Method for determining series arc fault of low voltage distribution cable based on disturbance current characteristics

Abstract

This invention belongs to the field of low-voltage power distribution cable fault identification technology, specifically involving a method for judging series arc faults in low-voltage power distribution cables based on disturbance current characteristics. The method includes: S1, acquiring current data of the three-phase current of the cable; S2, analyzing the similarity of current waveforms in adjacent current cycles using the Pearson correlation coefficient (PCC) and determining whether current disturbance has occurred in the cable; S3, when current disturbance is determined to have occurred, extracting current data from the disturbance initiation point and subsequent current cycles, and recording this as the disturbance current; calculating time-frequency domain characteristics of the extracted disturbance current; S4, inputting the time-frequency domain characteristics of the disturbance current into a preset fault identification model to detect and identify series arc faults in the cable, and determining whether the fault causing the current disturbance is a series arc fault. This method can efficiently, accurately, and promptly monitor and identify series arc faults in low-voltage power distribution cables, improving the safety and stability of the power system.

Description

Technical Field

[0001] This invention belongs to the field of low-voltage power distribution cable fault identification technology, and particularly relates to a method for judging series arc faults in low-voltage power distribution cables based on disturbance current characteristics. Background Technology

[0002] In power systems, the safe operation of low-voltage distribution cables is crucial. However, cables are often affected by various factors during actual operation, leading to faults, among which series arcing faults are relatively common. Long-term operation of low-voltage cables often reveals obvious signs of aging; decreased cable insulation and poor contact can easily lead to series arcing faults. A series arcing fault releases a large amount of heat, potentially igniting nearby flammable materials. The resulting electrical fire poses a serious threat to the power system, equipment, and personnel safety. If this fault is not detected and addressed promptly, it can severely threaten the stable operation of the power system, even leading to equipment damage or fires. Therefore, once a series arcing fault occurs in a low-voltage cable, personnel must promptly detect and eliminate the fault to prevent greater losses.

[0003] Currently, online monitoring of electric arcs in power distribution network cables mainly employs two methods: arc detection and ultrasonic monitoring. Arc detection determines the presence of partial arc discharge by detecting abnormal arcing from the cable, while ultrasonic monitoring detects the ultrasonic signals generated by the arc discharge. However, both methods have some significant problems in practical applications.

[0004] First, arc flash and ultrasonic monitoring methods have relatively limited coverage, making it difficult to comprehensively monitor the entire distribution network cables. This may lead to some potential arc faults being missed, increasing the operational risks of the power system. Second, the accuracy of these two methods is low. Because arc flash and ultrasonic signals are easily affected by external environmental interference, such as electromagnetic noise and mechanical vibration, false alarms or missed alarms often occur in practical applications. This not only reduces the reliability of the monitoring system but also increases the workload of maintenance personnel. Furthermore, the timeliness of arc flash and ultrasonic monitoring methods is low; they can only detect arc faults in cables when they are already very obvious. This means that by the time the monitoring system issues an alarm, the fault has often already developed to a relatively serious stage, easily leading to further expansion and exacerbation of the damage.

[0005] Therefore, existing online monitoring methods for arc faults in power distribution cables suffer from problems such as small coverage, low accuracy, and low timeliness. How to efficiently, accurately, and timely monitor and identify series arc faults in low-voltage power distribution cables to improve the safety and stability of the power system has become an urgent problem to be solved. Summary of the Invention

[0006] To address the shortcomings of the existing technology, this invention provides a method for judging series arc faults in low-voltage distribution cables based on disturbance current characteristics. This method can efficiently, accurately, and promptly monitor and identify series arc faults in low-voltage distribution cables, thereby improving the safety and stability of the power system.

[0007] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0008] The method for identifying series arc faults in low-voltage distribution cables based on disturbance current characteristics includes the following steps:

[0009] S1. Real-time acquisition of three-phase current data of low-voltage distribution network cables;

[0010] S2. For the acquired current data, the Pearson correlation coefficient (PCC) is used to analyze the similarity of the current waveforms in each adjacent current cycle, and to determine whether the cable is experiencing current disturbance.

[0011] S3. When it is determined that a current disturbance has occurred, take the current period in which the current disturbance occurs as the disturbance start point, extract the current data of the disturbance start point and multiple current periods thereafter, and record them as the disturbance current; calculate the time-frequency domain characteristic quantities of the extracted disturbance current.

[0012] S4. Input the time-frequency domain characteristics of the disturbance current into the preset fault identification model to detect and identify the cable series arc fault, and determine whether the fault causing the current disturbance is a series arc fault.

[0013] Compared with the prior art, the present invention has the following beneficial effects:

[0014] 1. Improved accuracy of monitoring and identification. By analyzing the similarity of current waveforms in adjacent current cycles using the Pearson correlation coefficient (PCC), this method can preliminarily determine whether current disturbances have occurred in the cable. This method effectively captures minute changes in the current waveform, thus accurately screening out currents suspected of causing cable series arcing faults. Compared to traditional methods, this approach improves the accuracy of monitoring and identifying cable fault arcs in the initial stage, laying a solid foundation for subsequent analysis and judgment.

