Ground fault processing method based on arc suppression coil parallel connection switchable small resistor control

Abstract

The invention discloses a ground fault processing method based on arc suppression coil parallel connection switchable small resistor control. The method comprises the steps that an arc suppression coil is set to be in an automatic compensation working mode, and a small resistor is not input; neutral point zero-sequence voltage and zero-sequence current signals of all lines are monitored in real time; when the neutral point zero-sequence voltage exceeds a preset neutral point zero-sequence voltage normal value upper limit value, determining that a grounding fault occurs; when the occurrence time of the grounding fault continuously exceeds a first duration or intermittent multiple discrete accumulations meet a first condition, inputting a small resistor; according to the neutral point zero-sequence voltage and the zero-sequence current signal of each line, analyzing the type of a grounding fault and the position of the line; and the line protection device acts to cut off the fault line. According to the invention, a novel control strategy that the arc suppression coil is connected in parallel with the switchable small resistor during a single-phase grounding fault is provided, and automatic elimination of a transient fault and accurate identification and processing of various faults such as a continuous grounding fault and intermittent arc grounding are realized.

Description

Technical Field

[0001] This invention relates to the field of ground fault handling technology, and specifically to a ground fault handling method based on the control of a parallel switchable small resistor connected to an arc suppression coil. Background Technology

[0002] The importance of the safe, stable, and reliable operation of power distribution networks is increasing. The vast majority of faults occurring in power distribution networks are single-phase grounding faults. These include single-phase metallic (low-resistance) grounding faults, high-resistance grounding faults, and intermittent arc grounding faults. Rapid and accurate fault identification in power distribution networks is crucial for timely fault isolation and restoration of power to non-faulty areas, ensuring the safe, reliable, and stable operation of the power distribution network.

[0003] Currently, there is still a lack of research in comprehensively addressing various single-phase grounding faults, effectively handling different types of single-phase grounding faults, and improving the reliability of single-phase grounding protection. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention provides a grounding fault handling method based on parallel control of a switchable small resistor using an arc suppression coil, effectively enabling accurate identification and handling of various faults, including transient faults, continuous grounding faults, and intermittent arcing grounding. The technical solution is as follows: Firstly, a ground fault handling method based on parallel switchable small resistor control of arc suppression coil is provided, including the following steps: Set the arc suppression coil to automatic compensation mode and disable the small resistor; Real-time monitoring of neutral point zero-sequence voltage and zero-sequence current signals of each line; When the neutral point zero-sequence voltage exceeds the preset upper limit of the normal value of the neutral point zero-sequence voltage, a ground fault is determined to have occurred. When the time of a ground fault exceeds the first duration continuously or the first condition is met intermittently and repeatedly, a small resistor is switched on. The first duration is determined based on the range of the self-healing time of the instantaneous ground fault. Based on the analysis of the neutral point zero-sequence voltage and the zero-sequence current signals of each line, the ground fault is either low-resistance grounding or high-resistance grounding, and the line where low-resistance grounding or high-resistance grounding occurs. The faulty line is disconnected by using the line protection device.

[0005] In some implementations, determining a ground fault when the neutral point zero-sequence voltage exceeds a preset upper limit of the normal value for the neutral point zero-sequence voltage further includes: If the zero-sequence voltage of the neutral point does not exceed the preset upper limit of the normal value of the zero-sequence voltage of the neutral point, but the zero-sequence voltage signal changes abruptly and the zero-sequence current signal of the line changes abruptly, a ground fault is determined to have occurred.

[0006] In some embodiments, the step of analyzing the ground fault based on the neutral point zero-sequence voltage and the zero-sequence current signals of each line to determine whether it is a low-resistance ground fault or a high-resistance ground fault, and the line where the low-resistance ground fault or high-resistance ground fault occurs, includes: (31) Analyze the magnitude of the zero-sequence current of each line and analyze whether there is a line with a zero-sequence current greater than the first preset current threshold. If there is, proceed to step (32); otherwise, proceed to step (33). (32) For a line with a zero-sequence current greater than the first preset current threshold, determine that a low-resistance grounding has occurred and determine the location of the low-resistance fault line where the low-resistance grounding has occurred; for a low-resistance fault line, use the line protection device to disconnect the fault line. (33) For lines with zero-sequence current less than the first preset current threshold, they are identified as candidate high-resistance ground fault lines; whether a ground fault still occurs is determined based on whether the neutral point zero-sequence voltage exceeds the preset upper limit of the normal value of the neutral point zero-sequence voltage; if so, proceed to step (34), otherwise proceed to step (35). (34) Analyze the candidate high-resistance ground fault lines, determine the location of the high-resistance ground fault lines, and use the line protection device to disconnect the fault lines. (35) Grounding fault eliminated.

