Arc fault detection method and related apparatus

Abstract

The application discloses an arc fault detection method and related device, the method comprises the following steps: determining the first statistical characteristics of the current signal in the time window; waveform mutation analysis is performed on the current signal to obtain the change interval with the largest fluctuation amplitude in the current signal; at least one time domain feature, at least one frequency domain feature and a plurality of second statistical characteristics of the current signal are extracted; whether there is a transient event is determined according to the first statistical characteristics, real-time current data and the change interval; if there is a transient event, a target fault detection model is determined according to at least one time domain feature and the change interval; the at least one time domain feature, the at least one frequency domain feature and the plurality of second statistical characteristics are subjected to feature screening processing and feature fusion processing to obtain target features; and a fault detection result is determined according to the target features and the target fault detection model. The application can improve the efficiency and accuracy of fault detection of the system.

Description

Technical Field

[0001] This application relates to the field of power system safety technology, and in particular to an arc fault detection method and related device. Background Technology

[0002] Electrical equipment generates electric arcs in circuits during operation. Electric arcs are the most intense form of self-sustaining gas discharge, characterized by high temperature, low current, and short duration. Electric arcs include normal arcs and fault arcs. Fault arcs generate high temperatures and energy, capable of igniting surrounding flammable and explosive materials, causing fires or even explosions. Therefore, it is crucial to detect and eliminate arc faults promptly and accurately before they cause fires. However, traditional arc fault detection technologies primarily rely on the time-frequency characteristics of current and voltage. When faced with complex loads and harmonic interference, traditional methods often exhibit limitations, prone to false alarms and missed alarms, resulting in low accuracy and poor reliability in arc fault detection. Summary of the Invention

[0003] This application provides an arc fault detection method and related apparatus to improve the efficiency and accuracy of system fault detection.

[0004] In a first aspect, embodiments of this application provide an arc fault detection method, including:

[0005] A first statistical feature of the current signal within a time window is determined, wherein the current signal includes an even number of sampling points, and the first statistical feature is used to characterize the local offset level and the overall dispersion of the current signal;

[0006] Waveform abrupt change analysis is performed on the current signal to obtain the range of change with the largest fluctuation amplitude in the current signal;

[0007] At least one time-domain feature, at least one frequency-domain feature, and multiple second statistical features are extracted from the current signal, wherein the multiple second statistical features are used to characterize the fluctuation amplitude and shape deviation of the current signal;

[0008] Based on the first statistical feature, real-time current data, and the change range, determine whether a transient event exists;

[0009] If the transient event exists, the target fault detection model is determined based on the at least one time-domain feature and the change interval;

[0010] The target feature is obtained by performing feature filtering and feature fusion processing on the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features;

[0011] Based on the target features and the target fault detection model, the fault detection result is determined.

[0012] The first statistical feature of the current signal within the determined time window includes:

[0013] Determine the variance of the current signal within the time window;

[0014] The time window is divided into a first sub-time window and a second sub-time window. The number of the first sampling point is the same as the number of the second sampling point. The number of the first sampling point is the number of sampling points of the current signal in the first sub-time window, and the number of the second sampling point is the number of sampling points of the current signal in the second sub-time window.

[0015] Determine a first mean value of the current signal within the first sub-time window; and determine a second mean value of the current signal within the second sub-time window;

[0016] The first statistical feature is obtained based on the first mean, the second mean, and the variance.

[0017] The step of determining whether a transient event exists based on the first statistical feature, real-time current data, and the change range includes:

[0018] The change in the current signal is determined based on the first mean, the second mean, the variance, and the real-time current data.

[0019] If the detected change amount is greater than a preset change amount, then the duration of the change interval is determined;

[0020] The existence of transient events is determined based on the duration of the interval.

[0021] The step of determining the change in the current signal based on the first mean, the second mean, the variance, and the real-time current data includes:

[0022] Determine a first difference between the first mean and the second mean, wherein the first difference is a positive number;

[0023] Determine the ratio of the first difference to the variance;

[0024] The first mean and the second mean are averaged to obtain the third mean;

[0025] Determine a second difference between the real-time current data and the third mean, wherein the second difference is a positive number;

[0026] The product of the second difference and the ratio value is determined to obtain the amount of change.

[0027] The step of performing feature filtering and feature fusion processing on the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features to obtain the target feature includes:

[0028] Determine the first load type corresponding to the current signal;

[0029] Obtain the monitoring requirements for the first load type;

[0030] Create a first feature subset, which includes at least the at least one time-domain feature, the at least one frequency-domain feature, and any two different dimensions of the plurality of second statistical features;

[0031] The first feature subset is subjected to feature expansion processing to obtain a second feature subset. The feature expansion processing is used to extend a single other feature into the current feature subset. The other feature is a feature other than the first feature subset among the at least one time-domain feature, the at least one frequency-domain feature and the plurality of second statistical features.

[0032] Determine the first relevance between the second feature subset and the monitoring requirement;

[0033] Based on the third difference between the first relevance and the preset relevance, the feature processing method of the second feature subset is determined. The feature processing method includes the feature expansion processing and / or feature removal processing. The feature removal processing is used to remove a single feature from the current feature subset.

[0034] The second feature subset is processed according to the aforementioned feature processing method to obtain the third feature subset;

[0035] Determine the second relevance between the third feature subset and the monitoring requirement;

[0036] If it is detected that the number of features in the third feature subset is a preset number, the second relevance is greater than or equal to the preset relevance, and each of the other features has been processed by the feature expansion process and / or the feature removal process, then each feature in the third feature subset is fused to obtain the target feature.

[0037] The step of determining the feature processing method for the second feature subset based on the third difference between the first relevance and the preset relevance includes:

[0038] If the third difference is within the first preset range, then the feature processing method is the feature expansion processing;

[0039] If the third difference is within the second preset range, then the feature processing method is the feature removal processing, where the upper limit of the first preset range is less than the lower limit of the second preset range.

[0040] If the third difference is within the third preset range, then the feature processing method is the feature removal processing and the feature expansion processing, and the upper limit of the second preset range is less than the lower limit of the third preset range.

[0041] The method further includes:

[0042] If the transient event does not exist, then obtain the second load type triggered by the preceding transient event;

[0043] Determine the target fault detection model based on the second load type;

[0044] The target feature is obtained by performing feature filtering and feature fusion processing on the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features;

[0045] Based on the target features and the target fault detection model, the fault detection result is determined.

