Leakage monitoring method, system and device applied to intelligent distribution box

Abstract

The application relates to the technical field of intelligent sensing systems, in particular to a leakage monitoring method, system and device applied to an intelligent distribution box, which comprises the following steps: obtaining a residual current signal of the intelligent distribution box and performing segmented modal decomposition to obtain modal components; analyzing the amplitude fluctuation and energy dispersion degree of extreme points of each modal component, determining local mutation strength, and screening out target components; determining transient interference recognition degree based on the steepness of the left side of the peak value in the local wave peak interval and the waveform asymmetry degree on both sides, and adaptively adjusting the filter window size to obtain filtered target components; and performing leakage detection based on the filtered target components. The application solves the problem of leakage false alarm caused by harmonics and transient impact interference, and improves the high accuracy and low false alarm rate of leakage monitoring of the intelligent distribution box.

Description

Technical Field

[0001] This application relates to the field of intelligent sensing system technology, specifically to leakage current monitoring methods, systems, and devices applied to intelligent distribution boxes. Background Technology

[0002] Traditional power distribution equipment relies on reactive disconnection, making it difficult to detect early, subtle, and progressive leakage faults. Therefore, intelligent distribution boxes, with their advantages of real-time monitoring and fault early warning, are widely used. Their built-in high-precision current transformers and edge computing modules can collect multi-dimensional data in real time and use filtering algorithms to eliminate transient interference, thereby accurately identifying true resistive leakage characteristics and ensuring power supply safety.

[0003] When performing leakage current detection, the residual current signal of the three-phase circuit needs to be obtained through a residual current sensor. However, this signal is often affected by two types of interference: first, the continuous high-frequency harmonics generated by equipment such as variable frequency air conditioners and water pumps, which cause distortion of the signal fundamental frequency; second, the transient impacts caused by the starting and stopping of high-power loads such as compressors and elevator brakes. These interferences intertwine and superimpose with the original leakage current signal, which can easily lead to frequent false alarms from the monitoring system. Summary of the Invention

[0004] In a first aspect, embodiments of this application provide a leakage current monitoring method for intelligent distribution boxes, the method comprising the following steps: The residual current signal of the intelligent distribution box is acquired, and the residual current signal is segmented and decomposed into modes to obtain the mode components of each segment of the residual current signal. The amplitude fluctuation at the extreme points of each modal component is analyzed, as well as the energy concentration and dispersion at the extreme points, to determine the intensity of local abrupt changes in each modal component, so as to screen out the target component from all modal components. The time interval between adjacent troughs on each modal component is denoted as the local peak interval. By analyzing the steepness of the signal change to the left of the peak and the temporal asymmetry of the waveforms on both sides of the peak within the local peak interval, the transient interference identification degree within the local peak interval is determined. Based on the transient interference identification degree, the size of the filtering window in the corresponding signal filtering process within the local peak interval is adaptively adjusted to obtain the filtered target component. The residual current signal is reconstructed based on the filtered target component to detect leakage current in the intelligent distribution box.

[0005] Preferably, the method for determining the intensity of local abrupt changes in each modal component is as follows: Analyze the amplitude fluctuation at the extreme points of each modal component to determine the amplitude fluctuation characteristic value of each modal component; Based on the energy concentration and dispersion at the extreme points of each modal component, the energy distribution characteristic value of each modal component is determined; The intensity of local mutations in each modal component is positively correlated with the amplitude fluctuation characteristic value and the energy distribution characteristic value, respectively.

[0006] Preferably, the process for determining the amplitude fluctuation characteristic values ​​of each modal component is as follows: Calculate the difference between any extreme value in each modal component and its left and right adjacent extreme values, and use the information entropy of all differences at all extreme values ​​as the amplitude fluctuation characteristic value of each modal component.

[0007] Preferably, the process for determining the energy distribution characteristic values ​​of each modal component is as follows: Based on the average level and dispersion of the difference between the energy corresponding to all extreme values ​​in the frequency domain and the average energy of the modal component in the frequency domain, the first energy characteristic value of each modal component is determined. Calculate the dispersion of the kurtosis of all extreme points in each modal component, and denote it as the second energy eigenvalue of each modal component; The energy distribution characteristic values ​​of each modal component are positively correlated with the first energy characteristic value and the second energy characteristic value, respectively.

[0008] Preferably, the step of filtering the target component from all modal components includes: For each segment of residual current signal, the local mutation intensities of all its modal components are sorted in ascending order. The minimum value in the first-order forward difference sequence of the sorting result is used as the segmentation threshold, and the modal components with local mutation intensities greater than the segmentation threshold are used as target components.