[0015] 2. Ensuring Timely Identification of Series Arc Faults. This method analyzes current data from the starting point of the disturbance current and multiple subsequent cycles, enabling rapid detection of the initial characteristics of arc faults. Since arc faults are often weak in their initial stages but gradually intensify over time, timely identification and intervention are crucial for preventing fault escalation and protecting equipment and personnel safety. This method allows for accurate identification in the early stages of arc fault development, ensuring timely identification of series arc faults and providing a valuable time window for subsequent fault handling.

[0016] 3. Reduce the amount of current data to be identified and lower monitoring latency. After confirming a current disturbance, this scheme only extracts data from multiple current cycles after the disturbance's inception point for further analysis. This avoids processing large amounts of irrelevant data, significantly reducing the amount of current data that needs to be calculated. Simultaneously, because only the data after the disturbance is processed, the entire monitoring and identification process is significantly shortened, thereby reducing monitoring latency and improving the system's real-time response capability.

[0017] 4. Ensuring Accuracy of Identification. When a series arc fault occurs in a low-voltage distribution network cable, the current data will exhibit obvious time-frequency domain characteristic changes, such as an increase in high-frequency components and a decrease in periodicity. This method calculates the time-frequency domain characteristics of the disturbance current and inputs them into a preset fault identification model, fully utilizing these characteristic changes to accurately determine the fault type. This method considers not only the time-domain characteristics of the current but also its frequency-domain characteristics, thereby ensuring the accuracy and reliability of cable series arc fault identification.

[0018] This method, by comprehensively applying Pearson correlation coefficient analysis, time-frequency domain feature calculation, and fault identification model, achieves efficient, accurate, and timely monitoring and identification of series arc faults in low-voltage distribution cables, effectively improving the safety and stability of the power system.

[0019] Preferably, in S2, when determining whether a current disturbance has occurred in the cable, the current disturbance is determined by combining the cumulative sum algorithm CUSUM based on double sliding windows W1 and W2; wherein, W1 is a front window used to obtain the expected distribution value E and determine the upper bound threshold UB and the lower bound threshold LB; the expected value E is the average value calculated based on the current waveform data distribution within the front window W1; the values ​​of the upper bound threshold UB and the lower bound threshold LB are set based on historical data; W2 is a back window used to calculate the offset accumulation value C of the data sequence within the window; the window length of both windows is M current cycles;

[0020] Set the initial accumulated value of positive offset and negative offset initial accumulated value All values ​​are 0; after each sliding displacement of the window, based on the expected value E and the thresholds UB and LB, the offset of each point data in W2 is calculated sequentially, and the positive offset is accumulated. and negative offset initial accumulated value The iteration process is as follows:

[0021]

[0022] In the formula, x k Let x be the value of the Pearson correlation coefficient for the k-th current cycle within W2; when x k When the positive offset accumulation value is greater than the upper threshold E+UB, Increase; when x k When the negative offset accumulation value is less than the lower threshold E-LB Decrease;

[0023] If satisfied or This indicates that the current in the window has been disturbed; where DL is the preset offset limit.

[0024] This setup has two advantages: 1. It improves the sensitivity of disturbance detection. By setting a front window, this technique can acquire the expected distribution value of current waveform data and determine upper and lower bound thresholds based on this. These thresholds are set based on historical data and can reflect the fluctuation range of normal current waveforms. When the data in the back window deviates from these thresholds, it indicates that a disturbance may have occurred in the current. Compared to a single threshold judgment, this method is better able to capture minute current changes, thereby improving the sensitivity of disturbance detection.

[0025] 2. It can enhance the accuracy of disturbance identification. The back window is used to calculate the accumulated offset value of the data sequence within the window. By iterating the positive and negative accumulated offset values, the offset of the current waveform data relative to the expected value can be accumulated. When the accumulated offset value reaches a preset offset limit, it can be determined that a disturbance change has occurred in the current. This method, by accumulating the offset, can more accurately reflect the overall trend of current waveform changes, thus enhancing the accuracy of disturbance identification.

[0026] 3. High adaptability, suitable for various scenarios. By adjusting parameters such as the window length, threshold, and offset limit of the pre-window and post-window, it can adapt to the current disturbance detection needs of different cable types and operating conditions. This makes the technology highly adaptable and widely applicable to various power system monitoring and fault diagnosis scenarios.

[0027] Preferably, in S2, when using the Pearson correlation coefficient (PCC) to analyze the similarity of current waveforms in adjacent current cycles, the formula for calculating the Pearson correlation coefficient for each current cycle is:

[0028]

[0029] In the formula, PCC m The Pearson correlation coefficient for the m-th current cycle is represented. This represents the instantaneous current value at the kth moment in the mth current cycle. This represents the average current in the m-th cycle; K is the number of instantaneous current values ​​in one current cycle.