[0046] Secondly, embodiments of this application provide an arc fault detection device, comprising:

[0047] The first determining unit is used to determine the first statistical characteristics of the current signal within a time window. The current signal includes an even number of sampling points. The first statistical characteristics are used to characterize the local offset level and the overall dispersion of the current signal.

[0048] The analysis unit is used to perform waveform abrupt change analysis on the current signal to obtain the range of change with the largest fluctuation amplitude in the current signal.

[0049] An extraction unit is used to extract at least one time-domain feature, at least one frequency-domain feature, and multiple second statistical features of the current signal, wherein the multiple second statistical features are used to characterize the fluctuation amplitude and shape deviation of the current signal;

[0050] The second determining unit is used to determine whether a transient event exists based on the first statistical feature, real-time current data, and the change range.

[0051] A matching unit is configured to determine a target fault detection model based on at least one time-domain feature and the change interval if the transient event exists.

[0052] The feature processing unit is used to perform feature filtering and feature fusion processing on the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features to obtain the target feature;

[0053] The third determining unit is used to determine the fault detection result based on the target features and the target fault detection model.

[0054] Thirdly, embodiments of this application provide an electronic device, including a memory, a processor, and executable program code stored in the memory and executable on the processor, wherein the processor executes the executable program code and performs the steps of the method described in the first aspect.

[0055] Fourthly, embodiments of this application provide a computer-readable storage medium storing executable program code, the executable program code including execution instructions for performing the steps of the method as described in the first aspect.

[0056] Fifthly, embodiments of this application provide a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps described in the first aspect of embodiments of this application. The computer program product may be a software installation package.

[0057] As can be seen, in this embodiment, firstly, a first statistical feature of the current signal within a time window is determined. The current signal includes an even number of sampling points. The first statistical feature is used to characterize the local offset level and overall dispersion of the current signal. Then, waveform mutation analysis is performed on the current signal to obtain the variation range with the largest fluctuation amplitude in the current signal. Next, at least one time-domain feature, at least one frequency-domain feature, and multiple second statistical features of the current signal are extracted. The multiple second statistical features are used to characterize the fluctuation amplitude and shape deviation of the current signal. Then, based on the first statistical feature, real-time current data, and the variation range, it is determined whether a transient event exists. If the transient event exists, a target fault detection model is determined based on the at least one time-domain feature and the variation range. Then, feature filtering and feature fusion processing are performed on the at least one time-domain feature, the at least one frequency-domain feature, and the multiple second statistical features to obtain the target feature. Finally, based on the target feature and the target fault detection model, the fault detection result is determined.

[0058] This application uses multi-dimensional data to comprehensively determine transient events, which helps to eliminate false events and improve the accuracy of the determination. Then, the transient event triggers the matching of the fault detection model. During feature processing, time domain, frequency domain features and statistical features are screened and combined to eliminate redundant features. Finally, the target features obtained after feature processing and the matched target fault detection model are combined to perform fault detection. This makes the fault detection results more consistent with the actual load situation, thereby improving the pertinence and accuracy of fault detection. It avoids the problem of misjudgment or missed judgment that may be caused by load characteristic mismatch, and improves the efficiency and accuracy of power system fault detection. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art 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.

[0060] Figure 1 This application provides a system architecture diagram of a fault detection system.

[0061] Figure 2 This is a flowchart illustrating the first arc fault detection method provided in the embodiments of this application;

[0062] Figure 3 This is a flowchart illustrating the process of determining whether a transient event exists, provided in an embodiment of this application.

[0063] Figure 4 This is a flowchart illustrating the second arc fault detection method provided in this application embodiment;

[0064] Figure 5 This is a flowchart illustrating the third arc fault detection method provided in the embodiments of this application;

[0065] Figure 6 This is a functional unit block diagram of an arc fault detection device provided in an embodiment of this application;

[0066] Figure 7 This is a functional unit block diagram of another arc fault detection device provided in the embodiments of this application;

[0067] Figure 8 This is a schematic diagram of the structure of an electronic device proposed in an embodiment of this application. Detailed Implementation

[0068] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0069] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0070] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0071] Electrical equipment generates electric arcs in circuits during operation. Electric arcs are the most intense form of self-sustaining gas discharge, characterized by high temperature, low current, and short duration. Electric arcs include normal arcs and fault arcs. Fault arcs generate high temperatures and energy, capable of igniting surrounding flammable and explosive materials, causing fires or even explosions. Therefore, it is crucial to detect and eliminate arc faults promptly and accurately before they cause fires. However, traditional arc fault detection technologies primarily rely on the time-frequency characteristics of current and voltage. When faced with complex loads and harmonic interference, traditional methods often exhibit limitations, prone to false alarms and missed alarms, resulting in low accuracy and poor reliability in arc fault detection.

[0072] In addition, arc fault signals are often accompanied by noise interference, and traditional methods perform poorly in low signal-to-noise ratio environments, affecting the accuracy of arc fault detection.

[0073] To address at least one of the aforementioned problems, embodiments of this application provide an arc fault detection method and related apparatus. The embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0074] Please see Figure 1 , Figure 1This is a system architecture diagram of a fault detection system provided in an embodiment of this application. Figure 1 As shown, the fault detection system 100 includes a discrimination module 101, a feature extraction module 102, and a fault detection module 103. The discrimination module 101, the feature extraction module 102, and the fault detection module 103 are interconnected in pairs.

[0075] The discrimination module 101 is used to determine whether a transient event has occurred in the current sliding window. It analyzes the current signal in the sliding window segment by segment, and determines the information of sudden changes or abnormal changes in the current signal by calculating the mean and variance of the current signal in each window. At the same time, it deletes false events according to the duration of abnormal changes to ensure that false alarms are eliminated, and finally determines whether a transient event has occurred.

[0076] The feature extraction module 102 is used to extract the time-domain features, frequency-domain features and statistical features of the current signal, and match the corresponding feature evaluation index according to the load type determined by the fault detection module 103. The extracted features are then subjected to feature screening and feature fusion processing based on the feature evaluation index to obtain the target features.

[0077] The fault detection module 103 receives information from the discrimination module 101 and the feature extraction module 102. When a transient event is detected, it triggers load type identification, determines the load type based on the duration of the abnormal change, the time range of the abnormal change, and transient characteristics, determines the corresponding fault detection model based on the load type, inputs the target features into the fault detection model, and outputs the fault detection result. Conversely, when no transient event occurs, the load type triggered by the previous transient event is determined as the current load type, and fault detection is performed based on the fault detection model corresponding to that load type.

