Unsupervised low voltage series arc fault periodicity detection method based on normalized energy
Patent Information
- Application Number
- CN202610929768.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-25
- Publication Date
- 2026-11-03
AI Technical Summary
Existing low-voltage AC series arc fault detection methods suffer from small sample size and data imbalance in complex power distribution scenarios, resulting in limited detection accuracy and difficulty in meeting real-time detection requirements.
An unsupervised low-voltage series arc fault cycle detection method based on normalized energy is adopted. Local large-class sets are divided by hierarchical clustering, the average current waveform sequence and residual sequence are calculated, the overall autocorrelation matrix is constructed by adaptive weighting, and the arc fault cycle is detected using normal current samples.
It achieves high-precision and low-complexity arc fault cycle detection, making it suitable for deployment in smart circuit breakers and arc fault protection devices, and meeting the online detection needs in complex power distribution scenarios.
Smart Images
Figure CN123029323A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of arc fault detection technology, and more specifically, to an unsupervised low-voltage series arc fault periodic detection method and system based on normalized energy. Background Technology
[0002] In low-voltage electrical systems, aging insulation, poor contact, and loose connections can easily induce series arc faults. The high temperatures generated by these faults can easily ignite surrounding combustibles, leading to electrical fires and seriously threatening the safety of people and property as well as the stable and reliable operation of the power distribution system. Therefore, developing efficient and accurate low-voltage AC series arc fault detection technology is of significant practical importance and application value for electrical fire prevention and control. Currently, the core principle of low-voltage AC series arc fault detection is based on current sequence analysis, which identifies and determines faults by judging whether the number of arc half-waves in the current sequence meets a preset fault standard.
[0003] Current mainstream methods for detecting low-voltage AC series arc faults include rule-driven and data-driven methods, both of which heavily rely on a large amount of diverse arc fault data. Rule-driven methods require the use of diverse features from a large number of fault samples to construct comprehensive and robust detection criteria, while data-driven methods rely on a large number of high-quality labeled samples to train learning algorithms to obtain detection models with strong generalization capabilities. However, in actual operating conditions, line arc faults are random and sporadic, making it difficult to collect a sufficient number and diversity of fault current samples, thus facing serious problems of small sample sizes and data imbalance. This data-level deficiency directly leads to limited accuracy of the criteria or intelligent detection models established by mainstream detection methods in complex power distribution scenarios, and their generalization ability cannot meet practical needs.
[0004] Furthermore, existing arc fault detection methods mostly rely on complex time-frequency domain feature analysis or intelligent algorithm training for arc period detection, which suffers from high computational complexity and difficulty in engineering implementation, making it difficult to meet the real-time detection requirements in complex power distribution scenarios. Therefore, an arc period detection method that balances reliability and sensitivity is needed. Summary of the Invention
[0005] This invention proposes an unsupervised low-voltage series arc fault cycle detection method and system based on normalized energy to solve the problem of how to efficiently detect arc fault cycles.
[0006] To address the aforementioned problems, according to one aspect of the present invention, an unsupervised low-voltage series arc fault cycle detection method based on normalized energy is provided, the method comprising: Collect multi-cycle line current sequences under normal operating conditions with different loads, and normalize each current sequence. The normalized current sequences are clustered to determine multiple local cluster sets, and the average current waveform sequence of all current sequences within each local cluster set is calculated. Calculate the mean current periodic waveform of all periods in the mean current waveform sequence corresponding to each local major category set, and calculate the residual between the current waveform of each period in the mean current waveform sequence and the corresponding mean current periodic waveform to obtain the residual sequence corresponding to each local major category set. For any local large category set, calculate the weight coefficient corresponding to each period based on the residual sequence corresponding to that local large category set; For any local major category set, based on the weight coefficients and average current waveform sequence corresponding to that local major category set, calculate the overall autocorrelation matrix corresponding to that local major category set; Acquire the current signal to be measured, calculate the similarity between the current signal to be measured and the average current waveform sequence corresponding to each local category set, determine the target local category set corresponding to the current signal to be measured, and calculate the normalized energy value of the current signal to be measured based on the overall autocorrelation matrix corresponding to the target local category set. When the normalized energy value is greater than the preset energy threshold, an arc fault cycle is determined to have occurred.
[0007] Preferably, the calculation of the average current waveform sequence of all current sequences within each local major category includes: For any local large class set S, if all current sequences in that local large class set S are Let the sampled values of all current sequences in any local large class set S be denoted as ,in, , , , K represents the number of current sequences; K represents the period. This represents the number of sampling points per period, then in any local large class set S The average current waveform sequence of the current sequences is: Each element in the sequence The calculation formula is: , in, This represents the average current at the j-th sampling point in the i-th period of the average current waveform sequence. For the local large class set S, the first... The current value at the j-th sampling point of the i-th period of a current sequence.