[0030] In this setup, the Pearson correlation coefficient, a statistical indicator measuring the degree of linear correlation between two variables, ranges from -1 to 1. When applied to analyze the similarity of current waveforms between adjacent current cycles, it can accurately capture subtle changes in the current waveform between cycles. These changes may reflect potential faults or anomalies in the cable, such as series arcing faults. When the current waveform changes significantly, the Pearson correlation coefficient will also change significantly. This change provides a sensitive indicator for fault detection, enabling the system to detect potential faults earlier, thereby improving the sensitivity of fault detection. In subsequent analysis, the Pearson correlation coefficient can serve as an important basis for determining whether current disturbances have occurred in the cable. When the Pearson correlation coefficient between adjacent current cycles falls below a certain preset threshold, it can be considered that the current waveform has changed significantly, thus indicating a possible fault in the cable. This provides strong data support for fault identification.

[0031] Preferably, in S3, when calculating the time-frequency domain characteristics of the extracted disturbance current, the calculated time-domain characteristics include zero rest time, peak-to-peak value, standard deviation, and kurtosis; the calculated frequency-domain characteristics include harmonic amplitude, spectral centroid, and spectral standard deviation.

[0032] This setup, by calculating the time-domain characteristics (zero rest time, peak-to-peak value, standard deviation, kurtosis) and frequency-domain characteristics (harmonic amplitude, spectral centroid, spectral standard deviation) of the disturbance current, can comprehensively reflect the characteristics of series arc faults in cables, providing strong data support for the detection and diagnosis of cable series arc faults. These characteristics are not only intuitive but also easy to calculate and extract, and have broad application prospects in power system monitoring and fault diagnosis.

[0033] Preferably, when calculating time-domain characteristic quantities, the formula for calculating peak-to-peak value is:

[0034]

[0035] In the formula PPV k This represents the normalized peak-to-peak value of the k-th current cycle. and This represents the maximum and minimum values ​​of the current collected in the k-th current cycle.

[0036] In this configuration, when an arc fault occurs, the current waveform often fluctuates significantly due to the nonlinear characteristics and instability of the arc, resulting in a substantial increase in the peak-to-peak value. Therefore, by calculating the peak-to-peak value, this abnormal change in the current waveform can be visually captured, providing a strong basis for arc fault detection.

[0037] Preferably, when calculating time-domain features, the formulas for calculating standard deviation and kurtosis are:

[0038]

[0039] In the formula, Std rms The standard deviation of the effective value of current, Kurt rms The kurtosis of the effective value of the current is represented by N, where N represents the total number of current cycles captured, and k represents the ordinal number of the cycle. I represents the average effective value of the intercepted current. m This represents the m-th current sample value within one cycle, where n represents the total number of current samples within one cycle.

[0040] In this configuration, when an arc fault occurs, the current waveform becomes irregular, with increased dispersion of the current value, leading to an increase in the standard deviation. Therefore, by calculating the standard deviation, the dispersion of the current waveform can be quantified, thereby determining the presence and severity of the arc fault. Furthermore, in current waveform analysis, kurtosis can be used to detect extreme values ​​or abnormal fluctuations in the current waveform. When an arc fault occurs, the current waveform may exhibit large instantaneous current values ​​or abnormal fluctuation patterns, causing changes in kurtosis. Therefore, by calculating kurtosis, these abnormal characteristics in the current waveform can be captured, providing additional information for the detection and diagnosis of arc faults.

[0041] Preferably, when calculating frequency domain characteristic quantities, the calculation process for harmonic amplitude includes:

[0042] First, the current signal of the disturbance current is filtered and preprocessed.

[0043] Then, a short-time Fourier transform (STFT) is performed on the preprocessed current signal;

[0044] Then, the average value of the results obtained from the short-time Fourier transform is calculated to obtain the harmonic amplitude of the disturbance current; among them, the Hanning window is selected as the window function of the short-time Fourier transform.

[0045] This setup allows for filtering preprocessing, removing high-frequency noise and interference from the current signal and improving signal quality. This is crucial for subsequent short-time Fourier transform (STFT) and accurate harmonic amplitude calculation. STFT enables joint analysis of the signal in both the time and frequency domains, providing information on the frequency components of the signal at different time points. This is particularly useful for analyzing non-stationary signals (such as disturbance currents). By selecting appropriate window functions (Hanning windows) and parameter settings, STFT can achieve high time and frequency resolution, thus capturing the detailed features of the signal more accurately. Averaging the STFT results reduces the impact of random errors and measurement noise, improving the accuracy of harmonic amplitude calculation; it also enhances the stability of the results, ensuring consistency across different time periods. This method can accurately calculate the harmonic amplitude of disturbance currents, providing strong support for power system analysis, signal processing, and fault diagnosis.