[0078] Based on this, this application provides an arc fault detection method and related apparatus, which will be described in detail below with reference to the accompanying drawings.

[0079] Please see Figure 2 , Figure 2 This is a flowchart illustrating an arc fault detection method provided in an embodiment of this application, as shown below. Figure 2 As shown, the method includes the following steps:

[0080] S210, determine the first statistical characteristic of the current signal within the time window.

[0081] The current signal includes an even number of sampling points, and the first statistical feature is used to characterize the local offset level and overall dispersion of the current signal.

[0082] The system captures voltage and current signals in real time, and the time window is sliding, advancing a preset step size each time, for example, advancing one step size each time.

[0083] Among them, current signal refers to the continuous or discrete waveform data of the entire current changing over time.

[0084] Each time window includes 2n current sampling points. The current signal within the time window is analyzed segment by segment. The mean of the current signal within each segment is calculated, as well as the variance of the current signal within the entire segment. The current signal sample in the time window is obtained. The transient changes of the current waveform are determined based on the current signal sample, mean, and variance.

[0085] The current signal sample is the current value being monitored in real time, and it is the latest data point in the sliding window.

[0086] In one possible embodiment, determining the first statistical feature of the current signal within a time window includes: determining the variance of the current signal within the time window; splitting the time window into a first sub-time window and a second sub-time window, wherein the number of first sampling points is the same as the number of second sampling points, the first number of sampling points being the number of sampling points of the current signal within the first sub-time window, and the second number of sampling points being the number of sampling points of the current signal within the second sub-time window; determining a first mean of the current signal within the first sub-time window; and determining a second mean of the current signal within the second sub-time window; and obtaining the first statistical feature based on the first mean, the second mean, and the variance.

[0087] The first statistical feature includes the mean and variance of each sub-time window.

[0088] The time window can be divided into two segments, each sub-time window including n current sampling points. The mean value of the current signal in each sub-time window is calculated, that is, the mean value of the n current sampling points in the first half and the mean value of the n current sampling points in the second half are calculated. The local offset level of the current signal is reflected by the first mean value and the second mean value.

[0089] In this process, the variance of 2n current sampling points is calculated simultaneously, and this variance reflects the overall dispersion of the current signal.

[0090] By using the mean and variance of each sub-time window, abrupt changes or abnormal variations in the current signal can be detected.

[0091] In one possible embodiment, the length of the time window can be adjusted according to the variation characteristics of the actual current signal to balance detection accuracy and computational efficiency.

[0092] In one possible embodiment, an automatic adaptive algorithm can be used to analyze the actual operating conditions of the power grid in real time, and then dynamically adjust the sampling frequency and sampling period to improve detection accuracy and reduce computational burden.

[0093] Specifically, the sampling frequency can be dynamically adjusted based on the grid load and arc fault characteristics. When the grid load is low and the current change is stable, the sampling frequency can be reduced; when the grid load is high or the probability of detecting an arc fault increases, the sampling frequency can be increased to ensure that the high-frequency components of the arc fault are captured in a timely manner.

[0094] Specifically, the sampling period can be dynamically adjusted based on the rate of change of the current waveform. When the current fluctuation is large, the sampling period is shortened to improve real-time performance; when the current fluctuation is small, the sampling period is extended to reduce data redundancy, reduce storage and computational pressure, and avoid interference from invalid data in the analysis process.

[0095] In one possible embodiment, the number of sampling points is dynamically adjusted based on the adjustment of the time window length, and / or the adjustment of the sampling frequency, and / or the adjustment of the sampling period.

[0096] The sub-time windows are redefined based on the adjusted number of sampling points.

[0097] For example, a first threshold and a second threshold are preset, and the second threshold is greater than the first threshold. If the number of additional sampling points is greater than the first threshold, then when splitting the time window, an additional sub-time window is added, that is, the time window is evenly divided into three sub-time windows; if the number of additional sampling points is greater than the second threshold, then when splitting the time window, two additional sub-time windows are added, that is, the time window is evenly divided into four sub-time windows.

[0098] S220, Perform waveform abrupt change analysis on the current signal to obtain the range of change with the largest fluctuation amplitude in the current signal.

[0099] Among them, the time period in which the waveform of the current signal changes drastically is identified, and the interval with the most significant fluctuation amplitude is selected to obtain the change interval.

[0100] The process involves preprocessing the current signal, including filtering and normalization, and extracting abrupt change features such as instantaneous change and rate of change. The intensity of the abrupt change is then determined based on these features. Instantaneous change refers to the difference in current between adjacent sampling points; a larger difference indicates a more pronounced abrupt change. The rate of change refers to the amount of current change per unit time, reflecting the speed of the abrupt change.

[0101] The continuous current signal is divided into several intervals by using a sliding window, the intensity of the sudden change in each interval is calculated, and finally the interval with the highest intensity of sudden change, that is, the interval with the largest fluctuation amplitude, is selected.

[0102] S230, extract at least one time-domain feature, at least one frequency-domain feature, and multiple second statistical features of the current signal.

[0103] The plurality of second statistical features are used to characterize the fluctuation amplitude and shape deviation of the current signal.

[0104] Among them, time-domain features are features extracted from the time series of current signals based on the time dimension, used to reflect the instantaneous changes or overall trends of the signal over time. These can include peak value, mean value, root mean square value, kurtosis, etc.

[0105] Among them, frequency domain features are features based on the frequency dimension. They are used to convert time-domain signals to the frequency domain through methods such as Fourier transform and wavelet transform, and are used to reflect the frequency components and energy distribution contained in the signal.

[0106] When an arc fault occurs, specific high-frequency noise components appear in the frequency domain. The current signal can be converted from the time domain to the frequency domain by using a fast Fourier transform, which can effectively extract the high-frequency components in the arc fault signal.

[0107] Specifically, frequency domain characteristics can include the dominant frequency, which is the frequency component with the highest energy content in the signal; harmonic amplitude, which is the amplitude of higher harmonics other than the dominant frequency, used to reflect the degree of spectral pollution of the signal; spectral entropy, which is used to measure the uniformity of frequency component distribution. High spectral entropy indicates that the signal contains multiple frequencies, while low entropy indicates that the frequency components are singular, such as a stable sine wave; and bandwidth, which is used to characterize the distribution range of the main frequency components of the signal.