[0008] Preferably, the residual between the current waveform of each period in the average current waveform sequence and the corresponding mean current period waveform is calculated to obtain the residual sequence corresponding to each local large category set, including: , , in, The first in the average current waveform sequence In the period, the 1st The residual values of each sampling point; This represents the average current value at the j-th sampling point in the i-th period of the average current waveform sequence. is the current value at the j-th sampling point in the current mean periodic waveform; K is the number of periods contained in the residual sequence.
[0009] Preferably, for any local large category set, the weight coefficient corresponding to each period is calculated based on the residual sequence corresponding to that local large category set, including: For any local large class set, calculate the total residual of N residual values in each period of the residual sequence corresponding to that local large class set. ; The total residual value for all periods is normalized, and based on the normalized total residual value for each period, the weighting coefficient corresponding to each period is calculated. ; in, The first in the average current waveform sequence In the period, the 1st The residual value of each sampling point; N is the number of sampling points per period; and These are the total residual values for the i-th period before and after normalization, respectively; The residual sequence is the first The weighting coefficients for each period; K is the number of periods contained in the parameter sequence.
[0010] Preferably, for any local large category set, based on the weighting coefficients and average current waveform sequence corresponding to that local large category set, the overall autocorrelation matrix corresponding to that local large category set is calculated, including: , , in, The first in the average current waveform sequence Periodic vectors The periodic autocorrelation matrix; The residual sequence is the first The weighting coefficients for each period; K is the number of periods contained in the parameter sequence; This is the overall autocorrelation matrix of the average current waveform sequence.
[0011] Preferably, the calculation of the normalized energy value of the measured current signal based on the overall autocorrelation matrix corresponding to the target local category set includes: The overall autocorrelation matrix corresponding to the target local category set is taken as the overall autocorrelation matrix R corresponding to the current signal to be measured. The normalized energy value of the measured current signal is calculated based on the overall autocorrelation matrix corresponding to the measured current signal. ; where y is the periodic current signal containing N sampling points in the current signal to be measured.
[0012] According to another aspect of the present invention, an unsupervised low-voltage series arc fault cycle detection system based on normalized energy is provided, the system comprising: The current sequence acquisition unit is used to acquire multi-cycle line current sequences under normal operating conditions under different loads and to normalize each current sequence. The clustering analysis unit is used to cluster the normalized current sequences, determine multiple local cluster sets, and calculate the average current waveform sequence of all current sequences within each local cluster set. The residual sequence calculation unit is used to calculate the mean current periodic waveform of all periods in the average current waveform sequence corresponding to each local major category set, and to calculate the residual between the current waveform of each period in the average current waveform sequence and the corresponding mean current periodic waveform, so as to obtain the residual sequence corresponding to each local major category set. The weight coefficient calculation unit is used to calculate the weight coefficient corresponding to each period based on the residual sequence corresponding to any local large category set. The overall autocorrelation matrix calculation unit is used to calculate the overall autocorrelation matrix corresponding to any local major category set based on the weight coefficients and average current waveform sequence corresponding to that local major category set. The normalized energy value calculation unit is used to acquire the current signal to be measured, calculate the similarity between the current signal to be measured and the average current waveform sequence corresponding to each local category set, determine the target local category set corresponding to the current signal to be measured, and calculate the normalized energy value of the current signal to be measured based on the overall autocorrelation matrix corresponding to the target local category set. The judgment unit is used to determine the occurrence of an arc fault cycle when the normalized energy value is greater than a preset energy threshold.
[0013] Preferably, the clustering analysis unit calculates the average current waveform sequence of all current sequences within each local cluster set, including: For any local large class set S, if all current sequences in that local large class set S are Let the sampled values of all current sequences in any local large class set S be denoted as ,in, , , , K represents the number of current sequences; K represents the period. This represents the number of sampling points per period, then in any local large class set S The average current waveform sequence of the current sequences is: Each element in the sequence The calculation formula is: , in, This represents the average current at the j-th sampling point in the i-th period of the average current waveform sequence. For the local large class set S, the first... The current value at the j-th sampling point of the i-th period of a current sequence.
[0014] Preferably, the residual sequence calculation unit calculates the residual between the current waveform of each period in the average current waveform sequence and the corresponding current mean period waveform to obtain the residual sequence corresponding to each local class set, including: , , in, The first in the average current waveform sequence In the period, the 1st The residual values of each sampling point; This represents the average current value at the j-th sampling point in the i-th period of the average current waveform sequence. is the current value at the j-th sampling point in the current mean periodic waveform; K is the number of periods contained in the residual sequence.
[0015] Preferably, the weight coefficient calculation unit, for any local large category set, calculates the weight coefficient corresponding to each period based on the residual sequence corresponding to that local large category set, including: For any local large class set, calculate the total residual of N residual values in each period of the residual sequence corresponding to that local large class set. ; The total residual value for all periods is normalized, and based on the normalized total residual value for each period, the weighting coefficient corresponding to each period is calculated. ; in, The first in the average current waveform sequence In the period, the 1st The residual value of each sampling point; N is the number of sampling points per period; and These are the total residual values for the i-th period before and after normalization, respectively; The residual sequence is the first The weighting coefficients for each period; K is the number of periods contained in the parameter sequence.