[0046] Preferably, when calculating frequency domain features, the spectral centroid f c The formula for calculation is:

[0047]

[0048] Spectral Standard Deviation Std f The formula for calculation is:

[0049]

[0050] In the formula, f k A represents the frequency value at position k in the frequency sequence. k Represents frequency f k The corresponding amplitude is given by N, where N is the defined spectral width.

[0051] With this setup, the spectral centroid can be used to detect spectral changes in current or voltage signals. When a fault or anomaly exists in the power system, such as a series arc fault, the spectral distribution of the current or voltage signal may change, causing a shift in the spectral centroid. Therefore, by calculating the spectral centroid, this change in spectral distribution can be captured, providing a strong basis for fault detection and diagnosis.

[0052] Spectral standard deviation can be used to assess the spectral stability of current or voltage signals. When instability or faults exist in a power system, the spectral distribution of current or voltage signals may become more dispersed, leading to an increase in the spectral standard deviation. Therefore, by calculating the spectral standard deviation, the degree of dispersion of this spectral distribution can be quantified, providing important reference information for assessing the stability and reliability of power systems.

[0053] Preferably, in S4, the preset fault identification model is a Stacking model composed of decision tree DT, support vector machine SVM, K nearest neighbor KNN and logistic regression LR, which is built based on the logic of ensemble learning algorithms; and the Stacking model is trained using historical data of the current time-frequency domain features of real cable series arc faults that have occurred in the past.

[0054] This setup has two advantages: 1. It can improve the accuracy of fault identification. By integrating multiple different base learners (such as decision trees, support vector machines, K-nearest neighbors, and logistic regression), the Stacking model can fully utilize the advantages of each base learner, thereby improving the accuracy of fault identification. Each base learner can extract different information from the current time-frequency domain features, while the Stacking model can effectively combine this information to form a more comprehensive fault identification capability.

[0055] 2. Enhanced Model Generalization Ability: Training the Stacking model using historical data of the time-frequency domain characteristics of actual cable series arc faults enhances its generalization ability. This means the model can not only accurately identify fault modes in the training data but also effectively predict and identify unseen fault data. This is crucial for fault detection and diagnosis in practical applications.

[0056] Preferably, in S4, if a series arc fault is determined to have occurred, the disturbance current waveform of the cable fault and the location of the faulty cable joint are sent to the back-end terminal, and the back-end terminal sends a fault warning message to the management terminal and / or maintenance terminal.

[0057] This setup, by automatically sending fault information, allows the system to respond quickly upon detecting a series arcing fault. This helps management and maintenance personnel to understand the fault situation promptly, enabling them to take necessary intervention measures quickly to prevent the fault from worsening or causing a larger safety accident.

[0058] Furthermore, the system can accurately send the location of the faulty cable connector to the backend, which greatly simplifies the troubleshooting process. Maintenance personnel can go directly to the designated location for repairs without conducting extensive searches and investigations, thereby improving maintenance efficiency and reducing maintenance costs. Attached Figure Description

[0059] To make the objectives, technical solutions, and advantages of the invention clearer, the invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0060] Figure 1 This is a flowchart of the method;

[0061] Figure 2This is a graph showing the change in the Pearson correlation coefficient of the cable after a current disturbance occurs in the embodiment.

[0062] Figure 3 This is a schematic diagram of the sliding window displacement logic in the embodiment;

[0063] Figure 4 This is a schematic diagram illustrating the training process of the Stacking model in the embodiment. Detailed Implementation

[0064] The following detailed explanation illustrates the specific implementation methods:

[0065] Example:

[0066] like Figure 1 As shown in the figure, this embodiment discloses a method for judging series arc faults in low-voltage distribution cables based on disturbance current characteristics, including the following steps:

[0067] S1. Real-time acquisition of three-phase current data of low-voltage distribution network cables.

[0068] In practical implementation, the three-phase current data of the cable joint can be collected in real time through the current carrier fluid and stored in the data storage module. The current sensor and the current carrier fluid of the cable joint are integrated to collect the current signal. The Hall sensor in the integrated module of the current carrier fluid and sensor of the cable joint stores the collected real-time three-phase current data in the data storage module and transmits the data through the communication module.

[0069] S2. For the acquired current data, the Pearson correlation coefficient (PCC) is used to analyze the similarity of the current waveforms in each adjacent current cycle, and to determine whether the cable is experiencing current disturbance.

[0070] When using the Pearson correlation coefficient (PCC) to analyze the similarity of current waveforms in adjacent current cycles, the formula for calculating the Pearson correlation coefficient for each current cycle is as follows:

[0071]

[0072] In the formula, PCC m The Pearson correlation coefficient for the m-th current cycle is represented. This represents the instantaneous current value at the k-th moment in the m-th current cycle. This represents the average current in the m-th cycle; K is the number of instantaneous current values ​​in one current cycle.

[0073] The change in Pearson correlation coefficient after current disturbance in the cable is as follows: Figure 2 As shown.