[0009] Preferably, the transient interference identification degree within the local peak interval is positively correlated with the steepness of the signal change to the left of the peak and negatively correlated with the temporal asymmetry of the waveforms on both sides of the peak.

[0010] Preferably, the adaptive adjustment of the filter window size during the signal filtering process within the local peak interval includes: The signal within the local peak interval corresponds to the adjusted filter window size. The expression is: In the formula, This indicates the preset basic control weight; Normalized value representing the detectability of transient interference within a local peak interval; `round[]` represents the preset base window radius; `round[]` represents the rounding function.

[0011] Preferably, the leakage current detection of the intelligent distribution box includes: Extract the time-domain and frequency-domain features of each segment of residual current signal after reconstruction, input the time-domain and frequency-domain features into the pre-trained leakage fault detection model, and output the leakage fault detection results of the smart distribution box for each segment of residual current signal in the corresponding time period.

[0012] Secondly, embodiments of this application provide a leakage current monitoring system for intelligent distribution boxes, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements any of the above-described leakage current monitoring methods for intelligent distribution boxes.

[0013] Thirdly, embodiments of this application also provide a leakage current monitoring device for a smart distribution box, wherein the device stores a computer program, and when the computer program is executed by a processor, it implements the leakage current monitoring method for a smart distribution box as described above.

[0014] As can be seen from the above embodiments, the leakage current monitoring method for intelligent distribution boxes provided in this application has at least the following beneficial effects: This application constructs local mutation intensity by quantifying the amplitude fluctuation and energy concentration of modal components, accurately distinguishes and eliminates regular high-frequency harmonic interference, and screens out low-frequency target components containing real leakage current and transient impact. It effectively cuts off the aliasing interference path of harmonic and leakage current signals, laying a clean and reliable signal foundation for subsequent adaptive filtering and noise reduction and accurate leakage current identification. Furthermore, this application constructs transient interference identification by quantifying the waveform asymmetry and steepness of local peak intervals, accurately distinguishing the peak shape differences between physical leakage current and transient impact, and adaptively adjusting the filter window size accordingly, thereby achieving strong smoothing of impact interference and accurate preservation of leakage current mutation characteristics. This effectively avoids the problems of fault signal loss or interference residue that are easily caused by traditional filtering, and greatly improves the accuracy and reliability of leakage current monitoring. Ultimately, this application obtains a clean residual current signal by reconstructing the filtered target component, and trains a support vector machine leakage detection model based on a multi-dimensional time-frequency domain feature dataset constructed by simulation. This achieves accurate classification and identification of real-time signal leakage status, completely eliminating the misleading effects of harmonic and transient impact interference on feature extraction and model judgment, and improving the high accuracy and low false alarm rate of leakage monitoring in intelligent distribution boxes. Attached Figure Description

[0015] Figure 1 A flowchart illustrating the steps of a leakage current monitoring method for a smart distribution box provided in one embodiment of this application; Figure 2 A flowchart of the target component screening process provided in one embodiment of this application. Detailed Implementation

[0016] The following description, in conjunction with the accompanying drawings, details the specific solutions for the leakage current monitoring method, system, and device for intelligent distribution boxes provided in this application.

[0017] Please see Figure 1 The diagram illustrates a flowchart of a leakage current monitoring method for a smart distribution box according to an embodiment of this application. The method includes the following steps: Step S1: Obtain the residual current signal of the intelligent distribution box, segment and decompose the residual current signal to obtain the modal components of each segment of the residual current signal.

[0018] Signal acquisition and transmission: The residual current signal is acquired in real time through the residual current transformer built into the intelligent distribution box. In order to meet the accuracy requirements of leakage current feature capture, the signal sampling rate is set to 10kHz in this embodiment.

[0019] A segmentation strategy based on frequency domain distribution differences: In leakage current monitoring scenarios, the actual leakage fault signal is mainly concentrated in the low-frequency range, while the transient impact interference generated by the start-up and shutdown of the intelligent distribution box is also distributed in the low-frequency band, resulting in severe overlap between the two in the frequency domain. In contrast, the harmonic interference generated by devices such as frequency converters is high-frequency modulation interference, distributed in a different frequency band than the actual leakage fault. Based on the above frequency domain distribution differences, prioritizing the removal of high-frequency harmonic interference is a key prerequisite for subsequent leakage current feature extraction. Furthermore, considering the suddenness and locality of leakage faults, the proportion of faults in long-sequence signals is relatively small, and leakage current features are easily overwhelmed. Therefore, it is necessary to segment the residual current signal. In this embodiment, the length of each residual current signal segment is 1 second. This is a trade-off between the amount of signal data and the completeness of feature extraction at a sampling rate of 10 kHz. One second can provide 10,000 sampling points, which is sufficient to cover the complete evolution cycle of leakage faults and transient impact interference to ensure the reliability of mode decomposition and frequency band feature extraction. At the same time, it avoids computational redundancy and real-time response delay caused by excessively long windows. This length can be other values. If the application scenario has extremely high real-time requirements, it can be shortened to 0.5 seconds to speed up the detection frequency. If it is necessary to analyze the characteristics of lower frequency slowly changing faults, it can be extended to 2 seconds or more. However, it is necessary to ensure that the window length is always greater than the single duration period of the target interference or fault signal to avoid the end effect and feature distortion of mode decomposition due to signal truncation.