[0016] Preferably, the overall autocorrelation matrix calculation unit, for any local major category set, calculates the overall autocorrelation matrix corresponding to that local major category set based on the weighting coefficients and average current waveform sequence corresponding to that local major category set, including: , , in, The first in the average current waveform sequence Periodic vectors The periodic autocorrelation matrix; The residual sequence is the first The weighting coefficients for each period; K is the number of periods contained in the parameter sequence; This is the overall autocorrelation matrix of the average current waveform sequence.
[0017] Preferably, the normalized energy value calculation unit calculates the normalized energy value of the measured current signal based on the overall autocorrelation matrix corresponding to the target local category set, including: The overall autocorrelation matrix corresponding to the target local category set is taken as the overall autocorrelation matrix R corresponding to the current signal to be measured. The normalized energy value of the measured current signal is calculated based on the overall autocorrelation matrix corresponding to the measured current signal. ; where y is the periodic current signal containing N sampling points in the current signal to be measured.
[0018] According to another aspect of the present invention, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the steps of an unsupervised low-voltage series arc fault cycle detection method based on normalized energy.
[0019] According to another aspect of the present invention, the present invention provides an electronic device, comprising: The aforementioned computer-readable storage medium; and One or more processors for executing a program in the computer-readable storage medium.
[0020] This invention provides an unsupervised low-voltage series arc fault cycle detection method and system based on normalized energy. It constructs a detection model using only normal current samples, divides load categories through hierarchical clustering, and adaptively weights the overall autocorrelation matrix. This achieves high-precision, low-complexity arc cycle detection without the need for fault samples or data annotation, effectively overcoming the problems of small sample size and data imbalance, and helping to improve the power safety monitoring level of low-voltage power distribution systems. The method has low computational complexity, making it suitable for deployment in embedded edge computing devices such as smart circuit breakers and arc fault protection devices. It possesses engineering advantages of low computational overhead and high real-time performance, meeting the online detection needs in complex power distribution scenarios. Attached Figure Description
[0021] Exemplary embodiments of the present invention can be more fully understood by referring to the following figures: Figure 1 A flowchart of an unsupervised low-voltage series arc fault cycle detection method 100 based on normalized energy according to an embodiment of the present invention; Figure 2 This is a general flowchart according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the normal current average sequence and current mean periodic waveform corresponding to a local major category set according to an embodiment of the present invention. Figure 4 This is a schematic diagram of the structure of an unsupervised low-voltage series arc fault cycle detection system 400 based on normalized energy according to an embodiment of the present invention. Detailed Implementation
[0022] Exemplary embodiments of the invention will now be described with reference to the accompanying drawings. However, the invention may be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided to fully and completely disclose the invention and to fully convey its scope to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the drawings is not intended to limit the invention. In the drawings, the same units / elements are referred to by the same reference numerals.
[0023] Unless otherwise stated, the terms used herein (including technical terms) have their common meaning as understood by one of ordinary skill in the art. Furthermore, it is understood that terms defined in commonly used dictionaries should be understood to have a meaning consistent with the context of their relevant field, and not to be interpreted as having an idealized or overly formal meaning.
[0024] Figure 1 This is a flowchart of an unsupervised low-voltage series arc fault cycle detection method 100 based on normalized energy according to an embodiment of the present invention. Figure 1 As shown, the unsupervised low-voltage series arc fault cycle detection method based on normalized energy provided by this invention constructs a detection model using only normal current samples. It divides the load into major categories through hierarchical clustering and adaptively weights the overall autocorrelation matrix. This achieves high-precision, low-complexity arc cycle detection without the need for fault samples and data labeling, effectively overcoming the problems of small samples and data imbalance, and helping to improve the power safety monitoring level of low-voltage power distribution systems. The method has low computational complexity and is suitable for deployment in embedded edge computing devices such as smart circuit breakers and arc fault protection devices. It possesses engineering advantages of low computational overhead and high real-time performance, meeting the online detection needs in complex power distribution scenarios. The unsupervised low-voltage series arc fault cycle detection method 100 based on normalized energy provided by this invention begins at step 101. In step 101, multi-cycle line current sequences under normal operating conditions under different loads are collected, and each current sequence is normalized.
[0025] In this invention, multi-cycle line current sequences under normal operating conditions and different loads are collected. Following the maximum-minimum normalization principle, the sampled values in each current sequence are normalized to obtain the corresponding normalized current sequence. Let the collected multi-cycle line current sequences under normal operating conditions be denoted as... ,in, For the first A current sequence, each sequence containing One cycle, denoted as: , in, Indicates the first In the current sequence, the first Current sampling values for each cycle, This represents the number of sampling points per cycle. The sampled values in the current sequence are normalized according to the minimum-maximum normalization principle. , In the formula, Indicates the first In the current sequence, the first In the period, the 1st Normalized result of each sample value; Represents current sequence The minimum value of the current sampled value; Represents current sequence The maximum value of the current sample value.
