A series arc fault identification method and system for low voltage switchgear

Abstract

The application discloses a series fault arc identification method and system of a low-voltage cabinet, and relates to the technical field of power system fault detection.The method comprises the following steps: collecting current signals and voltage signals of an outgoing line circuit of the low-voltage cabinet; decomposing the current signals into a plurality of IMF components by using CEEMDAN; performing Hilbert transform on each IMF component to extract an instantaneous phase, and calculating phase rearrangement entropy through phase space reconstruction; extracting an instantaneous amplitude sequence to calculate an amplitude modulation index; performing time-frequency transformation to calculate a time-frequency aggregation index; nonlinearly interacting and fusing the three kinds of features to obtain a comprehensive judgment index, and performing preliminary judgment by using a sliding window adaptive threshold; when the preliminary judgment exceeds the threshold, entering physical model verification, and calculating the correlation coefficient between the measured voltage-current and the pre-stored arc physical model to perform secondary confirmation.The application realizes early prediction of series fault arcs, and significantly improves the detection accuracy and anti-interference ability through nonlinear feature interaction and double verification of a physical model.

Description

Technical Field

[0001] This application relates to the field of power system fault detection technology, and in particular to a method and system for identifying series fault arcs in low-voltage switchgear. Background Technology

[0002] Critical components inside low-voltage switchgear, such as busbar connections, circuit breaker contacts, and cable joints, are prone to loosening or poor contact due to long-term vibration, thermal expansion and contraction, and material aging, which can induce series fault arcs. The loop current of a series fault arc is on the same order of magnitude as the normal load current, making it difficult for traditional overcurrent protection devices to operate effectively. Furthermore, the fault characteristics are easily masked by nonlinear load currents, making it one of the main hidden dangers causing electrical fires.

[0003] Currently, there are three main types of methods for detecting series fault arcs: First, methods based on signal modeling and threshold judgment, which extract time-domain or frequency-domain features of the current signal and compare them with a fixed threshold for discrimination. These methods are sensitive to load changes and have poor generalization ability. Second, methods based on signal decomposition and entropy features, which use algorithms such as empirical mode decomposition and variational mode decomposition combined with indicators such as sample entropy and permutation entropy for fault detection. However, most of these methods still rely on fixed thresholds and are difficult to adapt to fluctuations in operating conditions. Third, methods based on deep learning, which transform the current signal into a time-frequency graph and then input it into a neural network for recognition. These methods have high recognition accuracy but rely on a large amount of labeled data and consume a lot of computational resources, making them difficult to deploy in real time on low-voltage switchgear intelligent terminals.

[0004] The aforementioned existing technologies all aim to detect arcs that have already occurred, without addressing early prediction before arcs occur. Furthermore, the feature extraction and fusion methods of existing methods are relatively simple, lacking effective false alarm suppression mechanisms. Therefore, there is an urgent need for an identification method that can achieve early prediction of series fault arcs without requiring a large amount of training data and possesses high anti-interference capabilities. Summary of the Invention

[0005] To address the technical problems of the prior art, this application provides a method and system for identifying series fault arcs in low-voltage switchgear. The purpose is to: solve the technical problems of feature extraction being easily affected by load type, threshold setting relying on experience, and insufficient generalization ability in existing series fault arc identification methods; achieve early prediction of series fault arcs, unlike existing technologies that can only detect arcs that have already occurred; and improve detection sensitivity and anti-interference ability by introducing a cross-domain feature and nonlinear feature interaction mechanism.

[0006] To achieve the above objectives, this application provides a method for identifying series fault arcing in a low-voltage switchgear, comprising:

[0007] Step S10: Collect the current signal and voltage signal of the low-voltage switchgear outgoing circuit. After preprocessing the collected current signal, use fully adaptive noise ensemble empirical mode decomposition to decompose the preprocessed current signal into several intrinsic mode function components. Step S20: Based on the intrinsic mode function components obtained by decomposition, extract the phase rearrangement entropy, amplitude modulation index and time-frequency clustering index of each intrinsic mode function component; Step S30: Based on the phase rearrangement entropy, amplitude modulation index and time-frequency clustering index of each intrinsic mode function component, a nonlinear interactive fusion is performed to obtain a comprehensive judgment index, and a sliding window adaptive threshold is used to compare with the comprehensive judgment index. Step S40: When the comprehensive judgment index exceeds the adaptive judgment boundary value, the voltage signal and current signal are called to calculate the correlation coefficient between the voltage-current data pair and the pre-stored series fault arc physical model. If the correlation coefficient is greater than the model matching confirmation value, the occurrence of a series fault arc is confirmed, and an alarm signal is output.

[0008] Furthermore, step S10 includes the following detailed steps: Step S101: Acquire the current and voltage signals of the low-voltage switchgear outgoing circuit at a preset sampling frequency; Step S102: Bandpass filtering and noise reduction are performed on the acquired current signal; Step S103: The preprocessed current signal is decomposed using fully adaptive noise ensemble empirical mode decomposition. The specific process is as follows: Different adaptive white noises are added multiple times to the original preprocessed current signal. Empirical mode decomposition is performed on the signal after each addition of white noise. The average value of the first-order intrinsic mode function components obtained from multiple decompositions is taken as the first-order intrinsic mode function component. The residual components are calculated, and adaptive white noise is added to the residual components and decomposed again to obtain the intrinsic mode function components of each order in turn. The above process is repeated until the number of extreme points of the residual components meets the stopping condition, and several intrinsic mode function components are obtained.