[0020] Mode decomposition balancing timeliness and accuracy: For each segment of residual current signal, the integrated empirical mode decomposition (EEMD) algorithm is used to perform mode decomposition, adaptively decomposing the original residual current signal into mode components (IMF) of different frequency bands. In order to balance the timeliness of the calculation response under leakage current environment and the accuracy of end-point effect suppression in mode decomposition, specific constraints are imposed on the parameters in the integrated empirical mode decomposition algorithm: the added white noise intensity is set to 0.1 times the signal standard deviation, and the number of integrations is set to 5, thereby obtaining the decomposed mode components corresponding to each segment of residual current signal.

[0021] The process of using the Integrated Empirical Mode Decomposition (EEMD) algorithm to perform mode decomposition on a signal is a well-known technique and will not be described in detail here.

[0022] Step S2: Analyze the amplitude fluctuation at the extreme points of each modal component, as well as the energy concentration and dispersion at the extreme points, to determine the intensity of local mutations in each modal component, so as to screen out the target component from all modal components.

[0023] When high-frequency harmonic interference from devices such as frequency converters exists in the power distribution system, the PWM carrier is coupled into the residual current loop, and the interference source is a periodic steady-state signal. Combined with the continuous operation of the frequency converter, this causes the corresponding IMF component to exhibit dense, constant-amplitude or slowly varying-amplitude high-frequency oscillations. These waveforms are typically symmetrical (equal or slightly different absolute values), with stable peak values, rarely exhibiting isolated spikes, and have uniform energy distribution, showing a stable and unfluctuating instantaneous energy. Conversely, when leakage faults occur or transient impacts such as compressor startup, elevator brake activation, or high-power capacitive load power-on occur, these two types of signals are usually distributed in similar low-frequency bands and exist in an aliased manner, causing the corresponding IMF component waveform to oscillate irregularly. Simultaneously, the instantaneous large energy release of the start-stop impact and the instantaneous current jump caused by leakage disrupt the symmetrical distribution of the original signal's positive and negative peak values. Randomly appearing isolated peaks make the differences between adjacent peaks and valleys significant, and the energy in the IMF component is highly concentrated in the extremely short moments of abrupt changes or impacts, leading to drastic changes in the kurtosis of local energy at different times.

[0024] Based on the fundamental differences in physical behavior between high-frequency steady-state harmonics and low-frequency aliased transient signals, it is clear that the key to distinguishing between the two lies in the "degree of breaking" of the positive and negative symmetry of the quantized waveform and the "degree of local aggregation and abrupt change" in the energy distribution. Therefore, this embodiment analyzes the amplitude fluctuation at the extreme points of each modal component, as well as the degree of energy concentration and dispersion at the extreme points, to determine the intensity of local abrupt changes in each modal component, in order to screen out the target component from all modal components. The specific process is as follows: In this embodiment, the amplitude fluctuation at the extreme points of each modal component is analyzed to determine the amplitude fluctuation characteristic value of each modal component. Specifically, in this embodiment, the absolute difference between any extreme value of each modal component and its left and right adjacent extreme values ​​is calculated. All the obtained absolute differences form an amplitude difference sequence. Then, the numerical range of the amplitude difference sequence is evenly divided into N equal-width intervals and frequency statistics are performed. In this embodiment, N is 5. Further, the probability of the difference value occurring in each interval is calculated, and then the information entropy is solved based on the probability. In this embodiment, the information entropy is calculated using Shannon entropy, and the Shannon entropy solved based on the probability is used as the amplitude fluctuation characteristic value of each modal component.

[0025] The calculation process of Shannon entropy is a well-known technique and will not be elaborated further.