[0108] The second statistical characteristic may include variance, standard deviation, range, skewness, and kurtosis. Variance, standard deviation, and range are used to reflect the fluctuation amplitude of the current signal, while skewness and kurtosis are used to describe the distribution characteristics and shape deviation of the current signal.

[0109] Among them, by combining time-domain features, frequency-domain features, and second statistical features, fault signals and normal signals can be distinguished.

[0110] S240, based on the first statistical feature, real-time current data and the change range, determine whether a transient event exists.

[0111] In one possible embodiment, please refer to Figure 3 , Figure 3 This is a flowchart illustrating a method for determining the existence of a transient event, as provided in an embodiment of this application. Figure 3The steps for determining whether a transient event exists based on a first statistical characteristic, real-time current data, and the range of variation are shown below:

[0112] S310, determine the change in the current signal based on the first mean, the second mean, the variance, and the real-time current data.

[0113] The real-time current data consists of the current values ​​monitored in real time, representing the latest data within the sliding window. Based on the mean, variance, and real-time current data of the current signal, abrupt or abnormal changes in the current signal are detected.

[0114] In one possible embodiment, determining the change in the current signal based on the first mean, the second mean, the variance, and the real-time current data includes: determining a first difference between the first mean and the second mean, wherein the first difference is a positive number; determining a ratio between the first difference and the variance; performing a mean operation on the first mean and the second mean to obtain a third mean; determining a second difference between the real-time current data and the third mean, wherein the second difference is a positive number; and determining the product of the second difference and the ratio to obtain the change.

[0115] Among them, the change in current signal The formula is as follows:

[0116] ,

[0117] in, The first mean, The second mean, For variance, This is real-time current data.

[0118] In one possible embodiment, if the time window is divided into three sub-time windows, the mean of each sub-time window is calculated to obtain a first mean, a second mean, and a third mean, and the variance of the entire time window is calculated. Then, the differences between each pair of means are calculated to obtain a first difference, a second difference, and a third difference; all of these differences are positive. The mean of the three differences is calculated to obtain a fourth mean. The ratio of the fourth mean to the variance is calculated. Simultaneously, the mean of the first, second, and third means is calculated to obtain a fifth mean. The difference between the real-time current data and the fifth mean is calculated to obtain a fourth difference, which is positive. The product of the fourth difference and the ratio is calculated to obtain the change in the current signal.

[0119] S320, if the change is detected to be greater than a preset change, then the duration of the change interval is determined.

[0120] Specifically, when the change in the current signal exceeds a set threshold, the duration of the change interval is used to determine whether it is a false event. Events with too short an interval duration are judged as false events and are removed.

[0121] False events are signals that are mistakenly identified as transient events, but which do not actually have a corresponding physical process and may be caused by measurement noise, equipment errors, etc. The duration of false events may be extremely short or irregular.

[0122] S330, determine whether a transient event exists based on the duration of the interval.

[0123] Among these methods, events with an interval duration shorter than the minimum possible time of the physical process can be identified as false events; events with an interval duration that contradicts the physical process can be identified as false events, for example, by setting an upper limit based on the typical duration of a specific transient type, and if the interval duration exceeds this upper limit, it can be identified as a false event; events with irregular durations can also be identified as false events.

[0124] Transient events refer to sudden changes in current or voltage signals that occur within a short period of time. They are usually caused by real physical processes, such as switching operations, short-circuit faults, arc discharges, and sudden load changes.

[0125] If the event is detected to be a genuine event, then a transient event is determined to exist.

[0126] Transient events may be precursors to arc faults or sudden changes in current. By analyzing these events, the system can determine whether an arc fault has occurred.

[0127] S250, if the transient event exists, then the target fault detection model is determined based on the at least one time-domain feature and the change interval.

[0128] Specifically, transient features are extracted from at least one time-domain feature, the load type is determined based on the transient features and the change range, and the target fault detection model is determined based on the load type.

[0129] The load types include resistive loads, inductive loads, switchable loads, capacitive loads, and new energy loads.

[0130] Each load type exhibits unique transient characteristics in its current waveform. Resistive loads are characterized by short-term current fluctuations during transient events; inductive loads are characterized by large transient fluctuations in their current waveforms during startup, which last for a relatively long time; switchable loads are characterized by short-duration current pulses; capacitive loads are characterized by short-duration large current pulses; and new energy loads are characterized by small-amplitude, high-frequency fluctuations in current during startup, continuous and non-periodic changes in current amplitude during operation, and short-duration reverse current pulses when the energy storage system switches from charging to discharging mode.

[0131] Therefore, we can first determine the duration of the change interval, and then determine at least one candidate load type based on the duration of the interval; determine the time position of the change interval within the time window; and determine the load type from at least one candidate load type based on the time position and transient characteristics.

[0132] Among them, the electrical characteristics and fault modes of different load types are significantly different, so different load types correspond to different fault detection models.

[0133] The system receives characteristic data of the electric arc, analyzes the characteristics based on a trained fault detection model, and outputs results, such as determining whether the electric arc is a normal working arc, an abnormal arc caused by poor contact, or a short-circuit fault arc.

[0134] For example, for resistive loads, the model can use a temporal network with short-time sliding windows; for inductive loads, the model can use a deep network with long-time windows; for switchable loads, the model can use an adaptive network with hybrid temporal windows; for capacitive loads, the model can use a convolutional neural network with high-frequency convolutional kernels; and for renewable energy loads, the model can use a network that combines a sampling attention mechanism with a Transformer.

[0135] S260, perform feature filtering and feature fusion processing on the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features to obtain the target feature.

[0136] In one possible embodiment, the step of performing feature filtering and feature fusion processing on the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features to obtain target features includes: determining the first load type corresponding to the current signal; obtaining the monitoring requirements of the first load type; creating a first feature subset, the first feature subset including at least two different dimension features from the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features; performing feature expansion processing on the first feature subset to obtain a second feature subset, the feature expansion processing being used to expand a single other feature into the current feature subset, the other feature being a feature other than the first feature subset from the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features; determining... The second feature subset is determined to have a first correlation with the monitoring requirement; based on a third difference between the first correlation and a preset correlation, a feature processing method for the second feature subset is determined, the feature processing method including feature expansion processing and / or feature removal processing, the feature removal processing being used to remove a single feature from the current feature subset; the second feature subset is processed according to the feature processing method to obtain a third feature subset; a second correlation between the third feature subset and the monitoring requirement is determined; if it is detected that the number of features in the third feature subset is a preset number, and the second correlation is greater than or equal to the preset correlation, and each of the other features has been processed by the feature expansion processing and / or the feature removal processing, then each feature in the third feature subset is fused to obtain the target feature.