[0009] Further, in step S20, the extraction process of the phase rearrangement entropy includes: Perform Hilbert transform on each intrinsic mode function component to construct an analytic signal, and extract the instantaneous amplitude and instantaneous phase of each intrinsic mode function component from the analytic signal; The instantaneous phase sequences of each intrinsic mode function component are reconstructed and mapped to an m-dimensional phase space, where m is a preset embedding dimension. In the phase space, a preset distance threshold is used to group state points with a distance less than the threshold into the same replay state. The occurrence probability of each replay state is calculated, and the phase rearrangement entropy of each intrinsic mode function component is calculated according to the following formula:

[0010] in, For the first Phase rearrangement entropy of each eigenmode function component For the first phase space The probability of a recurring state occurring. The total number of replay states; the phase rearrangement entropy is used to characterize the replay mode of the random reignition characteristics of the series fault arc in the phase space.

[0011] The extraction process of the amplitude modulation index includes: The instantaneous amplitude sequences of each eigenmode function component obtained using the Hilbert transform are used to calculate the amplitude modulation index of each eigenmode function component using the following formula:

[0012] in, For the first The amplitude modulation index of each intrinsic mode function component For the first The instantaneous amplitude sequence of each intrinsic mode function component; , and These represent the functions for taking the maximum, minimum, and mean values, respectively; the amplitude modulation index is used to characterize the degree of current amplitude modulation caused by the series fault arc.

[0013] The extraction process of the time-frequency clustering index includes: Short-time Fourier transforms are performed on each intrinsic mode function component to obtain the time-frequency distribution matrix of each intrinsic mode function component. Based on the time-frequency distribution matrix, the ratio of the energy peak region in the time-frequency plane to the total energy is calculated as an index of the time-frequency clustering of each intrinsic mode function component.

[0014] in, For the first Time-frequency clustering index of each intrinsic mode function component, For the first The energy of the energy peak region on the time-frequency plane of each eigenmode function component. For the first The total energy of the intrinsic mode function components in the time-frequency plane; the time-frequency clustering index is used to characterize the degree of clustering of the arc energy of the series fault in the time-frequency plane.

[0015] Furthermore, step S30 includes the following detailed steps: Step S301: Select fault-sensitive intrinsic mode function components from all intrinsic mode function components obtained by decomposition. The selection criteria are the energy proportion, variance contribution rate, or deviation from the adaptive judgment boundary value of each intrinsic mode function component. Step S302: For each selected intrinsic mode function component, its phase rearrangement entropy, amplitude modulation index, and time-frequency clustering index are nonlinearly fused to obtain the nonlinear interactive characteristic value of the intrinsic mode function component. The nonlinear interactive fusion is performed according to the following formula:

[0016] in, For the first Nonlinear interactive eigenvalues ​​of each intrinsic mode function component , and They represent the first Phase rearrangement entropy, amplitude modulation index, and time-frequency clustering index of each intrinsic mode function component. This represents the preset coupling coefficient; Step S303: The nonlinear interactive eigenvalues ​​of each selected intrinsic mode function component are weighted and summed to obtain a comprehensive judgment index. Step S304: Set a sliding window, construct a historical feature sequence using the comprehensive judgment indicators of several times before the current time, calculate the mean and standard deviation of the historical feature sequence, and determine the adaptive judgment boundary value based on the mean and standard deviation. Step S305: Compare the current comprehensive judgment index with the adaptive judgment boundary value. If the current comprehensive judgment index is greater than the adaptive judgment boundary value, then execute the subsequent physical model verification step; otherwise, determine that it is in normal operation state and return to step S10 to continue monitoring.

[0017] Furthermore, step S40 includes the following detailed steps: Step S401: When the comprehensive judgment index exceeds the adaptive judgment boundary value, physical model verification is triggered, and a verification window with a preset duration is entered. Step S402: In the verification window, call the voltage signal and synchronously acquired current signal collected and stored in step S10, extract voltage-current data pairs from multiple sampling points near the current zero-crossing point, and obtain the measured voltage value. Step S403: Call the pre-stored series fault arc physical model and obtain the reference voltage value of the physical model under the corresponding current value at each sampling point; Step S404: Calculate the correlation coefficient between the measured voltage value and the reference voltage value according to the following formula. :

[0018] in, For the first The measured voltage values ​​at each sampling point For the physical model in the first The reference voltage value corresponding to the current value at each sampling point; To verify the average measured voltage of multiple sampling points within the verification window, To verify the average reference voltage value corresponding to the current value at multiple sampling points within the verification window; This represents the total number of sampling points; Step S405: If the correlation coefficient is greater than the preset model matching confirmation value, then a series fault arc is confirmed and an alarm signal is output; otherwise, it is determined to be a false alarm and the process returns to step S10 to continue monitoring.

[0019] This application also provides a series fault arc identification system for low-voltage switchgear, including: Signal acquisition, processing and mode decomposition module: used to acquire current and voltage signals of the low-voltage switchgear outgoing circuit. After preprocessing the acquired current signal, the fully adaptive noise ensemble empirical mode decomposition is used to decompose the preprocessed current signal into several intrinsic mode function components. Multidimensional feature extraction module: used to extract the phase rearrangement entropy, amplitude modulation index and time-frequency clustering index of each intrinsic mode function component based on the intrinsic mode function components obtained by decomposition; The comprehensive judgment index calculation module is used to perform nonlinear interactive fusion based on the phase rearrangement entropy, amplitude modulation index and time-frequency clustering index of each intrinsic mode function component to obtain a comprehensive judgment index, and compares it with the comprehensive judgment index using a sliding window adaptive threshold. Physical model verification module: When the comprehensive judgment index exceeds the adaptive judgment boundary value, it calls the voltage signal and current signal to calculate the correlation coefficient between the voltage-current data pair and the pre-stored series fault arc physical model. If the correlation coefficient is greater than the model matching confirmation value, it confirms that a series fault arc has occurred and outputs an alarm signal.

[0020] This application also provides a series fault arc identification device for a low-voltage switchgear. The series fault arc identification device for a low-voltage switchgear includes: a memory, a processor, and a series fault arc identification program for a low-voltage switchgear stored in the memory and executable on the processor. When the series fault arc identification program for a low-voltage switchgear is executed by the processor, it implements the above-described method.