[0074] The Pearson correlation coefficient is a statistical indicator that measures the degree of linear correlation between two variables, with a value ranging from -1 to 1. When applied to analyze the similarity of current waveforms between adjacent current cycles, it can accurately capture subtle changes in the current waveform between cycles. These changes may reflect potential faults or anomalies in the cable, such as series arcing faults. When the current waveform changes significantly, the Pearson correlation coefficient will also change significantly. This change provides a sensitive indicator for fault detection, enabling the system to detect potential faults earlier, thereby improving the sensitivity of fault detection. In subsequent analysis, the Pearson correlation coefficient can serve as an important basis for determining whether current disturbances have occurred in the cable. When the Pearson correlation coefficient between adjacent current cycles falls below a certain preset threshold, it can be considered that the current waveform has changed significantly, thus indicating that a fault may have occurred in the cable. This provides strong data support for fault identification.

[0075] like Figure 3 As shown. In specific implementation, when determining whether a current disturbance has occurred in the cable, assume that the obtained continuous Pearson correlation coefficient (PCC) data sequence X = {x1, x2, ..., x...} is obtained. K The Pearson correlation coefficient index is processed using the CUSUM algorithm based on double sliding windows W1 and W2 to determine current disturbances.

[0076] W1 is the pre-window, used to obtain the expected distribution value E and determine the upper and lower bound thresholds UB and LB. The expected value E is the average value calculated based on the current waveform data distribution within the pre-window W1. The values ​​of the upper and lower bound thresholds UB and LB are set based on historical data, typically μ ± kσ, where σ is the standard deviation and k is a preset coefficient. W2 is the post-window, used to calculate the cumulative offset value C of the data sequence within the window. The window length of both windows is M current cycles.

[0077] Set the initial accumulated value of positive offset and negative offset initial accumulated value All values ​​are 0; after each sliding displacement of the window, based on the expected value E and the thresholds UB and LB, the offset of each point data in W2 is calculated sequentially, and the positive offset is accumulated. and negative offset initial accumulated value The iteration process is as follows:

[0078]

[0079] In the formula, x k Let x be the value of the Pearson correlation coefficient for the k-th current cycle within W2; when x k When the positive offset accumulation value is greater than the upper threshold E+UB, Increase; when xk When the negative offset accumulation value is less than the lower threshold E-LB Decrease;

[0080] If satisfied or This indicates that the current in the window has been disturbed; where DL is the preset offset limit.

[0081] This setup improves the sensitivity of disturbance detection. By setting a pre-window, the technology can acquire the expected distribution value of the current waveform data and determine the upper and lower bound thresholds based on this. These thresholds are set based on historical data and can reflect the fluctuation range of normal current waveforms. When the data in the post-window deviates from these thresholds, it indicates that a disturbance may have occurred in the current. Compared to single threshold judgment, this method is better at capturing minute current changes, thus improving the sensitivity of disturbance detection. Furthermore, it enhances the accuracy of disturbance identification. The post-window is used to calculate the accumulated offset value of the data sequence within the window. By iterating the positive and negative accumulated offset values, the offset of the current waveform data relative to the expected value can be accumulated. When the accumulated offset value reaches a preset offset limit, it can be determined that a disturbance change has occurred in the current. This method, by accumulating the offset, can more accurately reflect the overall trend of current waveform changes, enhancing the accuracy of disturbance identification. It is also highly adaptable and suitable for various scenarios. By adjusting parameters such as the window length, threshold, and offset limit of the pre-window and post-window, it can adapt to the current disturbance detection needs of different cable types and operating conditions. This makes the technology highly adaptable and can be widely applied in various power system monitoring and fault diagnosis scenarios.

[0082] S3. When it is determined that a current disturbance has occurred, take the current period when the current disturbance occurs as the disturbance start point, extract the current data of the disturbance start point and multiple current periods thereafter, and record them as the disturbance current; calculate the time-frequency domain characteristic quantities of the extracted disturbance current.

[0083] When calculating the time-frequency domain characteristics of the extracted disturbance current, the calculated time-domain characteristics include zero-rest time, peak-to-peak value, standard deviation, and kurtosis; the calculated frequency-domain characteristics include harmonic amplitude, spectral centroid, and spectral standard deviation. Thus, by calculating the time-domain characteristics (zero-rest time, peak-to-peak value, standard deviation, and kurtosis) and frequency-domain characteristics (harmonic amplitude, spectral centroid, and spectral standard deviation) of the disturbance current, the characteristics of series arc faults in cables can be comprehensively reflected, providing strong data support for the detection and diagnosis of cable series arc faults. These characteristics are not only intuitive but also easy to calculate and extract, showing broad application prospects in power system monitoring and fault diagnosis.

[0084] In practical implementation, the zero-rest time is calculated using the following formula when calculating time-domain characteristic quantities:

[0085]

[0086] In the formula, T zero Indicates the zero-rest time within a single cycle, i k I represents the amplitude of the k-th current sampled within the period. rms This represents 5% of the effective value of the current in that period, f s Indicates the corresponding sampling frequency; N represents the total number of current samplings within the period; α k (i k <I rms ) indicates whether condition i is satisfied. k <I rms The statistical parameter, when condition i is satisfied k <I rms When α is taken k (i k <I rms ) = 1, when condition i is not satisfied k <I rms When α is taken k (i k <I rms ) = 0.