[0026] Furthermore, based on the energy concentration and dispersion at the extreme points of each modal component, the energy distribution characteristic values ​​of each modal component are determined, specifically as follows: Based on the average level and dispersion of the difference between the energy corresponding to all extreme values ​​in the frequency domain and the average energy of the modal component in the frequency domain, the first energy characteristic value of each modal component is determined. That is, in this embodiment, the average level and dispersion of the difference between the energy corresponding to all extreme values ​​in the frequency domain and the average energy of the modal component in the frequency domain are calculated respectively, and denoted as the energy characteristic mean and energy characteristic dispersion. The sum of the energy characteristic mean and energy characteristic dispersion is taken as the first energy characteristic value of each modal component. The dispersion of kurtosis at all extreme points in each modal component is denoted as the second energy characteristic value of each modal component. The calculation process of kurtosis is a well-known technique and will not be described in detail here.

[0027] It should be noted that before calculating the first energy characteristic value, the analytical signals of each modal component must first be extracted by Hilbert transform, and the instantaneous amplitude trend trajectory of the entire process must be calculated. Then, the corresponding instantaneous amplitude of the cross-section mapping at the extreme moment point is found, and its square is defined as the instantaneous energy expressed in the frequency domain at the extreme value. The process of extracting the corresponding energy of each modal component in the frequency domain using Hilbert transform is a well-known technique and will not be described in detail here.

[0028] It should be noted that, in this embodiment, the ratio of the energy corresponding to all extreme values ​​in the frequency domain to the average energy of the modal component in the frequency domain is used as the difference between the energy corresponding to all extreme values ​​in the frequency domain and the average energy of the modal component in the frequency domain; the mean of the difference between the energy corresponding to all extreme values ​​in the frequency domain and the average energy of the modal component in the frequency domain is used as the average level of the difference between the energy corresponding to all extreme values ​​in the frequency domain and the average energy of the modal component in the frequency domain; the standard deviation of the difference between the energy corresponding to all extreme values ​​in the frequency domain and the average energy of the modal component in the frequency domain is used as the dispersion of the difference between the energy corresponding to all extreme values ​​in the frequency domain and the average energy of the modal component in the frequency domain; and the standard deviation of the kurtosis at all extreme points in each modal component is used as the dispersion of the kurtosis at all extreme points in each modal component.

[0029] It should be noted that if kurtosis cannot be calculated (due to insufficient points or boundary constraints), it is assumed by default to not having abrupt change characteristics, that is, the second energy characteristic value is set to 0.

[0030] It should be understood that a positive correlation means that the dependent variable increases as the independent variable increases, and the dependent variable decreases as the independent variable decreases. The specific relationship can be additive or multiplicative, etc., and is determined by the actual application. This application does not impose any special restrictions.

[0031] Preferably, as one implementation, in this embodiment, the sum of the first energy characteristic value and the second energy characteristic value of each modal component is used as the energy distribution characteristic value of each modal component.

[0032] Furthermore, based on the amplitude fluctuation characteristic value and the energy distribution characteristic value, the local abrupt change intensity of each modal component is determined, specifically as follows: In this embodiment, the intensity of local mutations in each modal component is positively correlated with the amplitude fluctuation characteristic value and the energy distribution characteristic value, respectively.

[0033] Preferably, as one implementation method, in this embodiment, the product of the amplitude fluctuation characteristic value and the energy distribution characteristic value of each modal component is used as the local mutation intensity of each modal component.

[0034] Based on the local mutation intensity, it can be understood that the local mutation intensity is used to characterize the differences in peak-valley symmetry and energy distribution between harmonic interference and low-frequency aliasing signals (transient impulses and leakage faults), reflecting the comprehensive distortion degree of the signal in terms of amplitude fluctuation in the time domain and energy concentration in the frequency domain. Since it is composed of the product of amplitude fluctuation characteristic value and energy distribution characteristic value, the amplitude fluctuation characteristic value is a dimensionless statistical measure, and the energy distribution characteristic value is composed of the energy ratio and the dimensionless kurtosis standard deviation, the local mutation intensity is a dimensionless index. A larger amplitude fluctuation characteristic value indicates a more severe failure of the positive and negative symmetry characteristics of adjacent peaks and valleys in the corresponding modal component, and a more irregular distribution of amplitude differences. This reflects a deeper degree of damage to the signal from transient high-energy impacts or leakage current jumps, and a significant increase in the intensity of corresponding local abrupt changes. This leads to the corresponding modal component being identified as a target component that needs to be retained. Conversely, a smaller amplitude fluctuation characteristic value indicates that adjacent peaks and valleys are closer to equal amplitude symmetry with minimal differences. This reflects that the signal more closely conforms to the characteristics of periodic steady-state harmonics. Its effect is to reduce the intensity of local abrupt changes, causing the corresponding modal component to be directly treated as high-frequency harmonic interference. Furthermore, the larger the energy distribution characteristic value, the higher the instantaneous energy at the peaks and troughs of the modal components is compared to the overall average level. Moreover, the more drastic the energy kurtosis changes between different peak and trough intervals, the more concentrated the signal energy is at the specific moment of leakage current mutation or transient impact. The increased intensity of local mutations indicates a high probability of non-steady-state faults or impacts in the corresponding modal components. Conversely, the smaller the energy distribution characteristic value, the more uniform and stable the instantaneous energy distribution is, and the less obvious the local kurtosis fluctuations are. This indicates that the signal lacks transient energy release points, and the corresponding local mutation intensity is smaller. Such components should be directly removed to avoid aliasing misjudgment.