[0021] This application also provides a computer program product, which includes a series fault arc identification program for low-voltage switchgear, wherein the series fault arc identification program for low-voltage switchgear implements the above-described method when executed by a processor.

[0022] This application discloses the following technical effects: This application provides a method and system for identifying series fault arcs in low-voltage switchgear. It introduces phase space rearrangement analysis into low-voltage arc detection for the first time and specifically modifies the method to address the random reignition physical characteristics of series arcs. Phase rearrangement entropy, by statistically analyzing the distribution of states with replay properties in phase space, can sensitively capture phase disturbances caused by random arc reignition. These disturbances are already apparent during the slow degradation stage of contact resistance, enabling the invention to issue an early warning before the actual occurrence of a fault arc.

[0023] Unlike existing technologies that simply superimpose features using a linear weighted fusion method, this invention constructs a nonlinear interactive fusion of phase rearrangement entropy, amplitude modulation exponent, and time-frequency clustering index. Under normal operating conditions, all three features are in the low-value region, and their product remains at an extremely low level. When a series fault arc occurs, all three respond simultaneously, and their product grows exponentially, resulting in a detection margin exceeding that of the linear fusion method. Furthermore, it maintains a high level of redundancy detection capability even when any one feature fails.

[0024] Unlike existing technologies where purely data-driven methods lack physical constraints and are susceptible to nonlinear load interference, and where purely physical model methods are limited in their inability to cover complex working conditions, this invention combines data-driven sliding window adaptive threshold preliminary judgment with secondary verification of the electric arc physical model. This creates a collaborative division of labor between the two levels, enabling fast screening and precise confirmation, while also ensuring both real-time performance and reliability. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating a method for identifying series fault arcs in a low-voltage switchgear, as provided in an embodiment of this application.

[0026] Figure 2 This is a schematic diagram of a series fault arc identification system for a low-voltage switchgear provided in an embodiment of this application. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0028] Example 1: This application provides a method for identifying series fault arcing in a low-voltage switchgear, applied to a 380V / 50Hz low-voltage distribution cabinet, such as... Figure 1 As shown, the method includes: Step S10: Collect the current and voltage signals of the low-voltage switchgear outgoing circuit. After preprocessing the collected current signal, use fully adaptive noise ensemble empirical mode decomposition to decompose the preprocessed current signal into several intrinsic mode function components.

[0029] In this embodiment, open-type current transformers and voltage transformers are installed in the outgoing circuit of the low-voltage switchgear. The current transformer has a transformation ratio of 1000:1 and an accuracy class of 0.5, and is used to collect line current signals; the voltage transformer has a transformation ratio of 380:10 and an accuracy class of 0.5, and is used to collect line voltage signals. Both signals are processed by a signal conditioning circuit (including an amplification circuit and an anti-aliasing filter circuit) and then synchronously acquired by a data acquisition card at a sampling frequency of 200kHz, collecting 4000 data points per power frequency cycle (20ms).

[0030] The acquired current signal is bandpass filtered with a passband frequency range of 50Hz to 10kHz to remove DC components and high-frequency noise above 10kHz. The filter uses a finite impulse response (FIR) digital filter with a filter order of 100 and a Hamming window design to ensure passband flatness and linear phase characteristics.

[0031] The preprocessed current signal is subjected to fully adaptive ensemble noise empirical mode (CEEMDAN) decomposition with the following parameters: number of times white noise is added. =100, noise amplitude coefficient =0.2, maximum number of iterations 500, screening stop condition is that the standard deviation of two consecutive screening results is less than 0.01; decomposed to obtain 8 intrinsic mode function (IMF) components. arrive and residual components ,in and Corresponding to high-frequency noise components and local transient impacts, to The corresponding fault characteristics are mainly concentrated in the mid-frequency band. and Corresponding to the power frequency and its lower harmonic components. Experiments show that under normal operating conditions, the energy of each IMF component is mainly distributed in the low-frequency range ( and ), mid-to-high frequency band ( to The energy proportion of the IMF component in the mid-to-high frequency band is extremely small; however, under the condition of a series fault arc, due to the broadband random disturbance introduced by the arc, the energy of the IMF component in the mid-to-high frequency band increases significantly. to The proportion of energy in the component increases from less than 5% under normal operating conditions to 20% to 35% under fault conditions.

[0032] Step S20: Based on the intrinsic mode function components obtained by decomposition, extract the phase rearrangement entropy, amplitude modulation index and time-frequency clustering index of each intrinsic mode function component.

[0033] In this embodiment, a Hilbert transform is performed on each IMF component to construct an analytic signal, and the instantaneous amplitude and instantaneous phase of each IMF component are extracted from the analytic signal. Taking the fourth IMF component as an example... For example, perform a Hilbert transform on it:

[0034] in, for The Hilbert transform result, For the current moment, Let time be the integral variable; the above equation is the definition of the Hilbert transform, which is obtained by Cauchy principal value integration of the real signal. Transform it into its orthogonal components Based on this orthogonal component, an analytic signal is constructed. : ,in The imaginary unit, Instantaneous amplitude, For instantaneous phase, Phase factor in complex exponential form; instantaneous amplitude and instantaneous phase They are obtained from the following formulas respectively:

[0035]

[0036] in, This represents the arctangent function, whose principal value interval is... ;because The principal value range of a function needs to be restricted. and The symbols are dewound in phase to obtain a continuous instantaneous phase sequence. This embodiment employs an adaptive phase dewound algorithm, which detects phase jumps between adjacent sampling points exceeding a certain threshold. The position is compensated to restore the true continuous phase evolution. The above steps are repeated to obtain the instantaneous phase sequence of each IMF component. .