[0026] In step 102, the normalized current sequence is clustered to determine multiple local cluster sets, and the average current waveform sequence of all current sequences in each local cluster set is calculated.
[0027] Preferably, the calculation of the average current waveform sequence of all current sequences within each local major category includes: For any local large class set S, if all current sequences in that local large class set S are Let the sampled values of all current sequences in any local large class set S be denoted as ,in, , , , K represents the number of current sequences; K represents the period. This represents the number of sampling points per period, then in any local large class set S The average current waveform sequence of the current sequences is: Each element in the sequence The calculation formula is: , in, This represents the average current at the j-th sampling point in the i-th period of the average current waveform sequence. For the local large class set S, the first... The current value at the j-th sampling point of the i-th period of a current sequence.
[0028] In this invention, hierarchical clustering is used to classify the normalized current sequence into multiple local large category sets according to the consistency of waveform morphology. And calculate the average current waveform sequence of the current sequence within each local major category set.
[0029] With sets For example, suppose All current sequences are Then the set The sampled values of all current sequences in the sequence can be denoted as: ,in , , Obviously, The average current waveform sequence of the current sequences is: Each element in the sequence The calculation formula is: , Similarly, the set can be calculated. The average current waveform sequence.
[0030] In step 103, the mean current periodic waveform of all periods in the mean current waveform sequence corresponding to each local major category is calculated, and the residual between the current waveform of each period in the mean current waveform sequence and the corresponding mean current periodic waveform is calculated to obtain the residual sequence corresponding to each local major category.
[0031] Preferably, the residual between the current waveform of each period in the average current waveform sequence and the corresponding mean current period waveform is calculated to obtain the residual sequence corresponding to each local large category set, including: , , in, The first in the average current waveform sequence In the period, the 1st The residual values of each sampling point; This represents the average current value at the j-th sampling point in the i-th period of the average current waveform sequence. is the current value at the j-th sampling point in the current mean periodic waveform; K is the number of periods contained in the residual sequence.
[0032] In this invention, for all periods of the average current waveform sequence in each local major category set, the corresponding mean period waveform is calculated; and the residual between each period of the average current waveform sequence and its mean period waveform is calculated respectively, resulting in a residual sequence with the same number of sampling points as the average current waveform sequence. The normal current average sequence and the current mean period waveform corresponding to the local major category set are as follows: Figure 2 As shown.
[0033] average current waveform sequence For example, its average current periodic waveform Elements The calculation formula is: , Using the average current periodic waveform Calculate the average current waveform sequence The residual waveform corresponding to each period Each element The calculation formula is: , In the formula, Indicating the first digit in the average current waveform sequence In the period, the 1st The residual values at each sampling point. The residual sequence corresponding to the average current waveform sequence contains... Each cycle, the magnitude of the residual value reflects the volatility and randomness of the current signal waveform.
[0034] In step 104, for any local large category set, the weight coefficient corresponding to each period is calculated based on the residual sequence corresponding to that local large category set.
[0035] Preferably, for any local large category set, the weight coefficient corresponding to each period is calculated based on the residual sequence corresponding to that local large category set, including: For any local large class set, calculate the total residual of N residual values in each period of the residual sequence corresponding to that local large class set. ; The total residual value for all periods is normalized, and based on the normalized total residual value for each period, the weighting coefficient corresponding to each period is calculated. ; in, The first in the average current waveform sequence In the period, the 1st The residual value of each sampling point; N is the number of sampling points per period; and These are the total residual values for the i-th period before and after normalization, respectively; The residual sequence is the first The weighting coefficients for each period; K is the number of periods contained in the parameter sequence.
[0036] In this invention, for a residual sequence containing multiple periods, the sum of each residual value in each period is calculated, and the sum of the residual values corresponding to each period in the residual sequence is subjected to max-min normalization processing; based on the normalized sum of the residual values of each period, a weighting coefficient reflecting the periodic volatility is calculated.
[0037] With sets For example, for each containing For a periodic residual sequence, calculate the residual within each period. The sum of all residual values: , Then the entire residual sequence corresponds to The sum of the residuals is denoted as .right The sum of the residuals is subjected to min-max normalization: , Based on the sum of the normalized residuals for each period, calculate the weighting coefficients that reflect the periodic volatility: , in, Represents a set The corresponding residual sequence of the first The weighting of each period is determined by the number of cycles. Periods with higher volatility are given greater weight to enhance the assessment of random noise and disturbances.
[0038] In step 105, for any local major category set, the overall autocorrelation matrix corresponding to the local major category set is calculated based on the weight coefficients and average current waveform sequence corresponding to the local major category set.
[0039] Preferably, for any local large category set, based on the weighting coefficients and average current waveform sequence corresponding to that local large category set, the overall autocorrelation matrix corresponding to that local large category set is calculated, including: , , in, The first in the average current waveform sequence Periodic vectors The periodic autocorrelation matrix; The residual sequence is the first The weighting coefficients for each period; K is the number of periods contained in the parameter sequence; This is the overall autocorrelation matrix of the average current waveform sequence.