[0087] The zero-out period is a specific time in the periodic changes of alternating current where the arc extinguishes and reignites, characterized by a current value close to zero. This characteristic can intuitively reflect the periodicity of arc activity and is crucial for determining the existence and activity patterns of arc faults.

[0088] The formula for calculating peak-to-peak value is:

[0089]

[0090] In the formula PPV k This represents the normalized peak-to-peak value of the k-th current cycle. and This represents the maximum and minimum values ​​of the current collected in the k-th current cycle.

[0091] When an arc fault occurs, the current waveform often fluctuates significantly due to the nonlinear characteristics and instability of the arc, resulting in a substantial increase in peak-to-peak value. Therefore, by calculating the peak-to-peak value, this abnormal change in the current waveform can be visually captured, providing a strong basis for arc fault detection.

[0092] The formulas for calculating standard deviation and kurtosis are:

[0093]

[0094] In the formula, Stdrms The standard deviation of the effective value of current, Kurt rms The kurtosis of the effective value of the current is represented by N, where N represents the total number of current cycles captured, and k represents the ordinal number of the cycle. I represents the average effective value of the intercepted current. m This represents the m-th current sample value within one cycle, where n represents the total number of current samples within one cycle.

[0095] When an arc fault occurs, the current waveform becomes irregular, with increased dispersion of the current value, leading to an increase in the standard deviation. Therefore, by calculating the standard deviation, the dispersion of the current waveform can be quantified, thereby determining the presence and severity of the arc fault. In addition, kurtosis can be used in current waveform analysis to detect extreme values ​​or abnormal fluctuations in the current waveform. When an arc fault occurs, large instantaneous current values ​​or abnormal fluctuation patterns may appear in the current waveform, causing changes in kurtosis. Therefore, by calculating kurtosis, these abnormal characteristics in the current waveform can be captured, providing additional information for the detection and diagnosis of arc faults.

[0096] When calculating frequency domain characteristic quantities, the calculation process for harmonic amplitude includes:

[0097] First, the current signal of the disturbance current is filtered and preprocessed.

[0098] Then, a short-time Fourier transform (STFT) is performed on the preprocessed current signal;

[0099] Next, the average value of the results obtained from the short-time Fourier transform is calculated to obtain the harmonic amplitude of the disturbance current; the Hanning window is selected as the window function for the short-time Fourier transform. In specific implementation, before calculating the average value, higher harmonics in the embodiment with the smallest amplitude can be filtered out, retaining harmonics from the 2nd to the 21st harmonics.

[0100] Thus, filtering preprocessing can remove high-frequency noise and interference from the current signal, improving signal quality. This is crucial for subsequent short-time Fourier transform (STFT) and accurate calculation of harmonic amplitudes. STFT enables joint analysis of the signal in the time and frequency domains, providing information on the frequency components of the signal at different time intervals. This is particularly useful for analyzing non-stationary signals (such as disturbance currents). By selecting appropriate window functions (Hanning windows) and parameter settings, STFT can achieve high time and frequency resolution, thereby capturing the detailed features of the signal more accurately. Averaging the STFT results can reduce the impact of random errors and measurement noise, improving the accuracy of harmonic amplitude calculation; it also enhances the stability of the results, ensuring consistency of the harmonic amplitude calculation results across different time intervals. This method can accurately calculate the harmonic amplitudes of disturbance currents, providing strong support for power system analysis, signal processing, and fault diagnosis.

[0101] Spectral centroid f c The formula for calculation is:

[0102]

[0103] Spectral Standard Deviation Std f The formula for calculation is:

[0104]

[0105] In the formula, f k A represents the frequency value at position k in the frequency sequence. k Represents frequency f k The corresponding amplitude is given at this location, and N is the defined spectral width.

[0106] The centroid of a spectrum can be used to detect spectral changes in current or voltage signals. When a fault or anomaly exists in a power system, such as a series arc fault, the spectral distribution of the current or voltage signal may change, causing a shift in the centroid. Therefore, by calculating the centroid, this change in spectral distribution can be captured, providing a strong basis for fault detection and diagnosis.

[0107] Spectral standard deviation can be used to assess the spectral stability of current or voltage signals. When instability or faults exist in a power system, the spectral distribution of current or voltage signals may become more dispersed, leading to an increase in the spectral standard deviation. Therefore, by calculating the spectral standard deviation, the degree of dispersion of this spectral distribution can be quantified, providing important reference information for assessing the stability and reliability of power systems.

[0108] S4. Input the time-frequency domain characteristics of the disturbance current into the preset fault identification model to detect and identify the cable series arc fault, and determine whether the fault causing the current disturbance is a series arc fault.