[0035] Furthermore, this embodiment uses the local abrupt change intensity of each modal component to screen out the target component from all modal components. Specifically: In this embodiment, for each segment of residual current signal, the local mutation intensities of all its modal components are arranged in ascending order, and the minimum value in the first-order forward difference sequence of the arrangement result is used as the segmentation threshold. Modal components with local mutation intensities greater than the segmentation threshold are used as target components.

[0036] Preferably, the flowchart of the target component screening process provided in this embodiment is as follows: Figure 2 As shown.

[0037] The method for obtaining the first-order forward difference sequence is a well-known technique and will not be elaborated further.

[0038] Based on the target component, it can be understood that the target component is the set of low-frequency mode components to be processed that are screened from all modal components, after removing regular high-frequency harmonic interference, and may still contain transient impact interference and real leakage faults. It represents the effective frequency band components in the original residual current signal whose frequency overlaps with the leakage characteristics and whose waveform has local abrupt change characteristics. It is the direct target for subsequent quantitative analysis of transient impact and leakage peak shape differences and adaptive filtering noise reduction.

[0039] Thus, this embodiment constructs the local mutation intensity by quantifying the amplitude fluctuation and energy concentration of modal components, accurately distinguishes and eliminates regular high-frequency harmonic interference, and filters out low-frequency target components containing real leakage current and transient impact. It effectively cuts off the aliasing interference path of harmonic and leakage current signals, laying a clean and reliable signal foundation for subsequent adaptive filtering and noise reduction and accurate leakage current identification.

[0040] Step S3: The time interval between adjacent troughs on each modal component is denoted as the local peak interval. By analyzing the temporal asymmetry of the waveforms on both sides of the peak in the local peak interval and the steepness of the signal change to the left of the peak, the transient interference identification degree in the local peak interval is determined. Based on the transient interference identification degree, the size of the filtering window in the corresponding signal filtering process in the local peak interval is adaptively adjusted to obtain the filtered target component.

[0041] After screening for IMF components that may contain transient impact interference and leakage faults, median filtering is used to denoise the remaining components to be processed in order to further filter out the transient impact interference that still remains in the signal. However, the smoothing effect of median filtering depends on the window size: if the window is too small, it cannot effectively eliminate transient impacts with extremely high energy; if the window is too large, it will also erase the leakage fault signal whose waveform already changes abruptly. Therefore, the filtering window must be adaptively adjusted based on the differences in physical form between leakage faults and transient impact interference.

[0042] Physical morphology analysis is performed on the characteristic set of possible transient impact interference or leakage faults: When a leakage fault occurs, the current rises from the normal value to the leakage value, usually triggered by physical contact (such as insulation failure), which is extremely fast (microseconds to milliseconds). However, the decline phase is often slower due to the extinction of the arc, the action of the protector, or the removal of the human body (affected by muscle reaction and mechanical tripping time). Therefore, the signal peak shape exhibits strong asymmetry, and the overall duration is relatively long. At the same time, due to the limitation of dielectric polarization relaxation time, the current change tends to smooth out at the peak, and the slope decreases significantly. When the instrument starts and stops, causing transient impact interference, its essence is LC oscillation (energy release of line inductance and distributed capacitance). The rise and fall times of an ideal LC oscillation are consistent, which is manifested in the IMF component as a high degree of symmetry between the rise and fall edges, and a short overall duration. At the same time, according to Faraday's law of electromagnetic induction, the extremely high rate of change of current will generate a huge induced electromotive force, driving the current to form an extremely steep peak. The rise edge of this peak is much faster than any physical leakage process.

[0043] Based on the above analysis, this embodiment defines the time interval between adjacent troughs on each modal component as a local peak interval. By analyzing the temporal asymmetry of the waveforms on both sides of the peak within the local peak interval, as well as the steepness of the signal change to the left of the peak, the transient interference identification rate within the local peak interval is determined. Based on the transient interference identification rate, the size of the filtering window in the corresponding signal filtering process within the local peak interval is adaptively adjusted to obtain the filtered target component. The specific process is as follows: In this embodiment, the transient interference identification rate within the local peak interval is positively correlated with the steepness of the signal change to the left of the peak and negatively correlated with the temporal asymmetry of the waveforms on both sides of the peak.