[0037] The instantaneous phase sequences of each IMF component are reconstructed and mapped to the phase space. 3D phase space, in which This is a preset embedding dimension. In this embodiment, for the first... Instantaneous phase sequence of IMF components Take the embedding dimension Delay time = 2 sampling points, corresponding to 10μs. The specific operation of phase space reconstruction is as follows: for a length of Instantaneous phase sequence Reconstruct it as:

[0038] in, Table system in A state point at time t, all state points constitute the system at time t. The trajectory in phase space; in this invention, the system refers to the dynamic system corresponding to the instantaneous phase sequence of each IMF component obtained after the low-voltage switchgear outgoing circuit current signal is decomposed by CEEMDAN, that is, the object of analysis is the phase evolution behavior of each IMF component itself. For the 4000 sampling points of each power frequency cycle in this embodiment, approximately 3994 state points are obtained after phase space reconstruction.

[0039] In phase space, a preset distance threshold is defined. When two state points and The Euclidean distance between them is less than At that time, determine the next state point. The corresponding state is the previous state point. The corresponding state is replayed. In this embodiment, the distance threshold is set to... ,in For the first The standard deviation of the instantaneous phase sequence of each IMF component is set so that approximately 5% to 15% of state point pairs can be identified as recurring events, making it suitable for unified analysis under different load conditions.

[0040] All related replay events are clustered into several replay regions based on their spatial location, with each replay region corresponding to a replay state. The specific clustering method is as follows: using the first unvisited replay event as the initial seed, a breadth-first search is used to traverse all events whose spatial distance to the seed event is less than or equal to the seed event. The events are used to form the first cluster; this process is repeated until all replayed events have been visited, resulting in R cluster regions, each corresponding to a replayed state; the probability of each replayed state is then calculated. Define it as falling into the first The ratio of the number of phase points in each replay region to the total number of phase points.

[0041] The phase rearrangement entropy is calculated using the following formula: ,in For the first Phase rearrangement entropy of each IMF component For the first phase space The probability of a recurring state occurring. The total number of replay states; the unit of phase rearrangement entropy is nat. The larger the value, the more dispersed the replay states in the phase space, and the more complex and random the dynamic behavior of the system; the smaller the value, the more concentrated the replay states, and the more regular and deterministic the dynamic behavior of the system.

[0042] Under normal operating conditions of the low-voltage switchgear, the current signal strictly follows the periodic variation of the power frequency, and the instantaneous phase of each IMF component evolves according to a fixed pattern. Taking a power frequency period of T=20ms as an example, the instantaneous phase values ​​at the same moment in adjacent periods are approximately equal, that is... In phase space, this periodicity manifests as state points repeatedly cycling along a closed trajectory, only repeatedly passing through a few fixed replay regions. Therefore, the replay states are highly concentrated in a very small number of regions. The distribution is extremely uneven, at which point the phase rearrangement entropy... At a relatively low level. When a series fault arc occurs in the low-voltage switchgear, the nonlinear time-varying characteristics of the arc voltage exert a continuous random disturbance on the current within each half-cycle of the power frequency. The physical process of the series arc is a cycle of "ignition → combustion → extinction → reignition," and its reignition time and combustion intensity are random, directly causing unpredictable fluctuations in the instantaneous phase of each IMF component, i.e. Instead of approaching zero, the phase recursion exhibits a random distribution. In phase space, this random phase fluctuation manifests as state points no longer cycling along fixed trajectories but being randomly scattered across a large number of regions in the phase space. The clustering results of replayed events are no longer concentrated in a very few regions, the replayed states tend to be dispersed, and the phase rearrangement entropy increases. It then increased significantly.

[0043] Traditional sample entropy and permutation entropy only statistically analyze the probability distribution of state points in phase space, without distinguishing whether state points have recurrence properties. That is, they do not differentiate whether the system randomly accesses certain regions or repeatedly accesses the same region at specific times. For the normal operation of nonlinear loads (such as frequency converters and switching power supplies), the current signal itself has aperiodic characteristics. However, this aperiodicity is constrained by the deterministic nature of the circuit topology and control strategy. Although the state points are widely distributed in phase space, the access patterns still revolve around several deterministic trajectory clusters, resulting in a complete recurrence structure and a low level of phase rearrangement entropy. In contrast, the phase disturbance generated by the random reignition of a series fault arc is truly random and nondeterministic, directly disrupting the recurrence structure of state points. State points are no longer distributed along deterministic trajectory clusters but are randomly scattered. Phase rearrangement entropy, by specifically statistically analyzing states with recurrence properties, can effectively distinguish between these two scenarios. During normal operation of a nonlinear load, although the state points are dispersed, an identifiable recurrence structure exists. Lower; when a series fault arc occurs, the replay structure is destroyed. The efficiency is significantly increased. This characteristic makes the present invention highly robust to normal operation under nonlinear loads.

[0044] For this embodiment The components, under normal operating conditions, were obtained through the above cluster analysis. There are recurring states, with a probability distribution of . =0.72, =0.18, =0.07, =0.03; calculated as follows 0.92. Under the condition of series fault arcing, cluster analysis yielded the following results. The repeated states have a relatively dispersed probability distribution, with the maximum probability being approximately 0.12. The calculated... 0.85, approximately three times that under normal operating conditions; repeating the above process, the phase rearrangement entropy of each IMF component is obtained. Experimental statistics show that under normal operating conditions, the phase rearrangement entropy distribution of each IMF component is relatively uniform, with an average value of 0.85–1.20; under series fault arc conditions, the mid-frequency IMF component ( to The phase rearrangement entropy of the components increased significantly to 2.50–3.20, while the low-frequency components ( and The impact of the fault is relatively small, and the change in entropy value is not obvious.