[0109] The preset fault identification model is a Stacking model composed of decision tree (DT), support vector machine (SVM), k-nearest neighbor (KNN), and logistic regression (LR) built based on ensemble learning algorithms. Historical data of the time-frequency domain features of current from past cable series arc faults are used to train the Stacking model. The training process of the Stacking model is as follows: Figure 4 As shown.

[0110] This improves the accuracy of fault identification. The Stacking model, by integrating multiple different base learners (such as decision trees, support vector machines, K-nearest neighbors, and logistic regression), fully leverages the advantages of each base learner, thereby enhancing fault identification accuracy. Each base learner extracts different information from the current time-frequency domain features, while the Stacking model effectively combines this information to form a more comprehensive fault identification capability. Furthermore, it enhances the model's generalization ability. Training the Stacking model using historical data of the current time-frequency domain features from real-world cable series arc faults enhances its generalization ability. This means the model can not only accurately identify fault patterns in the training data but also effectively predict and identify unseen fault data. This is crucial for fault detection and diagnosis in practical applications.

[0111] In practice, if a series arc fault is detected, the disturbance current waveform of the cable fault and the location of the faulty cable joint are sent to the backend, which then sends a fault warning message to the management and / or maintenance end.

[0112] In this way, by automatically sending fault information, the system can respond quickly after detecting a series arcing fault. This helps management and maintenance personnel to understand the fault situation in a timely manner, enabling them to take necessary intervention measures quickly to prevent the fault from worsening or causing a larger safety accident. Furthermore, the system can accurately send the location of the faulty cable joint to the backend, which greatly simplifies the fault diagnosis process. Maintenance personnel can go directly to the designated location for repairs without conducting extensive searches and investigations, thereby improving maintenance efficiency and reducing maintenance costs.

[0113] Compared with existing technologies, this invention improves the accuracy of monitoring and identification. By analyzing the similarity of current waveforms in adjacent current cycles using the Pearson correlation coefficient (PCC), this method can initially determine whether a current disturbance has occurred in the cable. This method effectively captures minute changes in the current waveform, thus accurately screening out currents suspected of causing cable series arc faults. Compared with traditional methods, this scheme improves the accuracy of monitoring and identifying cable fault arcs in the initial stage, laying a solid foundation for subsequent analysis and judgment. It also ensures the timeliness of series arc fault identification. This method analyzes current data from the starting point of the disturbance current and multiple subsequent cycles, enabling rapid capture of the initial characteristics of arc fault formation. Since arc faults are often weak in the early stages but gradually intensify over time, timely identification and measures are crucial to prevent fault expansion and protect equipment and personnel safety. This method allows for accurate identification in the early stages of arc fault development, ensuring the timeliness of series arc fault identification and providing a valuable time window for subsequent fault handling.

[0114] In addition, this method can reduce the amount of current data to be identified and lower monitoring latency. After confirming the occurrence of a current disturbance, this scheme only extracts data from multiple current cycles after the disturbance's inception point for further analysis. This approach avoids processing a large amount of irrelevant data, significantly reducing the amount of current data that needs to be calculated. Simultaneously, because only the data after the disturbance is processed, the entire monitoring and identification process is significantly shortened, thereby reducing monitoring latency and improving the system's real-time response capability. It also ensures the accuracy of identification. When a series arc fault occurs in a low-voltage distribution network cable, the current data will exhibit obvious time-frequency domain characteristic changes, such as an increase in high-frequency components and a decrease in periodicity. This method calculates the time-frequency domain characteristics of the disturbance current and inputs them into a preset fault identification model, fully utilizing these characteristic changes to accurately determine the fault type. This method not only considers the time-domain characteristics of the current but also combines frequency-domain characteristics, thus ensuring the accuracy and reliability of cable series arc fault identification.

[0115] This method, by comprehensively applying Pearson correlation coefficient analysis, time-frequency domain characteristic quantity calculation, and fault identification model, achieves efficient, accurate, and timely monitoring and identification of series arc faults in low-voltage distribution cables, effectively improving the safety and stability of the power system.

[0116] 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 the technical solutions. Those skilled in the art should understand that any modifications or equivalent substitutions to the technical solutions of the present invention without departing from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A method for judging series arc faults in low-voltage distribution cables based on disturbance current characteristics, characterized in that, Includes the following steps: S1. Real-time acquisition of three-phase current data of low-voltage distribution network cables; S2. For the acquired current data, the Pearson correlation coefficient (PCC) is used to analyze the similarity of the current waveforms in each adjacent current cycle, and to determine whether the cable is experiencing current disturbance. S3. When it is determined that a current disturbance has occurred, take the current cycle in which the current disturbance occurs as the disturbance start point, extract the current data of the disturbance start point and multiple current cycles thereafter, and record it as the disturbance current. The time-frequency domain characteristics of the extracted disturbance current are calculated. S4. Input the time-frequency domain characteristics of the disturbance current into the preset fault identification model to detect and identify the cable series arc fault, and determine whether the fault causing the current disturbance is a series arc fault.