[0044] It should be understood that a positive correlation means that the dependent variable increases as the independent variable increases, and the dependent variable decreases as the independent variable decreases. The specific relationship can be additive or multiplicative, etc., and is determined by the actual application. This application does not impose any special restrictions. A negative correlation means that the dependent variable decreases as the independent variable increases, and the dependent variable increases as the independent variable decreases. The relationship can be subtractive or divisive, etc., and is determined by the actual application.

[0045] Preferably, as one implementation method, the transient interference recognition rate within the local peak interval in this embodiment is... The expression is: In the formula, It indicates the time-domain asymmetry of the waveforms on both sides of the peak within a local peak interval; This indicates the steepness of the signal change to the left of the peak within a local peak interval; This represents a preset constant greater than 0, used to prevent the denominator from being 0. In this embodiment... The value of is 0.01. Provided that the denominator is not zero and does not excessively affect the calculation result, the implementer may also set it according to the specific situation. This embodiment does not impose any special restrictions.

[0046] Additionally, the quantization process for the time-domain asymmetry of the waveforms on both sides of the peak and the steepness of the signal change to the left of the peak in this embodiment is as follows: The quantization process for the steepness of the signal change to the left of the peak is as follows: The steepness of the signal change to the left of the peak within a local peak interval The expression is: In the formula, This represents the first-order forward difference of the amplitude at the sampling point adjacent to the left of the peak value within a local peak interval. It represents the mean of the first-order forward difference values ​​of the amplitude at all sampling points to the left of the peak value within the local peak interval.

[0047] The quantification process of the temporal asymmetry of the waveforms on both sides of the peak is as follows: Calculate the absolute difference between the total number of sampling points on the left side and the total number of sampling points on the right side of the peak within the local peak interval, and record it as the sampling point number difference; further, calculate the sum of the total number of all sampling points within the local peak interval and the preset factor, and calculate the ratio of the sampling point number difference to the corresponding sum, which is used as the quantification result of the temporal asymmetry of the waveforms on both sides of the peak within the local peak interval. The preset factor is a constant greater than 0, used to prevent the denominator from being 0 when calculating the ratio, which would make the calculation meaningless. In this embodiment, the preset factor is set to 1. Under the premise of ensuring that the denominator is not 0 and does not excessively affect the calculation result, the implementer can also set it according to the specific situation. This embodiment does not impose any special restrictions.

[0048] Based on transient interference identification, it can be understood that transient interference identification is used to characterize the time-domain symmetry of the waveforms on both sides of the peak and the steepness of the signal change to the left of the peak. It reflects the difference between LC oscillating impulse interference and physical contact leakage faults in terms of rising edge steepness and overall peak structure. The steeper the signal change to the left of the peak within a local peak interval and the smaller the time-domain asymmetry of the waveforms on both sides of the peak, the greater the transient interference identification. This indicates that the peaks on both sides of the interval are more symmetrical and the rising edge is extremely steep (low smoothness), reflecting that the corresponding signal in the local peak interval is a start-up / shutdown trigger for power distribution equipment. The greater the likelihood of transient impact interference, the larger the filtering window should be allocated to the corresponding signal in the local peak interval to effectively smooth out such interference noise. Conversely, the smaller the steepness of the signal change to the left of the peak in the local peak interval and the greater the time-domain asymmetry of the waveforms on both sides of the peak, the lower the recognition of transient interference. This indicates that the more asymmetric the peaks are and the flatter the rising edge, the greater the likelihood that the waveform distortion in the corresponding local peak interval is caused by leakage fault. Its effect is to limit the size of the filtering window from being excessively smoothed, thereby effectively preserving the leakage sudden change characteristics and preventing them from being erased by the filtering operation.

[0049] Furthermore, based on transient interference recognition, this embodiment adaptively adjusts the size of the filtering window during the corresponding signal filtering process within the local peak interval to obtain the filtered target component. Specifically: In this embodiment, as one implementation method, the signal within the local peak interval corresponds to the adjusted filter window size. The expression is: In the formula, This indicates the preset basic control weight; Normalized value representing the detectability of transient interference within a local peak interval; `round[]` represents the preset base window radius; `round[]` represents the rounding function.

[0050] It should be noted that in this embodiment, the preset basic control weight is 0.5, and the preset basic window radius is 100, which means it contains 100 sampling points.

[0051] It should be further explained that the generation process of the normalized value Z of transient interference identification in this section is to place the currently calculated specific peak identification value into a large set interval consisting of the original transient interference identification values ​​of all intervals harvested by the 1-second time domain analysis segment. The maximum and minimum values ​​in this set are used as the maximum and minimum value parameters in the minimum-maximum normalization method, respectively, to normalize the transient interference identification.