[0045] Instantaneous amplitude sequences of each IMF component obtained using Hilbert transform The amplitude modulation index of each IMF component is calculated using the following formula:

[0046] in, For the first Amplitude modulation index of each IMF component, For the first Instantaneous amplitude sequence of IMF components; , and The maximum, minimum, and arithmetic mean of the instantaneous amplitude sequence within one power frequency cycle are given respectively; the amplitude modulation index reflects the relative fluctuation of the instantaneous amplitude within one power frequency cycle: under normal operating conditions, the instantaneous amplitudes of each IMF component are relatively stable. The value is relatively small; when a series fault arc occurs, due to the time-varying characteristics of the arc resistance, the current amplitude undergoes significant periodic modulation within each half-cycle of the power frequency. The value increased significantly.

[0047] For this embodiment Under normal operating conditions, the difference between the maximum and minimum instantaneous amplitude within one power frequency cycle is approximately 20% of the mean. Under series fault arcing conditions, the fluctuation of instantaneous amplitude increases significantly, with the difference between the maximum and minimum values ​​reaching 120% of the mean. Experimental statistics show that under normal working conditions to The amplitude modulation index of the components averages 0.15–0.35; under series fault arcing conditions, to The amplitude modulation index of the component increases to 0.8 to 2.0, which is about 3 to 6 times that under normal operating conditions.

[0048] Short-time Fourier transform (STFT) was performed on each IMF component with a window length of 256 sampling points and an overlap rate of 50%. A Hanning window was used to reduce spectral leakage, resulting in the time-frequency distribution matrix of each IMF component. , indicating the first Each IMF component at time... ,frequency Complex spectral values ​​at that location, Corresponding energy density; based on the time-frequency distribution matrix, calculate the ratio of energy peak regions on the time-frequency plane to total energy, as an indicator of the time-frequency clustering of each IMF component: ,in For the first Time-frequency clustering index of each IMF component For the first The energy of the energy peak region on the time-frequency plane of each IMF component For the first The total energy of the time-frequency plane of each IMF component is calculated using the following formula:

[0049] in, Indicates all moments The maximum value is used for calculation; the time-frequency clustering index reflects the degree of clustering of fault energy in the time-frequency plane. The larger the value, the more concentrated the energy is at a certain moment; the smaller the value, the more uniform the distribution of energy in the time-frequency plane.

[0050] For this embodiment Under normal operating conditions, the time-frequency energy distribution is relatively uniform. .08; Under the condition of series fault arcing, due to the amplitude modulation and phase disturbance caused by the arc, the energy is significantly concentrated in a specific frequency band and phase range. The value increased to 0.25–0.40. Experimental statistics show that under normal operating conditions, to The mean time-frequency clustering index of the components is 0.05–0.12; under the condition of series fault arcing, to The time-frequency clustering index of the component increased to 0.20–0.45.

[0051] Step S30: Based on the phase rearrangement entropy, amplitude modulation index and time-frequency clustering index of each intrinsic mode function component, a nonlinear interactive fusion is performed to obtain a comprehensive judgment index, and a sliding window adaptive threshold is used to compare it with the comprehensive judgment index.

[0052] In this embodiment, fault-sensitive IMF components are selected from the eight IMF components obtained by decomposition. The selection criterion is the degree of deviation between the energy proportion of each IMF component and the adaptive judgment boundary value. The energy proportion is expressed as the ratio of the total energy of each IMF component to the sum of the energies of all eight IMF components. A comparative analysis of the energy proportions of each IMF component under normal and fault conditions is performed. The results show that... to The energy proportions of the four components increase most significantly under fault conditions compared to normal conditions, while , The main component is high-frequency noise, and its energy percentage fluctuates significantly due to fault conditions, but the absolute value is extremely small. , The component is a low-frequency power frequency component, and its energy proportion does not change significantly before and after the fault. Therefore, this embodiment selects... to Four fault-sensitive IMF components participate in subsequent fusion. The weighting coefficients are determined based on the fault sensitivity of each component, i.e., the ratio of its eigenvalues ​​(phase rearrangement entropy, amplitude modulation index, and time-frequency clustering index) under fault conditions to those under normal conditions, and are 0.2, 0.35, 0.3, and 0.15, respectively. The component has the highest weight because it exhibits the highest fault sensitivity across all three feature dimensions.

[0053] For each selected IMF component, its phase rearrangement entropy, amplitude modulation index, and time-frequency clustering index are nonlinearly fused. In this embodiment, the coupling coefficient... The value of is 1.0, and the nonlinear interaction eigenvalue of this IMF component is obtained according to the following formula: in, For the first The nonlinear interactive eigenvalues ​​of the IMF components are designed based on the principle that: under normal conditions, all three characteristics are in a low-value region, and their product is very small; during a fault, all three values ​​increase simultaneously, and in addition... The amplification effect makes It is growing exponentially.

[0054] Taking the data in this embodiment as an example, The three eigenvalues ​​of the component under normal operating conditions are respectively , , , Under fault conditions, respectively , , , The nonlinear interactive characteristic value under fault conditions is about 22.7 times that under normal conditions, forming a significant detection margin, which is far superior to the linear weighted fusion method.

[0055] The nonlinear interaction eigenvalues ​​of each selected IMF component are weighted and summed to obtain a comprehensive evaluation index. : ,in , and They are respectively , and The nonlinear interaction eigenvalues ​​of the three IMF components; in this embodiment, under normal operating conditions, the nonlinear interaction eigenvalues ​​of each IMF component are all less than 0.3, which is a comprehensive judgment index. Approximately 0.15–0.25; under fault conditions, the nonlinear interaction eigenvalues ​​of each IMF component are all greater than 1.5, and the comprehensive judgment index is... It is approximately 3.5 to 5.0.