[0040] In this invention, for each period of the average current waveform sequence of each local category set, the periodic autocorrelation matrix is calculated respectively; the weight coefficient reflecting the periodic fluctuation is used as the weight value, and the autocorrelation matrix of each period is weighted and averaged to obtain the overall autocorrelation matrix corresponding to the average current waveform sequence.
[0041] With sets Average current waveform sequence The Middle Periodic vectors For example, its periodic autocorrelation matrix is calculated as follows: , Each average current waveform sequence The corresponding periodic autocorrelation matrix can be obtained for each period. By using weighting coefficients reflecting periodic fluctuations as weighting values, a weighted average is performed on the autocorrelation matrices of each period to obtain the overall autocorrelation matrix corresponding to the average current waveform sequence. , Similarly, the overall autocorrelation matrix of the average current waveform sequence in each of the other local major categories can be calculated, thereby constructing the abnormal period detection and discrimination model corresponding to each local major category.
[0042] In step 106, the current signal to be measured is acquired, the similarity between the current signal to be measured and the average current waveform sequence corresponding to each local category set is calculated, the target local category set corresponding to the current signal to be measured is determined, and the normalized energy value of the current signal to be measured is calculated based on the overall autocorrelation matrix corresponding to the target local category set.
[0043] Preferably, the calculation of the normalized energy value of the measured current signal based on the overall autocorrelation matrix corresponding to the target local category set includes: The overall autocorrelation matrix corresponding to the target local category set is taken as the overall autocorrelation matrix R corresponding to the current signal to be measured. The normalized energy value of the measured current signal is calculated based on the overall autocorrelation matrix corresponding to the measured current signal. y represents the periodic current signal containing N sampling points in the current signal to be measured.
[0044] In step 107, when the normalized energy value is greater than the preset energy threshold, an arc fault cycle is determined to have occurred.
[0045] In this invention, after performing max-min normalization on the test current signal, the similarity measure between the test current signal and the average current waveform sequence corresponding to each local cluster set is calculated. Based on the hierarchical clustering principle, the signal is assigned to the corresponding local cluster set, thereby determining the overall autocorrelation matrix corresponding to the test current signal. The overall autocorrelation matrix is used to calculate each cycle in the test current signal. Normalized energy value: Then, the calculated normalized energy value is... With a preset threshold A comparison is made. When the normalized energy value is greater than the threshold, i.e. If the test signal period is abnormal, it can be determined as an arc period; otherwise, if This indicates that the test signal period is a normal current period containing noise. By judging each period in the test current sequence one by one, the cycle detection of low-voltage series arc faults can be achieved.
[0046] like Figure 3As shown, the unsupervised low-voltage series arc fault cycle detection method based on normalized energy provided by this invention is an anomaly detection model architecture constructed solely from samples of normal line current. First, from the perspective of current waveform consistency, a hierarchical clustering algorithm is used to divide the normal line current under different loads into a finite number of local large-scale clusters, maximizing the distance between clusters. Second, considering the inconsistencies in the number and waveform differences of current waveform sequences within each local large-scale cluster, an average current waveform sequence is defined and used as the representative of the cluster, minimizing the sample spacing within the cluster. Then, for the average current waveform sequences of each local large-scale cluster, an adaptive weighted construction strategy considering the randomness and volatility of the overall autocorrelation matrix is defined, thereby establishing a discriminant model for anomaly cycle detection. In the detection stage, a hierarchical clustering algorithm is used to assign the test current waveform to the corresponding local large-scale cluster, thereby determining the overall autocorrelation matrix corresponding to the test current waveform, and using this matrix to calculate the normalized energy value of each cycle of the test current. Finally, the normalized energy value is compared with a threshold to achieve arc cycle discrimination.
[0047] The unsupervised low-voltage series arc fault cycle detection method based on normalized energy provided by this invention has the following advantages compared with the prior art: (1) In the model building stage, this invention only uses current samples under normal operating conditions. It divides local large class sets through hierarchical clustering and calculates the average current waveform sequence to construct the overall autocorrelation matrix of each set as the discrimination model, without requiring fault samples and data annotation. This method effectively avoids the problems of difficult collection, insufficient quantity and lack of diversity of arc fault samples in actual operating conditions. At the same time, it overcomes the defects such as decreased model generalization ability and limited detection accuracy caused by the imbalance between normal samples and fault samples.
[0048] (2) This invention uses a hierarchical clustering algorithm to classify the normal line current under different loads into a finite number of local large categories based on the consistency of waveform morphology. This solves the technical bottleneck of open and unbounded line current waveforms under different loads, which makes it difficult to construct a unified feature description and detection criterion. The waveform morphology within each category is highly consistent, and the differences between categories are maximized, enabling the detection model to adaptively match the corresponding discrimination parameters for different load types.
[0049] (3) When constructing the overall autocorrelation matrix, this invention introduces an adaptive calculation method for weight coefficients based on the volatility of residual sequences, which gives greater weight to periods with large waveform volatility, enhances attention to random noise and interference, and enables the detection model to more sensitively capture subtle deviations between abnormal periods and normal patterns, thereby improving the sensitivity and robustness of arc period detection.