[0052] The process of normalizing data using the minimum-maximum normalization method is a well-known technique and will not be elaborated further.

[0053] Based on the adjusted filter window size, it can be understood that the adjusted filter window size is the actual median filtering range dynamically adapted according to the transient interference recognition degree of each local peak interval. It reflects the targeted adjustment result of the filter on the smoothing strength of the signal in the current interval. When the interval is identified as transient impulse interference, the window length is increased to enhance the smoothing effect and thus completely filter out high-frequency oscillation noise. When the interval is identified as leakage fault characteristics, the window length is correspondingly reduced to weaken the smoothing effect and thus retain the true abrupt change edge of the signal. This achieves differentiated adaptive denoising based on the physical properties of the signal within the same component to be processed, ensuring a dynamic balance between interference filtering and feature preservation.

[0054] Furthermore, each target component is used as input to the median filtering algorithm. The adjusted filter window size corresponding to each local peak interval in the target component is used as the filter window size parameter in the median filtering algorithm to filter the target component, and finally the filtered target component is output.

[0055] Thus, this embodiment constructs transient interference identification by quantifying the waveform asymmetry and steepness of local peak intervals, accurately distinguishing the peak shape differences between physical leakage current and transient impact, and adaptively adjusting the filter window size accordingly. This achieves strong smoothing of impact interference and accurate preservation of leakage current mutation characteristics, effectively avoiding the problems of fault signal loss or interference residue that are easily caused by traditional filtering, and greatly improving the accuracy and reliability of leakage current monitoring.

[0056] Step S4: Reconstruct the residual current signal based on the filtered target component to perform leakage current detection on the intelligent distribution box.

[0057] First, the filtered target components are reconstructed to obtain a clean residual current signal after removing high-frequency harmonics and transient impact interference. The process of reconstructing the modal components into a signal is a well-known technique and will not be described in detail here.

[0058] Furthermore, to accurately identify leakage current status based on the reconstructed signal, a leakage fault detection model needs to be pre-built and trained. The specific process is as follows: A power distribution network leakage simulation model with parameters consistent with the smart distribution box under test is constructed using MATLAB / Simulink. The power supply, distribution transformer, feeder lines, and line parameters (including resistance, inductance, and capacitance per unit length corresponding to the positive and zero sequence of the three-phase lines) in the model are all consistent with those of the distribution box under test. A single-phase grounding fault is set on the target feeder to obtain leakage fault signals. Combining the constructed leakage simulation model, the simulation parameters are adjusted to obtain a simulated zero-sequence current signal containing both normal and leakage states. Since the zero-sequence current is the average of the three-phase currents, and the residual current is the sum of the three-phase currents, the obtained simulated zero-sequence current signal is multiplied by three to convert it into a simulated residual current signal, and then labeled according to the corresponding state. The system labels the signal as either normal or a leakage fault. Then, the labeled simulated residual current signal is divided into time windows, with a window size of 1 second in this embodiment. For each time window, time-domain features (maximum peak value, mean, variance, kurtosis, waveform factor, impulse factor) and frequency-domain features (maximum spectral peak value, dominant frequency, spectral centroid, frequency variance) are extracted. All extracted features are combined to construct a multi-dimensional training vector. Finally, all multi-dimensional training vectors are used as input to train the model using a support vector machine. The boundary loss metric for training is configured as a hinge loss function to maximize the classification margin. The optimization method utilizes the Sequence Minimum Optimization (SMO) algorithm as the core strategy to derive the SVM margin maximization process. The maximum training iterations are set to 100 epochs, with a learning rate of 0.001. The trained leakage fault detection model is then output.

[0059] Finally, leakage detection is performed based on the real-time acquired and reconstructed residual current signal: the time-domain and frequency-domain features of each segment of the reconstructed residual current signal are extracted to construct a multi-dimensional feature vector; the multi-dimensional feature vector is input into the pre-trained leakage fault detection model, and the leakage fault detection result of the smart distribution box for each segment of residual current signal in the corresponding time period is output.

[0060] Thus, this embodiment obtains a clean residual current signal by reconstructing the filtered target component, and trains a support vector machine leakage detection model based on the multi-dimensional time-frequency domain feature dataset constructed by simulation. This achieves accurate classification and identification of the leakage status of real-time signals, completely eliminates the misleading effects of harmonic and transient impact interference on feature extraction and model judgment, and improves the high accuracy and low false alarm rate of leakage monitoring in intelligent distribution boxes.

[0061] Based on the same inventive concept as the above methods, this application also provides a leakage current monitoring system for intelligent distribution boxes, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements any one of the above-described leakage current monitoring methods for intelligent distribution boxes.