[0056] Set the length of the sliding window =100 power frequency cycles, based on the current time The comprehensive judgment indicators from the previous 100 moments constitute a historical feature sequence, and the mean of this sequence is calculated. and standard deviation Determine the adaptive decision boundary value for: , This is the threshold coefficient (dimensionless), and its value is based on the assumption that the historical feature sequence follows a normal distribution. The corresponding cumulative probability is 99.98%, meaning that under normal operation and without faults, the theoretical probability of the comprehensive judgment index exceeding the adaptive judgment boundary value is approximately 0.02%, effectively controlling the false alarm rate. This adaptive judgment boundary value is dynamically adjusted according to load conditions: when a change in load causes an overall shift in the comprehensive judgment index, the sliding window synchronously updates the historical feature sequence. and As it changes, Automatically tracks load changes.

[0057] Compare the current Comprehensive judgment indicators of time With adaptive decision boundary value :like If the comprehensive judgment index at the current moment deviates significantly from the historical normal level, proceed to step S40 to execute the subsequent physical model verification step; otherwise, it is judged to be in normal operation state, return to step S10 to continue monitoring, slide the sliding window forward one moment, include the comprehensive judgment index at the current moment and remove the data from the earliest moment, and keep the window length constant at 100.

[0058] In this embodiment, under normal operating conditions... and The ratio is approximately 0.6 to 0.9, providing ample margin; when a series fault arc occurs, It rises rapidly to [a certain value] within 3–5 power frequency cycles (60–100 ms). The value is 1.5 to 3 times higher, triggering subsequent verification processes.

[0059] Step S40: When the comprehensive judgment index exceeds the adaptive judgment boundary value, the voltage signal and current signal are called to calculate the correlation coefficient between the voltage-current data pair and the pre-stored series fault arc physical model. If the correlation coefficient is greater than the model matching confirmation value, the occurrence of a series fault arc is confirmed, and an alarm signal is output.

[0060] When the comprehensive judgment index in step S30 exceeds the adaptive judgment boundary value, physical model verification is triggered, and the verification window is entered. In this embodiment, the duration of the verification window is 3 power frequency cycles (60ms). The reason for choosing 3 power frequency cycles is that the physical process of the series arc presents a complete cycle of "ignition → combustion → extinguishing" within each power frequency half-cycle. 3 power frequency cycles cover 6 half-cycles, which is sufficient to reliably identify the steady-state characteristics of the arc. If the verification is triggered due to a sudden load change or external interference, but no series fault arc actually occurs, it will be judged as a false alarm after the verification window ends (60ms) and normal monitoring will resume. Therefore, the setting of the verification window will not have a significant impact on the real-time monitoring performance of the system.

[0061] Within the verification window, the voltage signal and the synchronously acquired current signal collected and stored in step S10 are invoked. In this embodiment, the data acquisition card collects and caches the voltage and current data of the most recent 500ms in real time at a sampling rate of 200kHz. When verification is triggered, the system directly retrieves the data segment of 3 power frequency cycles (60ms) corresponding to the verification window from the cache, without the need for re-collection.

[0062] Voltage-current data pairs are extracted from 100 sampling points near the current zero-crossing point. The method for detecting the current zero-crossing point is as follows: detect adjacent points where the sign of the current sampling value changes, and determine the precise location of the zero-crossing point through linear interpolation. Taking the zero-crossing point as the center, 50 sampling points are taken forward and 50 points backward (corresponding to 0.5 ms, covering approximately 9° of electrical angle before and after the zero-crossing point). The instantaneous voltage and instantaneous current values ​​of these sampling points are extracted to form voltage-current data pairs. ,in For the first The measured voltage values ​​at each sampling point For the first The measured current values ​​at each sampling point. The volt-ampere characteristics of the series arc exhibit the most significant nonlinear features near the current zero-crossing point (voltage distortion before and after the zero-crossing point), therefore this region is the optimal analysis window.

[0063] The pre-stored physical model of a series fault arc is invoked to obtain the reference voltage value of the physical model at the corresponding current value at each sampling point. In this embodiment, the pre-stored physical model of the series fault arc is an improved Mayr arc model. The Mayr arc model is a classic mathematical model describing the dynamic characteristics of an arc. Its basic assumptions are: the heat dissipation power of the arc channel is constant, and the change of arc conductance over time is determined by the difference between the instantaneous arc power and the heat dissipation power.

[0064] For detecting arc faults in series, the model needs to be solved in conjunction with the line parameters to obtain the theoretical voltage waveform under a given current waveform. In this embodiment, the Mayr arc model is expressed by the following formula:

[0065] in, It is the conductance of the electric arc, and , Arc voltage It is the arc current. The instantaneous power of the electric arc. is the arc time constant, a parameter characterizing the thermal inertia of the arc; in this embodiment, it is taken as 2μs. The heat dissipation power is the rate at which the electric arc dissipates heat to the surrounding environment; in this embodiment, it is taken as 50W. Let be the time derivative of the arc conductance. This differential equation describes how the arc conductance changes with the difference between instantaneous power and heat dissipation power: when the instantaneous power is greater than the heat dissipation power, the conductance increases (arc strengthens); conversely, the conductance decreases (arc weakens). The Cassie arc model can also be used as an alternative, which assumes that the arc channel diameter varies with the current and is suitable for high-current arc scenarios.

[0066] The measured current sequence within the verification window is substituted into the above model, and the fourth-order Runge-Kutta method is used for numerical solution to obtain the reference voltage value corresponding to each sampling point. Calculate the measured voltage value Compared with reference voltage value Correlation coefficient between :

[0067] in, To verify the average measured voltage of 100 sampling points within the verification window, To verify the average reference voltage value corresponding to the current value at 100 sampling points within the verification window; The value range of is [-1, 1]. When When the measured voltage waveform approaches 1, it closely matches the theoretical waveform of the electric arc physical model, indicating that the measured voltage-current characteristics are consistent with the nonlinear resistance characteristics of the electric arc, thus confirming it as a real electric arc; when When the value is below 0, it indicates that the measured waveform does not match the arc model and is judged as a false alarm.

