in the previous iteration, and β2 is a uniformly generated random number in the range [-1, 1]. It is a random individual within a wild dog population.

[0096] (3) Scavenger

[0097] The behavior of wild dogs randomly wandering through their habitat and finding carrion to eat is defined as: in, This is the new location for search agents. This is the current search agent, and β2 is a uniformly generated random number in the range [-1, 1]. σ is a random individual from the wild dog population, and σ is a binary number randomly generated in step 2, where σ∈{0,1}.

[0098] (4) Dingoes' Survival Rates

[0099] After hunting, wild dogs face survival challenges. Each wild dog's survival probability is related to its fitness value. Individuals with lower survival probabilities need to return to the vicinity of the current best individual to forage for food in order to improve their survival probability. The survival rate of a wild dog is defined as: Among them, fitness max and fitness min These are the worst and best fitness values ​​in the current generation, respectively, while fitness(i) is the current fitness value of the i-th search agent. The survival vector in the above formula includes the normalized fitness in the interval [0,1]. For example, if its survival probability is less than 0.3, the individual uses the following formula to calculate the new position: in, It is the current search agent. It is the best search agent found in the last iteration. and Let σ be two distinct random individuals in the population, where σ∈{0,1}.

[0100] Fuzzy entropy is defined as:

[0101] (1) For a given time series of length N {x(i), i = 1, 2, ..., N}, initialize the embedding dimension m, and define the phase space reconstruction of the above time series as: X(i) = {x(i), x(i+1), ..., x(i+m-1)} - u(i); where X(i) is the new time series after reconstruction, i = 1, 2, ..., N-m+1; u(i) is the mean of m consecutive x(i), and its expression is:

[0102] (2) Define the distance between two vectors X(i) and X(j) as: Where 1≤i,j≤N-m+1, and i≠j.

[0103] (3) Introduce the fuzzy membership function to define the similarity between vectors X(i) and X(j) as:

[0104] (4) Define function for The relational dimension under the m-dimensional dimension can be obtained, that is

[0105] (5) Increase the embedding dimension by 1, and then repeat steps (1) to (4) above for the m+1 dimension vector to obtain the relation dimension under the m+1 dimension, i.e.

[0106] (6) Finally, the expression for the fuzzy entropy can be obtained, namely FuzzyEn(m,r,N)=lnΦ m (r)-lnΦ m+1 (r); where m is the embedding dimension parameter, which is taken as m = 2 in this embodiment; r is the similarity tolerance parameter, which is taken as r = 0.2std in this embodiment, where std is the standard deviation of the original signal; and N is the length of the original time series.

[0107] The Relief-F algorithm is defined as follows:

[0108] Assume the dataset is D F The dataset contains |y| categories. For example x... F If it belongs to the kth F Class (k) F If ∈{1,2,…,|y|}, then the Relief-F algorithm first... F Find x in the sample of class F nearest neighbor x F,l,nh(l=1,2,…,|y|;1≠k F ), as sample x F If the wrong nearest neighbor is guessed, then the correlation statistic corresponds to attribute j. F The components are: Where p1 represents the first class of samples in dataset D F The proportion it accounts for.

[0109] As one or more implementation methods, a comprehensive time-domain feature is constructed based on the selected time-domain features according to their feature weights. The principle of weight value × feature is adopted, and each selected time-domain feature × weight is added in turn to construct the comprehensive time-domain feature. Finally, the selected time-domain features and the comprehensive time-domain features are output as the time-domain feature arc fault dataset and the comprehensive feature arc fault dataset, respectively.

[0110] That is, based on the o time-domain features with a weight greater than 0.1, a comprehensive time-domain feature is constructed. Using the weight × feature principle, the selected time-domain features (T) are then combined. f (f = 1, 2, ..., o) × weight (σ) f (f = 1, 2, ..., o) are added sequentially to construct a comprehensive time-domain feature (the comprehensive time-domain feature G is defined as: G = T1 × σ1 + T2 × σ2 + ... + T f ×σ f (f=1,2,…,o)) and finally output the time-domain feature arc fault dataset and the comprehensive feature arc fault dataset respectively.