[0050] (4) The method of the present invention has low computational complexity. The construction of the autocorrelation matrix only involves basic matrix operations. It does not require complex time-frequency domain feature extraction and deep learning model training. It is suitable for deployment in embedded edge computing devices such as smart circuit breakers and arc fault protection devices. It has engineering advantages of low computing power consumption and high real-time performance, and meets the online detection needs in complex power distribution scenarios.
[0051] Figure 4 This is a schematic diagram of the structure of an unsupervised low-voltage series arc fault cycle detection system 400 based on normalized energy according to an embodiment of the present invention. Figure 4 As shown, the unsupervised low-voltage series arc fault cycle detection system 400 based on normalized energy provided in this embodiment of the invention includes: a current sequence acquisition unit 401, a cluster analysis unit 402, a residual sequence calculation unit 403, a weight coefficient calculation unit 404, an overall autocorrelation matrix calculation unit 405, a normalized energy value calculation unit 406, and a judgment unit 407.
[0052] Preferably, the current sequence acquisition unit 401 is used to acquire multi-cycle line current sequences under normal operating conditions under different loads, and to normalize each current sequence.
[0053] Preferably, the clustering analysis unit 402 is used to cluster the normalized current sequence, determine multiple local cluster sets, and calculate the average current waveform sequence of all current sequences within each local cluster set.
[0054] Preferably, the clustering analysis unit 402 calculates the average current waveform sequence of all current sequences within each local cluster set, including: For any local large class set S, if all current sequences in that local large class set S are Let the sampled values of all current sequences in any local large class set S be denoted as ,in, , , , K represents the number of current sequences; K represents the period. This represents the number of sampling points per period, then in any local large class set S The average current waveform sequence of the current sequences is: Each element in the sequence The calculation formula is: , in, This represents the average current at the j-th sampling point in the i-th period of the average current waveform sequence. For the local large class set S, the first... The current value at the j-th sampling point of the i-th period of a current sequence.
[0055] Preferably, the residual sequence calculation unit 403 is used to calculate the mean current periodic waveform of all periods in the average current waveform sequence corresponding to each local major category set, and to calculate the residual between the current waveform of each period in the average current waveform sequence and the corresponding mean current periodic waveform, so as to obtain the residual sequence corresponding to each local major category set.
[0056] Preferably, the residual sequence calculation unit 403 calculates the residual between the current waveform of each period in the average current waveform sequence and the corresponding current mean period waveform to obtain the residual sequence corresponding to each local class set, including: , , in, The first in the average current waveform sequence In the period, the 1st The residual values of each sampling point; This represents the average current value at the j-th sampling point in the i-th period of the average current waveform sequence. is the current value at the j-th sampling point in the current mean periodic waveform; K is the number of periods contained in the residual sequence.
[0057] Preferably, the weight coefficient calculation unit 404 is used to calculate the weight coefficient corresponding to each period based on the residual sequence corresponding to any local large category set.
[0058] Preferably, the weight coefficient calculation unit 404, for any local large category set, calculates the weight coefficient corresponding to each period based on the residual sequence corresponding to that local large category set, including: For any local large class set, calculate the total residual of N residual values in each period of the residual sequence corresponding to that local large class set. ; The total residual value for all periods is normalized, and based on the normalized total residual value for each period, the weighting coefficient corresponding to each period is calculated. ; in, The first in the average current waveform sequence In the period, the 1st The residual value of each sampling point; N is the number of sampling points per period; and These are the total residual values for the i-th period before and after normalization, respectively; The residual sequence is the first The weighting coefficients for each period; K is the number of periods contained in the parameter sequence.
[0059] Preferably, the overall autocorrelation matrix calculation unit 405 is used to calculate the overall autocorrelation matrix corresponding to any local major category set based on the weight coefficients and average current waveform sequence corresponding to the local major category set.
[0060] Preferably, the overall autocorrelation matrix calculation unit 405, for any local major category set, calculates the overall autocorrelation matrix corresponding to that local major category set based on the weighting coefficients and average current waveform sequence corresponding to that local major category set, including: , , in, The first in the average current waveform sequence Periodic vectors The periodic autocorrelation matrix; The residual sequence is the first The weighting coefficients for each period; K is the number of periods contained in the parameter sequence; This is the overall autocorrelation matrix of the average current waveform sequence.
[0061] Preferably, the normalized energy value calculation unit 406 is used to acquire the current signal to be measured, calculate the similarity between the current signal to be measured and the average current waveform sequence corresponding to each local category set, determine the target local category set corresponding to the current signal to be measured, and calculate the normalized energy value of the current signal to be measured based on the overall autocorrelation matrix corresponding to the target local category set.