[0068] Set model matching confirmation value =0.8, this value was selected based on statistical results from a large amount of experimental data: under real arc conditions, the correlation coefficient is usually greater than 0.85, while under non-arc conditions, the correlation coefficient is usually less than 0.60. A threshold of 0.8 effectively suppresses false alarms while ensuring a high detection rate. (Comparison of correlation coefficients) Confirmation value for model matching: If If a series fault arc is detected, an alarm signal is output. The alarm signal is uploaded to the background monitoring system via RS485 communication, simultaneously activating the audible and visual alarm and triggering the circuit breaker to disconnect the faulty circuit; otherwise, it is determined to be a false alarm, and the process returns to step S10 to continue monitoring.

[0069] Example 2: The series fault arc identification system for low-voltage switchgear provided in this embodiment of the invention can execute the series fault arc identification method for low-voltage switchgear provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method, such as... Figure 2 As shown, it includes: Signal acquisition, processing and mode decomposition module: used to acquire current and voltage signals of the low-voltage switchgear outgoing circuit. After preprocessing the acquired current signal, the fully adaptive noise ensemble empirical mode decomposition is used to decompose the preprocessed current signal into several intrinsic mode function components. Multidimensional feature extraction module: used to extract the phase rearrangement entropy, amplitude modulation index and time-frequency clustering index of each intrinsic mode function component based on the intrinsic mode function components obtained by decomposition; The comprehensive judgment index calculation module is used to perform nonlinear interactive fusion based on the phase rearrangement entropy, amplitude modulation index and time-frequency clustering index of each intrinsic mode function component to obtain a comprehensive judgment index, and compares it with the comprehensive judgment index using a sliding window adaptive threshold. Physical model verification module: When the comprehensive judgment index exceeds the adaptive judgment boundary value, it calls the voltage signal and current signal to calculate the correlation coefficient between the voltage-current data pair and the pre-stored series fault arc physical model. If the correlation coefficient is greater than the model matching confirmation value, it confirms that a series fault arc has occurred and outputs an alarm signal.

[0070] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0071] Example 3: This application provides a series fault arc identification device for a low-voltage switchgear. The series fault arc identification device for a low-voltage switchgear includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the series fault arc identification method for a low-voltage switchgear described in Example 1 above.

[0072] Example 4: This application provides a computer program product including a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication system, or installed from a storage system. When the computer program is executed by a processing system, it performs the functions defined in the method of Example 1 of this application.

[0073] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A method for identifying series fault arcs in a low-voltage switchgear, characterized in that, The method includes: Step S10: Collect the current signal and voltage signal of the low-voltage switchgear outgoing circuit. After preprocessing the collected current signal, use fully adaptive noise ensemble empirical mode decomposition to decompose the preprocessed current signal into several intrinsic mode function components. Step S20: Based on the intrinsic mode function components obtained by decomposition, extract the phase rearrangement entropy, amplitude modulation index and time-frequency clustering index of each intrinsic mode function component; Step S30: Based on the phase rearrangement entropy, amplitude modulation index, and time-frequency clustering index of each intrinsic mode function component, a nonlinear interactive fusion is performed to obtain a comprehensive judgment index. An adaptive judgment boundary value is then used to compare this comprehensive judgment index with the obtained index. Step S301: Select fault-sensitive intrinsic mode function components from all intrinsic mode function components obtained by decomposition. The selection criteria are the energy proportion, variance contribution rate, or deviation from the adaptive judgment boundary value of each intrinsic mode function component. Step S302: For each selected intrinsic mode function component, its phase rearrangement entropy, amplitude modulation index, and time-frequency clustering index are nonlinearly fused to obtain the nonlinear interactive characteristic value of the intrinsic mode function component. The nonlinear interactive fusion is performed according to the following formula: in, For the first Nonlinear interactive eigenvalues ​​of each intrinsic mode function component , and They represent the first Phase rearrangement entropy, amplitude modulation index, and time-frequency clustering index of each intrinsic mode function component. This represents the preset coupling coefficient; Step S303: The nonlinear interactive eigenvalues ​​of each selected intrinsic mode function component are weighted and summed to obtain a comprehensive judgment index. Step S304: Set a sliding window, construct a historical feature sequence using the comprehensive judgment indicators of several times before the current time, calculate the mean and standard deviation of the historical feature sequence, and determine the adaptive judgment boundary value based on the mean and standard deviation. Step S305: Compare the current comprehensive judgment index with the adaptive judgment boundary value. If the current comprehensive judgment index is greater than the adaptive judgment boundary value, then execute the subsequent physical model verification step; otherwise, determine that it is in normal operation state and return to step S10 to continue monitoring. Step S40: When the comprehensive judgment index exceeds the adaptive judgment boundary value, the voltage signal and current signal are called to calculate the correlation coefficient between the voltage-current data pair and the pre-stored series fault arc physical model. If the correlation coefficient is greater than the model matching confirmation value, the occurrence of a series fault arc is confirmed, and an alarm signal is output.

2. The method for identifying series fault arcs in a low-voltage switchgear as described in claim 1, characterized in that, Step S10 includes the following detailed steps: Step S101: Acquire the current and voltage signals of the low-voltage switchgear outgoing circuit at a preset sampling frequency; Step S102: Bandpass filtering and noise reduction are performed on the acquired current signal; Step S103: The preprocessed current signal is decomposed using fully adaptive noise ensemble empirical mode decomposition. The specific process is as follows: Different adaptive white noises are added to the original preprocessed current signal multiple times. Empirical mode decomposition is performed on the signal after each addition of white noise. The average value of the first-order intrinsic mode function components obtained from multiple decompositions is taken as the first-order intrinsic mode function components. Calculate the residual components, add adaptive white noise to the residual components and decompose them to obtain the eigenmode function components of each order in turn; repeat the above process until the number of extreme points of the residual components meets the stopping condition, and obtain several eigenmode function components.