[0137] The current load type is determined based on transient characteristics and the range of change.

[0138] The monitoring requirements differ for different load types. For example, the monitoring requirements for resistive loads include identifying poor contact faults, warning of overload risks, and detecting short-circuit hazards; while the monitoring requirements for capacitive loads include detecting partial discharge faults and warning of overvoltage breakdown risks.

[0139] The first feature subset includes features in at least two dimensions, such as two time-domain features and a second statistical feature.

[0140] When calculating the correlation between a feature subset and monitoring requirements using statistical methods, such as the Pearson correlation coefficient, the monitoring requirements are transformed into evaluation indicators for the features. The evaluation value of the feature subset is determined based on the evaluation indicators, and the evaluation value is used to measure the degree of contribution of the selected feature subset to meeting the monitoring requirements, i.e., the correlation.

[0141] For example, the characteristic evaluation metrics for inductive loads may include metrics for evaluating the characteristic's ability to capture initiation spikes and metrics for evaluating the characteristic's ability to characterize harmonics. For example, the characteristic evaluation metrics for capacitive loads may include metrics for evaluating the characteristic's tracking accuracy during impact processes and metrics for evaluating the characteristic's ability to characterize stable operation.

[0142] The feature expansion process involves selecting one feature from other features to add to the current feature subset. The added feature subset is then evaluated based on the feature evaluation index to obtain an evaluation value. This evaluation value is then used to determine the relevance between the feature subset and the detection requirements, thus ensuring that the new subset after adding the feature has the optimal relevance to the monitoring requirements.

[0143] The feature expansion process is used to add two features from other features to the current feature subset. The added feature subset is evaluated according to the feature evaluation index to obtain the evaluation value. The evaluation value is determined as the relevance between the feature subset and the detection requirements, so that the new subset after adding features has the optimal relevance to the monitoring requirements.

[0144] Specifically, based on the feature evaluation index, the importance score of each feature in the feature subset is calculated and summed to obtain the evaluation value.

[0145] In one possible embodiment, determining the feature processing method for the second feature subset based on the third difference between the first relevance and the preset relevance includes: if the third difference is within a first preset range, the feature processing method is the feature expansion processing; if the third difference is within a second preset range, the feature processing method is the feature removal processing, where the upper limit of the first preset range is less than the lower limit of the second preset range; if the third difference is within a third preset range, the feature processing method is both the feature removal processing and the feature expansion processing, where the upper limit of the second preset range is less than the lower limit of the third preset range.

[0146] The feature removal process is used to remove a feature from the current feature subset. The removed feature subset is then evaluated based on the feature evaluation index to obtain an evaluation value. This evaluation value is then used to determine the relevance between the feature subset and the detection requirements, so that the new subset after feature removal has the optimal relevance to the monitoring requirements.

[0147] If the aforementioned feature expansion process selects two features to add to the current feature subset, the feature removal process can be used to remove one or two features from the current feature subset. The removed feature subset is then evaluated based on the feature evaluation index to obtain an evaluation value. This evaluation value is then determined as the relevance between the feature subset and the detection requirements, so that the new subset after feature removal has the optimal relevance to the monitoring requirements.

[0148] Among them, the maximum value of the first preset range is less than the minimum value of the second preset range; the maximum value of the second preset range is less than the minimum value of the third preset range, and the three do not overlap and are ordered in ascending order.

[0149] Specifically, when the third difference is within the first preset range, it indicates that the relevance between the feature subset and the monitoring requirements is close to the preset relevance. In this case, feature expansion processing is performed to add new relevant feature dimensions to compensate for the deficiencies of the current features, thereby optimizing the current relevance. It is also possible to backtrack and remove old features after adding new ones.

[0150] Specifically, when the third difference falls within the second preset range, it indicates that the relevance between the current feature subset and the monitoring requirements is somewhat different from the preset threshold, indicating feature redundancy. Feature removal is then performed to remove weakly relevant or irrelevant features, retaining core features to optimize the current relevance. Furthermore, even better features can be added after feature removal.

[0151] When the third difference is within the third preset range, it indicates that the correlation between the current feature subset and the monitoring requirements is much lower than the preset threshold. Feature expansion processing and feature removal processing can be performed simultaneously to make the current correlation optimal.

[0152] After obtaining the third feature subset, the second relevance between the third feature subset and the monitoring requirements is determined. If the number of features in the third feature subset is a preset number, the second relevance is greater than or equal to the preset relevance, and each other feature has been processed by feature expansion and / or feature removal, then each feature in the third feature subset is fused to obtain the target feature.

[0153] If the above conditions are not met simultaneously, the difference between the second relevance and the preset relevance is calculated. Based on the difference, the feature expansion processing and / or feature removal processing are repeatedly performed according to the method described above until the relevance between the feature subset and the monitoring requirements reaches the preset relevance, and the relevance cannot be improved by adding or removing a single feature, and the number of features in the feature subset reaches the preset number. Then, the obtained feature subset is subjected to feature fusion processing to obtain the target feature.

[0154] Feature fusion can be determined by weighted summation or by normalization and concatenation. For example, features of different magnitudes can be normalized to the same interval and then concatenated into a vector, preserving their respective information while eliminating magnitude differences to obtain the target features.

[0155] As can be seen, in this embodiment of the application, starting from the load monitoring requirements, an effective subset of features is constructed through feature engineering and its relevance to the monitoring requirements is quantified, providing targeted input features for subsequent model analysis.

[0156] In one possible embodiment, the method further includes: if the transient event does not exist, obtaining a second load type for the preceding transient event trigger identification; determining a target fault detection model based on the second load type; performing feature filtering and feature fusion processing on the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features to obtain target features; and determining a fault detection result based on the target features and the target fault detection model.

[0157] If no transient event is detected, the load type triggered by the previous transient event is determined as the current load type, and then the fault detection model is determined. Feature filtering and feature fusion processing are performed on the multi-dimensional features. Finally, the obtained target features are input into the target fault detection model, and the fault detection result is output.

[0158] S270, determine the fault detection result based on the target features and the target fault detection model.

[0159] The fault detection results may include fault type, fault occurrence time, fault characteristic parameters, fault level, and associated load information.