[0111] As one or more implementation methods, the following are adopted: Figure 8 The random forest algorithm and the improved support vector machine algorithm shown are used for fault diagnosis. Based on the stored dataset of 100 sets of normal state data, labeled "0", 100 sets of arc fault data and 100 sets of arc fault data of comprehensive feature data are used, labeled "1". The datasets of normal state data and the datasets of arc fault data of comprehensive feature data are used as input to the random forest diagnostic model, and the datasets of normal state data and the datasets of arc fault data of comprehensive feature data of comprehensive feature data are used as input to the improved support vector machine model. The prediction results of the random forest and the prediction results of the improved support vector machine are output respectively.

[0112] Support Vector Machine (SVM) is defined as follows: y i (ω T ω(x i )+b≥1-ζ i ), ζ i ≥0i=1,2,...,l; where ω is the weight vector, C is the penalty factor, ζ is the slack variable, and x i y represents the vector samples input into the training set of the SVM classifier.i For classification labeling, ω(x) i ) is a function that maps training samples from a low-dimensional space to a high-dimensional space.

[0113] Support Vector Machines (SVMs) map data samples from the original feature space to a reconstructed high-dimensional feature space, construct a classification hyperplane in this feature space, and achieve good classification results by finding the optimal solution for the penalty factor C and the weight vector ω. This embodiment employs the following... Figure 9 The DOA algorithm shown optimizes these two parameters to construct an improved support vector machine model. The optimization steps for VMD and SVM in the DOA algorithm are the same. First, the number of iterations in the DOA algorithm is set to 50, and the population size to 30. The optimization range for VMD parameters is set to K∈[5,10], α∈[500,4000], while the optimization range for SVM parameters is set to c=[0,4], ω=[0.4,2]. The fitness function for VMD parameters is the minimum envelope entropy of each IMF component decomposed from its components, and the fitness function for SVM is the highest accuracy among different optimization values.

[0114] As one or more implementation methods, the two sets of prediction results output by the fault diagnosis module are again classified using an improved support vector machine model to output the final secondary arc fault classification result, thereby achieving high-precision arc fault diagnosis.

[0115] Specifically, the two sets of prediction results are then reclassified using an improved support vector machine model to output the final secondary arc fault classification result, achieving high-precision arc fault diagnosis. If the final SVM classification result is "1", it proves that an arc fault has indeed occurred on the line, and an alarm is immediately triggered by feedback from the cloud platform monitoring system. If the final SVM classification result is "0", it proves that the current fluctuation may be caused by factors such as the line being connected to a load, and no arc fault has actually occurred. In this case, feedback is sent to the cloud platform monitoring system, and the line status is changed from warning to safe.

[0116] It should be noted that the improved support vector machine here is... Figure 9 The results are consistent, with the only difference being the input and output values. Here, the input to the primary classification SVM model is the selected fault features. The primary SVM model outputs its classification result, and the random forest model also outputs its classification result. The classification results of these two models are used as new input values ​​for the secondary classification SVM to obtain the final output result.

[0117] This embodiment also includes, for example Figure 10The cloud platform monitoring shown can obtain the processing information of each module and the current operation of each line in the integrated energy system in real time. When an arc fault is suspected, a warning mode will be prompted. If the line is operating normally and no arc fault occurs, the normal status will be displayed. If an arc fault is indeed found after fault diagnosis, an alarm will be immediately triggered. If no arc fault occurs, the warning status will be changed to a safe status.

[0118] This embodiment can accurately diagnose series arc faults occurring in integrated energy systems, improving the safety of integrated energy systems and providing a new approach for the accurate diagnosis of arc faults in future integrated energy systems. By using a cloud platform for real-time monitoring and alarm prompts, it can clearly indicate which line has an arc fault, saving manpower and resources and effectively ensuring the stability of power system lines.

[0119] Example 2

[0120] Embodiment 2 of the present invention introduces a series arc fault detection system.

[0121] like Figure 11 The series arc fault detection system shown includes:

[0122] The acquisition module is configured to acquire the bus current signal;

[0123] The calculation module is configured to analyze the acquired current signal, calculate and determine the time-domain features of the current signal when an arc fault occurs, and obtain the filtered time-domain features; perform energy decomposition on the filtered time-domain features to obtain the fuzzy entropy of the variational mode components, and combine the obtained fuzzy entropy to obtain the feature evaluation weights.

[0124] The detection module is configured to diagnose arc faults based on support vector machines and the obtained feature evaluation weights, thereby realizing the fault detection of series arcs.