3. The method for identifying series fault arcs in a low-voltage switchgear as described in claim 1, characterized in that, In step S20, the extraction process of the phase rearrangement entropy includes: Perform Hilbert transform on each intrinsic mode function component to construct an analytic signal, and extract the instantaneous amplitude and instantaneous phase of each intrinsic mode function component from the analytic signal; The instantaneous phase sequences of each intrinsic mode function component are reconstructed and mapped to an m-dimensional phase space, where m is a preset embedding dimension. In the phase space, a preset distance threshold is used to group state points with a distance less than the threshold into the same replay state. The occurrence probability of each replay state is calculated, and the phase rearrangement entropy of each intrinsic mode function component is calculated according to the following formula: in, For the first Phase rearrangement entropy of each eigenmode function component For the first phase space The probability of a recurring state occurring. The total number of replay states; the phase rearrangement entropy is used to characterize the replay mode of the random reignition characteristics of the series fault arc in the phase space.

4. The method for identifying series fault arcs in a low-voltage switchgear as described in claim 1, characterized in that, In step S20, the extraction process of the amplitude modulation index includes: The instantaneous amplitude sequences of each eigenmode function component obtained using the Hilbert transform are used to calculate the amplitude modulation index of each eigenmode function component using the following formula: in, For the first The amplitude modulation index of each intrinsic mode function component For the first The instantaneous amplitude sequence of each intrinsic mode function component; , and These represent the functions for taking the maximum, minimum, and mean values, respectively; the amplitude modulation index is used to characterize the degree of current amplitude modulation caused by the series fault arc.

5. The method for identifying series fault arcs in a low-voltage switchgear as described in claim 1, characterized in that, In step S20, the extraction process of the time-frequency clustering index includes: Short-time Fourier transforms are performed on each intrinsic mode function component to obtain the time-frequency distribution matrix of each intrinsic mode function component. Based on the time-frequency distribution matrix, the ratio of the energy peak region in the time-frequency plane to the total energy is calculated as an index of the time-frequency clustering of each intrinsic mode function component. in, For the first Time-frequency clustering index of each intrinsic mode function component, For the first The energy of the energy peak region on the time-frequency plane of each eigenmode function component. For the first The total energy of the intrinsic mode function components in the time-frequency plane; the time-frequency clustering index is used to characterize the degree of clustering of the arc energy of the series fault in the time-frequency plane.

6. The method for identifying series fault arcs in a low-voltage switchgear as described in claim 1, characterized in that, Step S40 includes the following detailed steps: Step S401: When the comprehensive judgment index exceeds the adaptive judgment boundary value, physical model verification is triggered, and a verification window with a preset duration is entered. Step S402: In the verification window, call the voltage signal and synchronously acquired current signal collected and stored in step S10, extract voltage-current data pairs from multiple sampling points near the current zero-crossing point, and obtain the measured voltage value. Step S403: Call the pre-stored series fault arc physical model and obtain the reference voltage value of the physical model under the corresponding current value at each sampling point; Step S404: Calculate the correlation coefficient between the measured voltage value and the reference voltage value according to the following formula. : in, For the first The measured voltage values ​​at each sampling point For the physical model in the first The reference voltage value corresponding to the current value at each sampling point; To verify the average measured voltage of multiple sampling points within the verification window, To verify the average reference voltage value corresponding to the current value at multiple sampling points within the verification window; This represents the total number of sampling points; Step S405: If the correlation coefficient is greater than the preset model matching confirmation value, then a series fault arc is confirmed and an alarm signal is output; otherwise, it is determined to be a false alarm and the process returns to step S10 to continue monitoring.

7. A series fault arc identification system for a low-voltage switchgear, characterized in that, The system is used to implement the series fault arc identification method for a low-voltage switchgear according to any one of claims 1-6, and the system comprises: Signal acquisition, processing and mode decomposition module: used to acquire current and voltage signals of the low-voltage switchgear outgoing circuit. After preprocessing the acquired current signal, the fully adaptive noise ensemble empirical mode decomposition is used to decompose the preprocessed current signal into several intrinsic mode function components. Multidimensional feature extraction module: used to extract the phase rearrangement entropy, amplitude modulation index and time-frequency clustering index of each intrinsic mode function component based on the intrinsic mode function components obtained by decomposition; The comprehensive judgment index calculation module is used to perform nonlinear interactive fusion based on the phase rearrangement entropy, amplitude modulation index and time-frequency clustering index of each intrinsic mode function component to obtain a comprehensive judgment index, and compares the comprehensive judgment index with the adaptive judgment boundary value. Physical model verification module: When the comprehensive judgment index exceeds the adaptive judgment boundary value, it calls the voltage signal and current signal to calculate the correlation coefficient between the voltage-current data pair and the pre-stored series fault arc physical model. If the correlation coefficient is greater than the model matching confirmation value, it confirms that a series fault arc has occurred and outputs an alarm signal.

8. A series fault arc identification device for a low-voltage switchgear, characterized in that, The low-voltage switchgear series fault arc identification device includes: a memory, a processor, and a low-voltage switchgear series fault arc identification program stored in the memory and executable on the processor. When the low-voltage switchgear series fault arc identification program is executed by the processor, it implements a low-voltage switchgear series fault arc identification method according to any one of claims 1 to 6.

9. A computer program product, characterized in that, The computer program product includes a series fault arc identification program for low-voltage switchgear. When the series fault arc identification program for low-voltage switchgear is executed by the processor, it implements a series fault arc identification method for low-voltage switchgear according to any one of claims 1 to 6.

Images