[0160] If an arc fault is detected, the system outputs an alarm message and executes corresponding isolation or control measures based on the load type and detection results, such as isolating the fault area or disconnecting the power supply.

[0161] Different response measures are determined based on the load type and fault detection results. For example, for resistive load faults, the power supply may be directly cut off; while for inductive load faults, a more refined control strategy may be required.

[0162] If no arc fault is detected, the system will continue to monitor the power grid status in real time and update the load classification information to ensure that the system is always in the best detection state.

[0163] As can be seen, in this embodiment, the comprehensive determination of transient events through multi-dimensional data helps to eliminate false events and improve the accuracy of the determination. Then, the fault detection model is matched based on the transient event. During feature processing, time domain, frequency domain features and statistical features are screened and combined to eliminate redundant features. Then, the target features obtained after feature processing and the target fault detection model obtained by matching are combined to perform fault detection. This makes the fault detection results more consistent with the actual load situation, thereby improving the pertinence and accuracy of fault detection, avoiding the problem of misjudgment or missed judgment that may be caused by load characteristic mismatch, and improving the efficiency and accuracy of system fault detection.

[0164] In one possible embodiment, please refer to Figure 4 , Figure 4 This is a flowchart illustrating the second arc fault detection method provided in this application embodiment, as shown below. Figure 4 As shown, the current signal in the distribution network is monitored in real time, and transient events are detected based on changes in the current signal. When an abnormality is detected in the current waveform, load classification is triggered.

[0165] Among them, the load classification uses an event-driven approach based on the transient event detection results to identify the load type, classifying the load into resistive loads, inductive loads, and switchable loads, ensuring that appropriate model algorithms are used when detecting arc faults.

[0166] Specifically, for different load types, arc fault features are extracted through feature extraction and feature selection, and the most distinctive features are selected for arc fault detection to obtain arc fault results.

[0167] If an arc fault is detected, the system outputs an alarm message and executes corresponding isolation or control measures based on the load type and detection results; if no fault is detected, monitoring continues and the next round of data collection and analysis is performed.

[0168] In one possible embodiment, please refer to Figure 5 , Figure 5 This is a flowchart illustrating the third arc fault detection method provided in this application embodiment, as shown below. Figure 5 As shown, the system captures current signals in real time within the acquisition window and monitors transient events based on these signals. Specifically, a sliding window analysis method is used to analyze the current signal segment by segment, detecting abrupt changes or abnormal variations. When an abnormal change is detected, false events are detected and deleted based on the duration of the abnormal change, ensuring that false alarms are eliminated, and the system continues to capture current signals for the next round of analysis.

[0169] If a transient event is determined, load type identification is performed based on the duration and transient characteristics of the event. For example, if the duration is within a certain time range and the transient characteristic is a large transient fluctuation in the current waveform at startup, the current load is determined to be an inductive load. An arc fault detector is matched to the inductive load, and load features are extracted and selected to obtain the most discriminative features. These features are then input into the matched arc fault detector for arc fault detection, and the arc fault detection result is output.

[0170] If no abnormal change is detected, the load type of the previous transient event is determined as the current load type. Then, the arc fault detector is matched, load feature extraction and selection are performed to obtain the most distinguishable feature, which is then input into the matched arc fault detector for arc fault detection, and the arc fault detection result is output.

[0171] For examples consistent with the above embodiments, please refer to... Figure 6 , Figure 6 This is a functional unit block diagram of an arc fault detection device provided in an embodiment of this application, such as... Figure 6 As shown, the arc fault detection device 60 includes: a first determining unit 61, used to determine a first statistical feature of the current signal within a time window, the current signal including an even number of sampling points, the first statistical feature being used to characterize the local offset level and overall dispersion of the current signal; an analysis unit 62, used to perform waveform mutation analysis on the current signal to obtain the variation range with the largest fluctuation amplitude in the current signal; an extraction unit 63, used to extract at least one time-domain feature, at least one frequency-domain feature, and multiple second statistical features of the current signal, the multiple second statistical features being used to characterize the fluctuation amplitude and shape deviation of the current signal; a second determining unit 64, used to determine whether a transient event exists based on the first statistical feature, real-time current data, and the variation range; a matching unit 65, used to determine a target fault detection model based on the at least one time-domain feature and the variation range if the transient event exists; a feature processing unit 66, used to perform feature filtering and feature fusion processing on the at least one time-domain feature, the at least one frequency-domain feature, and the multiple second statistical features to obtain target features; and a third determining unit 67, used to determine the fault detection result based on the target features and the target fault detection model.

[0172] In one possible embodiment, in determining a first statistical characteristic of the current signal within a time window, the first determining unit 61 is specifically configured to: determine the variance of the current signal within the time window; divide the time window into a first sub-time window and a second sub-time window, wherein the number of first sampling points is the same as the number of second sampling points, the first number of sampling points being the number of sampling points of the current signal within the first sub-time window, and the second number of sampling points being the number of sampling points of the current signal within the second sub-time window; determine a first mean of the current signal within the first sub-time window; and determine a second mean of the current signal within the second sub-time window; and obtain the first statistical characteristic based on the first mean, the second mean, and the variance.

[0173] In one possible embodiment, in determining whether a transient event exists based on the first statistical feature, real-time current data, and the change interval, the second determining unit 64 is further configured to: determine the amount of change in the current signal based on the first mean, the second mean, the variance, and the real-time current data; if the amount of change is detected to be greater than a preset amount of change, determine the interval duration of the change interval; and determine whether a transient event exists based on the interval duration.

[0174] In one possible embodiment, in determining the change in the current signal based on the first mean, the second mean, the variance, and the real-time current data, the second determining unit 64 is specifically configured to: determine a first difference between the first mean and the second mean, wherein the first difference is a positive number; determine a ratio between the first difference and the variance; perform a mean operation on the first mean and the second mean to obtain a third mean; determine a second difference between the real-time current data and the third mean, wherein the second difference is a positive number; and determine the product of the second difference and the ratio to obtain the change.