[0125] The detailed steps are the same as those provided in Example 1 for detecting series arc faults, and will not be repeated here.

[0126] Example 3

[0127] Embodiment 3 of the present invention provides a computer-readable storage medium.

[0128] A computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the series arc fault detection method as described in Embodiment 1 of the present invention.

[0129] The detailed steps are the same as those provided in Example 1 for detecting series arc faults, and will not be repeated here.

[0130] Example 4

[0131] Embodiment 4 of the present invention provides an electronic device.

[0132] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the series arc fault detection method as described in Embodiment 1 of the present invention.

[0133] The detailed steps are the same as those provided in Example 1 for detecting series arc faults, and will not be repeated here.

[0134] The above description is merely a preferred embodiment of this practice and is not intended to limit the scope of this practice. Various modifications and variations can be made to this practice by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this practice should be included within the protection scope of this practice.

Claims

1. A method for detecting series arc faults, characterized in that, include: Obtain the bus current signal; Analyze the acquired current signal, calculate and determine the time-domain characteristics of the current signal when an arc fault occurs, and obtain the filtered time-domain characteristics. The selected time-domain features are decomposed into energy to obtain the fuzzy entropy of the variational mode components. The feature evaluation weights are obtained by combining the obtained fuzzy entropy. The system diagnoses arc faults based on support vector machines and the obtained feature evaluation weights, thereby achieving fault detection of series arcs. After filtering, the time-domain features are decomposed to obtain the fuzzy entropy of each variational mode component. The three variational mode components with the largest fuzzy entropy are selected, and the current signals are merged and reconstructed to obtain new signal components. The current time-domain features of the new signal components are calculated, and the feature evaluation weights of the current time-domain features of the new signal components are calculated using the Relief-F feature selection algorithm.

2. The method for detecting series arc faults as described in claim 1, characterized in that, During the analysis of the acquired current signal, the time-domain characteristics of the current signal are calculated and stored; the time-domain characteristics include peak-to-peak value, variance, standard deviation, mean, RMS value, kurtosis, waveform index, peak factor, impulse factor, and skewness.

3. The method for detecting series arc faults as described in claim 2, characterized in that, The time-domain characteristics of the obtained current signal are monitored. When the time-domain characteristics exceed the set threshold, it is determined that there may be an arc fault. The time-domain characteristics are then selected for fault detection.

4. The method for detecting series arc faults as described in claim 1, characterized in that, In the process of obtaining the fuzzy entropy of the variational mode components, an improved variational mode decomposition is adopted, namely, the Australian Wild Dog algorithm is used to optimize the variational mode decomposition.

5. The method for detecting series arc faults as described in claim 1, characterized in that, In the process of diagnosing arc faults based on support vector machines and the obtained feature evaluation weights, a comprehensive time-domain feature is constructed based on the screened time-domain features and feature evaluation weights. The constructed comprehensive time-domain feature and the preset fault diagnosis model are combined to perform arc fault diagnosis. The obtained fault detection results are then used for secondary fault diagnosis through an improved support vector machine to realize the fault detection of series arcs.

6. The method for detecting series arc faults as described in claim 5, characterized in that, The fault diagnosis model employs a random forest algorithm and an improved support vector machine; the improved support vector machine uses the Australian Wild Dogs algorithm to optimize the penalty factor and weight vector.

7. A series arc fault detection system, characterized in that, include: The acquisition module is configured to acquire the bus current signal; The calculation module is configured to analyze the acquired current signal, calculate and determine the time-domain characteristics of the current signal when an arc fault occurs, and obtain the filtered time-domain characteristics. The selected time-domain features are decomposed into energy to obtain the fuzzy entropy of the variational mode components. The feature evaluation weights are obtained by combining the obtained fuzzy entropy. After filtering, the time-domain features are decomposed to obtain the fuzzy entropy of each variational mode component. The three variational mode components with the largest fuzzy entropy are selected, and the current signals are merged and reconstructed to obtain new signal components. The current time-domain features of the new signal components are calculated, and the feature evaluation weights of the current time-domain features of the new signal components are calculated using the Relief-F feature selection algorithm. The detection module is configured to diagnose arc faults based on support vector machines and the obtained feature evaluation weights, thereby realizing the fault detection of series arcs.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the series arc fault detection method as described in any one of claims 1-6.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the series arc fault detection method as described in any one of claims 1-6.

Images