[0062] Based on the same inventive concept as the above method, this application also provides a leakage current monitoring device for a smart distribution box. The device stores a computer program, which, when executed by a processor, implements the leakage current monitoring method for a smart distribution box as described above.

[0063] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.

Claims

1. A leakage current monitoring method applied to intelligent distribution boxes, characterized in that, The method includes the following steps: The residual current signal of the intelligent distribution box is acquired, and the residual current signal is segmented and decomposed into modes to obtain the mode components of each segment of the residual current signal. The amplitude fluctuation at the extreme points of each modal component is analyzed, as well as the energy concentration and dispersion at the extreme points, to determine the intensity of local abrupt changes in each modal component, so as to screen out the target component from all modal components. The time interval between adjacent troughs on each modal component is denoted as the local peak interval. By analyzing the steepness of the signal change to the left of the peak and the temporal asymmetry of the waveforms on both sides of the peak within the local peak interval, the transient interference identification degree within the local peak interval is determined. Based on the transient interference identification degree, the size of the filtering window in the corresponding signal filtering process within the local peak interval is adaptively adjusted to obtain the filtered target component. The residual current signal is reconstructed based on the filtered target component to detect leakage current in the intelligent distribution box.

2. The leakage current monitoring method applied to intelligent distribution boxes as described in claim 1, characterized in that, The method for determining the intensity of local abrupt changes in each modal component is as follows: Analyze the amplitude fluctuation at the extreme points of each modal component to determine the amplitude fluctuation characteristic value of each modal component; Based on the energy concentration and dispersion at the extreme points of each modal component, the energy distribution characteristic value of each modal component is determined; The intensity of local mutations in each modal component is positively correlated with the amplitude fluctuation characteristic value and the energy distribution characteristic value, respectively.

3. The leakage current monitoring method applied to intelligent distribution boxes as described in claim 2, characterized in that, The process for determining the amplitude fluctuation characteristic values ​​of each modal component is as follows: Calculate the difference between any extreme value in each modal component and its left and right adjacent extreme values, and use the information entropy of all differences at all extreme values ​​as the amplitude fluctuation characteristic value of each modal component.

4. The leakage current monitoring method applied to intelligent distribution boxes as described in claim 2, characterized in that, The process for determining the energy distribution characteristic values ​​of each modal component is as follows: Based on the average level and dispersion of the difference between the energy corresponding to all extreme values ​​in the frequency domain and the average energy of the modal component in the frequency domain, the first energy characteristic value of each modal component is determined. Calculate the dispersion of the kurtosis of all extreme points in each modal component, and denote it as the second energy eigenvalue of each modal component; The energy distribution characteristic values ​​of each modal component are positively correlated with the first energy characteristic value and the second energy characteristic value, respectively.

5. The leakage current monitoring method applied to intelligent distribution boxes as described in claim 1, characterized in that, The process of filtering the target component from all modal components includes: For each segment of residual current signal, the local mutation intensities of all its modal components are sorted in ascending order. The minimum value in the first-order forward difference sequence of the sorting result is used as the segmentation threshold, and the modal components with local mutation intensities greater than the segmentation threshold are used as target components.

6. The leakage current monitoring method applied to intelligent distribution boxes as described in claim 1, characterized in that, The transient interference identification rate within the local peak interval is positively correlated with the steepness of the signal change to the left of the peak and negatively correlated with the temporal asymmetry of the waveforms on both sides of the peak.

7. The leakage current monitoring method applied to intelligent distribution boxes as described in claim 1, characterized in that, The adaptive adjustment of the filter window size during the filtering process of the corresponding signal within the local peak interval includes: The signal within the local peak interval corresponds to the adjusted filter window size. The expression is: In the formula, This indicates the preset basic control weight; Normalized value representing the detectability of transient interference within a local peak interval; `round[]` represents the preset base window radius; `round[]` represents the rounding function.

8. The leakage current monitoring method applied to intelligent distribution boxes as described in claim 1, characterized in that, The leakage current detection of the intelligent distribution box includes: Extract the time-domain and frequency-domain features of each segment of residual current signal after reconstruction, input the time-domain and frequency-domain features into the pre-trained leakage fault detection model, and output the leakage fault detection results of the smart distribution box for each segment of residual current signal in the corresponding time period.

9. A leakage current monitoring system applied to an intelligent distribution box, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the leakage current monitoring method for intelligent distribution boxes as described in any one of claims 1-8.

10. A leakage current monitoring device applied to an intelligent distribution box, wherein the device stores a computer program, characterized in that, When the computer program is executed by the processor, it implements the leakage current monitoring method for intelligent distribution boxes as described in any one of claims 1-8.

Images