2. The method for determining series arc faults in low-voltage distribution cables based on disturbance current characteristics as described in claim 1, characterized in that: In S2, when determining whether the cable has experienced a current disturbance, the CUSUM algorithm based on the cumulative sum of two sliding windows W1 and W2 is used to determine the current disturbance. Wherein, W1 is the front window, used to obtain the expected value E of the distribution and determine the upper bound threshold UB and the lower bound threshold LB; the expected value E is the average value calculated based on the current waveform data distribution within the front window W1; the values ​​of the upper bound threshold UB and the lower bound threshold LB are set based on historical data; W2 is the back window, used to calculate the offset accumulation value C of the data sequence within the window; the window length of both windows is M current cycles; Set the initial accumulated value of positive offset and negative offset initial accumulated value All values ​​are 0; after each sliding displacement of the window, based on the expected value E and the thresholds UB and LB, the offset of each point in W2 is calculated sequentially, and the positive offset is accumulated. and negative offset initial accumulated value The iteration process is as follows: In the formula, x k Let x be the value of the Pearson correlation coefficient for the k-th current cycle within W2; when x k When the positive offset accumulation value is greater than the upper threshold E+UB, Increase; when x k When the negative offset accumulation value is less than the lower threshold E-LB Decrease; If satisfied or This indicates that the current in the window has been disturbed; where DL is the preset offset limit.

3. The method for determining series arc faults in low-voltage distribution cables based on disturbance current characteristics as described in claim 2, characterized in that: In S2, when using the Pearson correlation coefficient (PCC) to analyze the similarity of current waveforms in adjacent current cycles, the formula for calculating the Pearson correlation coefficient for each current cycle is: In the formula, PCC m The Pearson correlation coefficient for the m-th current cycle is represented. This represents the instantaneous current value at the kth moment in the mth current cycle. This represents the average current in the m-th cycle; K is the number of instantaneous current values ​​in one current cycle.

4. The method for determining series arc faults in low-voltage distribution cables based on disturbance current characteristics as described in claim 1, characterized in that: In S3, when calculating the time-frequency domain characteristics of the extracted disturbance current, the calculated time-domain characteristics include zero rest time, peak-to-peak value, standard deviation, and kurtosis; the calculated frequency-domain characteristics include harmonic amplitude, spectral centroid, and spectral standard deviation.

5. The method for determining series arc faults in low-voltage distribution cables based on disturbance current characteristics as described in claim 4, characterized in that: When calculating time-domain characteristic quantities, the formula for calculating peak-to-peak value is: In the formula PPV k This represents the normalized peak-to-peak value of the k-th current cycle. and This represents the maximum and minimum values ​​of the current collected in the k-th current cycle.

6. The method for determining series arc faults in low-voltage distribution cables based on disturbance current characteristics as described in claim 4, characterized in that: When calculating time-domain features, the formulas for standard deviation and kurtosis are: In the formula, Std rms The standard deviation of the effective value of current, Kurt rms The kurtosis of the effective value of the current is represented by N, where N represents the total number of current cycles captured, and k represents the ordinal number of the cycle. I represents the average effective value of the intercepted current. m This represents the m-th current sample value within one cycle, where n represents the total number of current samples within one cycle.

7. The method for determining series arc faults in low-voltage distribution cables based on disturbance current characteristics as described in claim 4, characterized in that: When calculating frequency domain characteristic quantities, the calculation process for harmonic amplitude includes: First, the current signal of the disturbance current is filtered and preprocessed. Then, a short-time Fourier transform (STFT) is performed on the preprocessed current signal; Then, the average value of the results obtained from the short-time Fourier transform is calculated to obtain the harmonic amplitude of the disturbance current; among them, the Hanning window is selected as the window function of the short-time Fourier transform.

8. The method for determining series arc faults in low-voltage distribution cables based on disturbance current characteristics as described in claim 4, characterized in that: When calculating frequency domain features, the spectral centroid f c The formula for calculation is: Spectral Standard Deviation Std f The formula for calculation is: In the formula, f k A represents the frequency value at position k in the frequency sequence. k Represents frequency f k The corresponding amplitude is given at this location, and N is the defined spectral width.

9. The method for determining series arc faults in low-voltage distribution cables based on disturbance current characteristics as described in claim 1, characterized in that: In S4, the preset fault identification model is a Stacking model composed of decision tree DT, support vector machine SVM, K nearest neighbor KNN and logistic regression LR, which is built based on the logic of ensemble learning algorithms. The Stacking model is trained using historical data of the current time-frequency domain features of real cable series arc faults.

10. The method for determining series arc faults in low-voltage distribution cables based on disturbance current characteristics as described in claim 1, characterized in that: In S4, if a series arc fault is detected, the disturbance current waveform of the cable fault and the location of the faulty cable joint are sent to the backend, and the backend sends a fault warning message to the management end and / or maintenance end.

Images