[0175] In one possible embodiment, in performing feature filtering and feature fusion processing on the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features to obtain target features, the feature processing unit 66 is further configured to: determine the first load type corresponding to the current signal; obtain the monitoring requirements of the first load type; create a first feature subset, the first feature subset including at least two different dimension features from the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features; perform feature expansion processing on the first feature subset to obtain a second feature subset, the feature expansion processing being used to expand a single other feature into the current feature subset, the other feature being any one of the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features other than the first feature subset. The process involves: determining the characteristics of the second feature subset; determining a first relevance between the second feature subset and the monitoring requirement; determining a feature processing method for the second feature subset based on a third difference between the first relevance and a preset relevance, the feature processing method including feature expansion processing and / or feature removal processing, the feature removal processing being used to remove a single feature from the current feature subset; performing feature processing on the second feature subset according to the feature processing method to obtain a third feature subset; determining a second relevance between the third feature subset and the monitoring requirement; detecting that the number of features in the third feature subset is a preset number, and that the second relevance is greater than or equal to the preset relevance, and that each of the other features has been subjected to feature expansion processing and / or feature removal processing, then fusing each feature in the third feature subset to obtain the target feature.

[0176] In one possible embodiment, in determining the feature processing method of the second feature subset based on the third difference between the first relevance and the preset relevance, the feature processing unit 66 is further configured to: if the third difference is within a first preset range, then the feature processing method is the feature expansion processing; if the third difference is within a second preset range, then the feature processing method is the feature removal processing, wherein the upper limit of the first preset range is less than the lower limit of the second preset range; if the third difference is within a third preset range, then the feature processing method is both the feature removal processing and the feature expansion processing, wherein the upper limit of the second preset range is less than the lower limit of the third preset range.

[0177] In one possible embodiment, the arc fault detection device 60 is further configured to: if the transient event does not exist, acquire a second load type for the pre-transient event trigger identification; determine a target fault detection model based on the second load type; perform feature filtering and feature fusion processing on the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features to obtain target features; and determine the fault detection result based on the target features and the target fault detection model.

[0178] It is understood that since the method embodiments and the device embodiments are different presentations of the same technical concept, the content of the method embodiment section in this application should be adapted to the device embodiment section in a synchronous manner, and will not be repeated here.

[0179] In the case of using integrated units, please refer to Figure 7 , Figure 7 This is a functional unit block diagram of another arc fault detection device provided in the embodiments of this application, such as... Figure 7 As shown, the arc fault detection device 60 includes a processing module 602 and a communication module 601. The processing module 602 controls and manages the operation of the arc fault detection device 60, for example, executing the steps of the first determination unit 61, the analysis unit 62, the extraction unit 63, the second determination unit 64, the matching unit 65, the feature processing unit 66, and the third determination unit 67, and / or executing other processes of the technology described herein. The communication module 601 is used for interaction between the arc fault detection device 60 and other devices.

[0180] Among them, such as Figure 7 As shown, the arc fault detection device 60 may also include a storage module 603, which is used to store the program code and data of the arc fault detection device 60.

[0181] The processing module 602 can be a processor or controller, such as a central processing unit (CPU), a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. The processor can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0182] The communication module 601 can be a transceiver, RF circuit, or communication interface, etc. The storage module 603 can be a memory.

[0183] All relevant content in each scenario involved in the above method embodiments can be referenced from the functional descriptions of the corresponding functional modules, and will not be repeated here. The above-mentioned arc fault detection device 60 can perform the above-mentioned... Figure 2 The arc fault detection method shown is illustrated.

[0184] Please see Figure 8 , Figure 8 This is a schematic diagram of the structure of an electronic device proposed in an embodiment of this application, as shown below. Figure 8 As shown, the electronic device 800 includes a processor 810, a memory 820, a communication interface 830, and one or more programs 821. The one or more programs 821 are stored in the memory and configured to be executed by the processor. When the program is executed, it includes some or all of the steps of any arc fault detection method described in the above method embodiments. The processor, memory, and communication interface are interconnected and complete communication between them.

[0185] The memory can be volatile memory such as Dynamic Random Access Memory (DRAM) or non-volatile memory such as a hard disk drive. The memory stores a set of executable program code, and the processor calls the executable program code stored in the memory to execute some or all of the steps of any arc fault detection method described in the above embodiments of the arc fault detection method.

[0186] As can be seen, the electronic device 800 described in this application embodiment first determines a first statistical feature of the current signal within a time window. The current signal includes an even number of sampling points. The first statistical feature is used to characterize the local offset level and overall dispersion of the current signal. Then, waveform mutation analysis is performed on the current signal to obtain the variation range with the largest fluctuation amplitude in the current signal. Next, at least one time-domain feature, at least one frequency-domain feature, and multiple second statistical features of the current signal are extracted. The multiple second statistical features are used to characterize the fluctuation amplitude and shape deviation of the current signal. Then, based on the first statistical feature, real-time current data, and the variation range, it is determined whether a transient event exists. If the transient event exists, a target fault detection model is determined based on the at least one time-domain feature and the variation range. Then, feature filtering and feature fusion processing are performed on the at least one time-domain feature, the at least one frequency-domain feature, and the multiple second statistical features to obtain the target feature. Finally, based on the target feature and the target fault detection model, the fault detection result is determined.

[0187] This application uses multi-dimensional data to comprehensively determine transient events, which helps to eliminate false events and improve the accuracy of the determination. Then, the transient event triggers the matching of the fault detection model. During feature processing, time domain, frequency domain features and statistical features are screened and combined to eliminate redundant features. Finally, the target features obtained after feature processing and the matched target fault detection model are combined to perform fault detection. This makes the fault detection results more consistent with the actual load situation, thereby improving the pertinence and accuracy of fault detection. It avoids the problem of misjudgment or missed judgment that may be caused by load characteristic mismatch, and improves the efficiency and accuracy of power system fault detection.

[0188] This application also provides a computer storage medium storing a computer program for electronic data interchange, which causes a computer to perform some or all of the steps of any of the methods described in the above method embodiments, wherein the computer includes an electronic device.

[0189] This application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps of any of the methods described in the above method embodiments. The computer program product may be a software installation package, and the computer may include an electronic device.

[0190] It should be noted that, for the sake of simplicity, the aforementioned methods are described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are optional, and the actions and modules involved are not necessarily essential to this application.

[0191] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0192] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical or other forms.

[0193] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0194] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software program module.

[0195] If the integrated unit is implemented as a software program module and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, essentially, or a related part of 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 memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0196] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage device, which may include: a flash drive, a read-only memory, a random access memory, a magnetic disk, or an optical disk, etc.