[0062] Preferably, the normalized energy value calculation unit 406 calculates the normalized energy value of the current signal under test based on the overall autocorrelation matrix corresponding to the target local category set, including: The overall autocorrelation matrix corresponding to the target local category set is taken as the overall autocorrelation matrix R corresponding to the current signal to be measured. The normalized energy value of the measured current signal is calculated based on the overall autocorrelation matrix corresponding to the measured current signal. y represents the periodic current signal containing N sampling points in the current signal to be measured.
[0063] Preferably, the judgment unit 407 is used to determine the occurrence of an arc fault cycle when the normalized energy value is greater than a preset energy threshold.
[0064] The unsupervised low-voltage series arc fault cycle detection system 400 based on normalized energy in one embodiment of the present invention corresponds to the unsupervised low-voltage series arc fault cycle detection method 100 based on normalized energy in another embodiment of the present invention, and will not be described again here.
[0065] According to another aspect of the present invention, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the steps of an unsupervised low-voltage series arc fault cycle detection method based on normalized energy.
[0066] According to another aspect of the present invention, the present invention provides an electronic device, comprising: The aforementioned computer-readable storage medium; and One or more processors for executing a program in the computer-readable storage medium.
[0067] The present invention has been described with reference to a few embodiments. However, it will be apparent to those skilled in the art that other embodiments besides those disclosed above fall equivalently within the scope of the present invention.
[0068] Generally, all terms used in this invention are interpreted according to their ordinary meaning in the art, unless otherwise expressly defined herein. All references to “a / the / the [device, component, etc.]” are openly interpreted as at least one instance of said device, component, etc., unless otherwise expressly stated. The steps of any method disclosed herein need not be performed in the exact order disclosed, unless explicitly stated otherwise.
[0069] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0070] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams.Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0071] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0072] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the present invention.
Claims
1. An unsupervised low-voltage series arc fault cycle detection method based on normalized energy, characterized in that, The method includes: Collect multi-cycle line current sequences under normal operating conditions with different loads, and normalize each current sequence. The normalized current sequences are clustered to determine multiple local cluster sets, and the average current waveform sequence of all current sequences within each local cluster set is calculated. Calculate the mean current periodic waveform of all periods in the mean current waveform sequence corresponding to each local major category set, and calculate the residual between the current waveform of each period in the mean current waveform sequence and the corresponding mean current periodic waveform to obtain the residual sequence corresponding to each local major category set. For any local large category set, calculate the weight coefficient corresponding to each period based on the residual sequence corresponding to that local large category set; For any local major category set, based on the weight coefficients and average current waveform sequence corresponding to that local major category set, calculate the overall autocorrelation matrix corresponding to that local major category set; Acquire the current signal to be measured, calculate the similarity between the current signal to be measured and the average current waveform sequence corresponding to each local category set, determine the target local category set corresponding to the current signal to be measured, and calculate the normalized energy value of the current signal to be measured based on the overall autocorrelation matrix corresponding to the target local category set. When the normalized energy value is greater than the preset energy threshold, an arc fault cycle is determined to have occurred.
2. The method according to claim 1, characterized in that, Calculate the average current waveform sequence of all current sequences within each local major category set, including: For any local large class set S, if all current sequences in that local large class set S are Let the sampled values of all current sequences in any local large class set S be denoted as ,in, , , , K represents the number of current sequences; K represents the period. This represents the number of sampling points per period, then in any local large class set S The average current waveform sequence of the current sequences is: Each element in the sequence The calculation formula is: ， in, The average current is the average current at the j-th sampling point in the i-th period of the average current waveform sequence; For the local large class set S, the first... The current value at the j-th sampling point of the i-th period of a current sequence.
3. The method according to claim 1, characterized in that, Calculate the residuals between the current waveform of each period and the corresponding mean current period waveform in the average current waveform sequence to obtain the residual sequence corresponding to each local large category set, including: ， ， in, The first in the average current waveform sequence In the period, the 1st The residual values of each sampling point; This represents the average current value at the j-th sampling point in the i-th period of the average current waveform sequence. is the current value at the j-th sampling point in the current mean periodic waveform; K is the number of periods contained in the residual sequence.
4. The method according to claim 1, characterized in that, For any local large category set, calculate the weight coefficients for each period based on the residual sequence corresponding to that local large category set, including: For any local large class set, calculate the total residual of N residual values in each period of the residual sequence corresponding to that local large class set. ; The total residual value for all periods is normalized, and based on the normalized total residual value for each period, the weighting coefficient corresponding to each period is calculated. ; in, The first in the average current waveform sequence In the period, the 1st The residual value of each sampling point; N is the number of sampling points per period; and These are the total residual values for the i-th period before and after normalization, respectively; The residual sequence is the first The weighting coefficients for each period; K is the number of periods contained in the parameter sequence.
5. The method according to claim 1, characterized in that, For any local large category set, based on the weighting coefficients and average current waveform sequence corresponding to that local large category set, calculate the overall autocorrelation matrix corresponding to that local large category set, including: ， ， in, The first in the average current waveform sequence Periodic vectors The periodic autocorrelation matrix; The residual sequence is the first The weighting coefficients for each period; K is the number of periods contained in the parameter sequence; This is the overall autocorrelation matrix of the average current waveform sequence.