[0197] The embodiments of this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The above description of the embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for detecting electric arc faults, characterized in that, include: A first statistical feature of the current signal within a time window is determined, wherein the current signal includes an even number of sampling points, and the first statistical feature is used to characterize the local offset level and the overall dispersion of the current signal; Waveform abrupt change analysis is performed on the current signal to obtain the range of change with the largest fluctuation amplitude in the current signal; At least one time-domain feature, at least one frequency-domain feature, and multiple second statistical features are extracted from the current signal, wherein the multiple second statistical features are used to characterize the fluctuation amplitude and shape deviation of the current signal; Based on the first statistical feature, real-time current data, and the change range, determine whether a transient event exists; If the transient event exists, the target fault detection model is determined based on the at least one time-domain feature and the change interval; The process involves performing feature filtering and feature fusion on the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features to obtain target features; wherein, the first load type corresponding to the current signal is determined; the monitoring requirements of the first load type are obtained; a first feature subset is created, the first feature subset including at least two different dimensions of the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features; the first feature subset is subjected to feature expansion processing to obtain a second feature subset, the feature expansion processing being used to expand a single other feature into the current feature subset, the other feature being features other than those in the first feature subset among the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features; and the second feature is determined. The first relevance of the subset to the monitoring requirement; based on the third difference between the first relevance and a preset relevance, the feature processing method of the second feature subset is determined, the feature processing method including the feature expansion processing and / or feature removal processing, the feature removal processing being used to remove a single feature from the current feature subset; the second feature subset is processed according to the feature processing method to obtain a third feature subset; the second relevance of the third feature subset to the monitoring requirement is determined; if it is detected that the number of features in the third feature subset is a preset number, and the second relevance is greater than or equal to the preset relevance, and each of the other features has been processed by the feature expansion processing and / or the feature removal processing, then each feature in the third feature subset is fused to obtain the target feature; Based on the target features and the target fault detection model, the fault detection result is determined.

2. The method according to claim 1, characterized in that, The first statistical feature of the current signal within the determined time window includes: Determine the variance of the current signal within the time window; The time window is divided into a first sub-time window and a second sub-time window. The number of the first sampling point is the same as the number of the second sampling point. The number of the first sampling point is the number of sampling points of the current signal in the first sub-time window, and the number of the second sampling point is the number of sampling points of the current signal in the second sub-time window. Determine a first mean value of the current signal within the first sub-time window; and determine a second mean value of the current signal within the second sub-time window; The first statistical feature is obtained based on the first mean, the second mean, and the variance.

3. The method according to claim 2, characterized in that, The step of determining whether a transient event exists based on the first statistical feature, real-time current data, and the change range includes: The change in the current signal is determined based on the first mean, the second mean, the variance, and the real-time current data. If the detected change amount is greater than a preset change amount, then the duration of the change interval is determined; The existence of transient events is determined based on the duration of the interval.

4. The method according to claim 3, characterized in that, The step of determining the change in the current signal based on the first mean, the second mean, the variance, and the real-time current data includes: Determine a first difference between the first mean and the second mean, wherein the first difference is a positive number; Determine the ratio of the first difference to the variance; The first mean and the second mean are averaged to obtain the third mean; Determine a second difference between the real-time current data and the third mean, wherein the second difference is a positive number; The product of the second difference and the ratio value is determined to obtain the amount of change.

5. The method according to claim 1, characterized in that, The step of determining the feature processing method for the second feature subset based on the third difference between the first relevance and the preset relevance includes: If the third difference is within the first preset range, then the feature processing method is the feature expansion processing; If the third difference is within the second preset range, then the feature processing method is the feature removal processing, where the upper limit of the first preset range is less than the lower limit of the second preset range. If the third difference is within the third preset range, then the feature processing method is the feature removal processing and the feature expansion processing, and the upper limit of the second preset range is less than the lower limit of the third preset range.

6. The method according to any one of claims 1-5, characterized in that, The method further includes: If the transient event does not exist, then obtain the second load type triggered by the preceding transient event; Determine the target fault detection model based on the second load type; The target feature is obtained by performing feature filtering and feature fusion processing on the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features; Based on the target features and the target fault detection model, the fault detection result is determined.

7. An arc fault detection device, characterized in that, include: The first determining unit is used to determine the first statistical characteristics of the current signal within a time window. The current signal includes an even number of sampling points. The first statistical characteristics are used to characterize the local offset level and the overall dispersion of the current signal. The analysis unit is used to perform waveform abrupt change analysis on the current signal to obtain the range of change with the largest fluctuation amplitude in the current signal. An extraction unit is used to extract at least one time-domain feature, at least one frequency-domain feature, and multiple second statistical features of the current signal, wherein the multiple second statistical features are used to characterize the fluctuation amplitude and shape deviation of the current signal; The second determining unit is used to determine whether a transient event exists based on the first statistical feature, real-time current data, and the change range. A matching unit is configured to determine a target fault detection model based on at least one time-domain feature and the change interval if the transient event exists. A feature processing unit is configured to perform feature filtering and feature fusion processing on the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features to obtain target features. Specifically, the feature processing unit is configured to: determine the first load type corresponding to the current signal; obtain the monitoring requirements of the first load type; create a first feature subset, the first feature subset including at least two different dimensions of the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features; and perform feature expansion processing on the first feature subset to obtain a second feature subset, the feature expansion processing being used to expand a single other feature into the current feature subset, the other feature being any feature other than those in the at least one time-domain feature, the at least one frequency-domain feature, and the plurality of second statistical features, excluding those in the first feature subset. The process involves: determining a first relevance between the second feature subset and the monitoring requirement; determining a feature processing method for the second feature subset based on a third difference between the first relevance and a preset relevance, the feature processing method including feature expansion processing and / or feature removal processing, the feature removal processing being used to remove a single feature from the current feature subset; performing feature processing on the second feature subset according to the feature processing method to obtain a third feature subset; determining a second relevance between the third feature subset and the monitoring requirement; detecting that the number of features in the third feature subset is a preset number, and the second relevance is greater than or equal to the preset relevance, and each of the other features has been processed by the feature expansion processing and / or the feature removal processing, then fusing each feature in the third feature subset to obtain the target feature; The third determining unit is used to determine the fault detection result based on the target features and the target fault detection model.

8. An electronic device, characterized in that, The device includes: The method includes a memory, a processor, and executable program code stored in the memory and executable on the processor, wherein the processor executes the executable program code to perform the steps of the arc fault detection method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores executable program code, the executable program code including execution instructions for performing the steps of the arc fault detection method as described in any one of claims 1-6.

Images