6. The method according to claim 1, characterized in that, The normalized energy value of the measured current signal is calculated based on the overall autocorrelation matrix corresponding to the target local category set, including: The overall autocorrelation matrix corresponding to the target local category set is taken as the overall autocorrelation matrix R corresponding to the current signal to be measured. The normalized energy value of the measured current signal is calculated based on the overall autocorrelation matrix corresponding to the measured current signal. ; where y is the periodic current signal containing N sampling points in the current signal to be measured.
7. An unsupervised low-voltage series arc fault cycle detection system based on normalized energy, characterized in that, The system includes: The current sequence acquisition unit is used to acquire multi-cycle line current sequences under normal operating conditions under different loads and to normalize each current sequence. The clustering analysis unit is used to cluster the normalized current sequences, determine multiple local cluster sets, and calculate the average current waveform sequence of all current sequences within each local cluster set. The residual sequence calculation unit is used to calculate the mean current periodic waveform of all periods in the average current waveform sequence corresponding to each local major category set, and to calculate the residual between the current waveform of each period in the average current waveform sequence and the corresponding mean current periodic waveform, so as to obtain the residual sequence corresponding to each local major category set. The weight coefficient calculation unit is used to calculate the weight coefficient corresponding to each period based on the residual sequence corresponding to any local large category set. The overall autocorrelation matrix calculation unit is used to calculate the overall autocorrelation matrix corresponding to any local major category set based on the weight coefficients and average current waveform sequence corresponding to that local major category set. The normalized energy value calculation unit is used to acquire the current signal to be measured, calculate the similarity between the current signal to be measured and the average current waveform sequence corresponding to each local category set, determine the target local category set corresponding to the current signal to be measured, and calculate the normalized energy value of the current signal to be measured based on the overall autocorrelation matrix corresponding to the target local category set. The judgment unit is used to determine the occurrence of an arc fault cycle when the normalized energy value is greater than a preset energy threshold.
8. The system according to claim 7, characterized in that, The clustering analysis unit calculates the average current waveform sequence of all current sequences within each local cluster set, including: For any local large class set S, if all current sequences in that local large class set S are Let the sampled values of all current sequences in any local large class set S be denoted as ,in, , , , K represents the number of current sequences; K represents the period. This represents the number of sampling points per period, then in any local large class set S The average current waveform sequence of the current sequences is: Each element in the sequence The calculation formula is: ， in, The average current is the average current at the j-th sampling point in the i-th period of the average current waveform sequence; For the local large class set S, the first... The current value at the j-th sampling point of the i-th period of a current sequence.
9. The system according to claim 7, characterized in that, The residual sequence calculation unit calculates the residual between the current waveform of each period and the corresponding mean current period waveform in the average current waveform sequence to obtain the residual sequence corresponding to each local major category set, including: ， ， in, The first in the average current waveform sequence In the period, the 1st The residual values of each sampling point; This represents the average current value at the j-th sampling point in the i-th period of the average current waveform sequence. is the current value at the j-th sampling point in the current mean periodic waveform; K is the number of periods contained in the residual sequence.
10. The system according to claim 7, characterized in that, The weight coefficient calculation unit, for any local large category set, calculates the weight coefficient corresponding to each period based on the residual sequence corresponding to that local large category set, including: For any local large class set, calculate the total residual of N residual values in each period of the residual sequence corresponding to that local large class set. ; The total residual value for all periods is normalized, and based on the normalized total residual value for each period, the weighting coefficient corresponding to each period is calculated. ; in, The first in the average current waveform sequence In the period, the 1st The residual value of each sampling point; N is the number of sampling points per period; and These are the total residual values for the i-th period before and after normalization, respectively; The residual sequence is the first The weighting coefficients for each period; K is the number of periods contained in the parameter sequence.
11. The system according to claim 7, characterized in that, The overall autocorrelation matrix calculation unit, for any local major category set, calculates the overall autocorrelation matrix corresponding to that local major category set based on the weighting coefficients and average current waveform sequence corresponding to that local major category set, including: ， ， in, The first in the average current waveform sequence Periodic vectors The periodic autocorrelation matrix; The residual sequence is the first The weighting coefficients for each period; K is the number of periods contained in the parameter sequence; This is the overall autocorrelation matrix of the average current waveform sequence.
12. The system according to claim 7, characterized in that, The normalized energy value calculation unit calculates the normalized energy value of the current signal under test based on the overall autocorrelation matrix corresponding to the target local category set, including: The overall autocorrelation matrix corresponding to the target local category set is taken as the overall autocorrelation matrix R corresponding to the current signal to be measured. The normalized energy value of the measured current signal is calculated based on the overall autocorrelation matrix corresponding to the measured current signal. ; where y is the periodic current signal containing N sampling points in the current signal to be measured.
13. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method as described in any one of claims 1-6.
14. An electronic device, characterized in that, include: The computer-readable storage medium as described in claim 13; as well as One or more processors for executing a program in the computer-readable storage medium.