[0007] In some implementations, step (34), analyzing candidate high-resistance ground fault lines, includes: For each line, before the small resistor is applied, obtain the angle between the zero-sequence current and the zero-sequence voltage, and record it as the first angle. For each line, after a small resistor is applied, the angle between the zero-sequence current and the zero-sequence voltage is recorded as the second angle. For the line with the largest difference between the first and second included angles, analyze whether it is the line with the largest zero-sequence current among all lines. If so, it is identified as a high-resistance grounding fault line.

[0008] In some implementations, step (34), analyzing candidate high-resistance ground fault lines, includes: After the small resistor is put into operation, the phase data of the zero-sequence current signal and the zero-sequence voltage signal on each line are obtained. Based on the phase data of the zero-sequence current signal and the zero-sequence voltage signal on each line, cluster analysis is performed to obtain the set of normal lines and the set of candidate lines for high-resistance grounding faults. Based on the zero-sequence current signal of the candidate lines in the high-resistance grounding fault candidate line set, identification and analysis are performed to determine the line where the high-resistance grounding fault has occurred.

[0009] In some implementations, the identification and analysis of the zero-sequence current signals of candidate lines in the high-resistance grounding fault candidate line set to determine the line where the high-resistance grounding fault has occurred includes: The zero-sequence current signal based on the candidate line is decomposed by wavelet packet to obtain multiple sub-signals of different frequencies. The sub-signals of different frequencies are discretely sampled to obtain signal values ​​of different frequencies at different times, and the frequency-time signal matrix is ​​obtained. The first feature is obtained by inputting a frequency-time signal matrix into a convolutional network to extract deep features. Simultaneously, the second feature is obtained by inputting the frequency-time signal matrix into an attention-based encoding network. The convolutional network comprises three residual convolutional units, with the input of each unit directly connected to its output. Each convolutional unit sequentially includes a convolutional layer, a normalization layer, an activation function layer, another convolutional layer, and a normalization layer. The attention-based encoding network is constructed based on a Transformer model encoder and sequentially includes a self-attention layer and a feedforward network. The first and second features are concatenated and input into the classification layer for identification and analysis to determine whether a high-resistance grounding fault has occurred.

[0010] In some embodiments, the step of analyzing the ground fault based on the neutral point zero-sequence voltage and the zero-sequence current signals of each line to determine whether it is a low-resistance ground fault or a high-resistance ground fault, and the line where the low-resistance ground fault or high-resistance ground fault occurs, includes: When the fault occurs, the zero-sequence voltage signal and the zero-sequence current signal are acquired within the first preset time window before and after the fault, and are denoted as the first zero-sequence voltage signal and the first zero-sequence current signal. When the small resistor is connected, the zero-sequence voltage signal and the zero-sequence current signal are obtained within a second preset time window before and after the connection, and are denoted as the second zero-sequence voltage signal and the second zero-sequence current signal. Based on the signal within the first preset time window, transient change signals and stationary signals are identified, and the time period of transient change signal after fault, the time period of signal before fault, and the time period of stationary signal after fault are obtained. The transient change signal represents a change signal without a fixed period, including the gradually increasing change signal of zero-sequence voltage after fault and the oscillating change signal of zero-sequence current after fault. Based on the signal time period before the fault, the transient change signal time period after the fault, and the steady-state signal time period after the fault within the first preset time window, the signal characteristics and signal change trend characteristics of each signal segment, as well as the signal change characteristics of adjacent signal segments, are analyzed within the first preset time window. Based on the signal within the second preset time window, transient change signals and stable sinusoidal wave signals are identified, and the signal segment before the small resistor is put into operation, the transient change signal segment, and the stable signal segment are obtained. The transient change signal represents a change signal without a fixed period, including the decrease of the zero-sequence voltage after the small resistor is put into operation and the increase of the zero-sequence current after the small resistor is put into operation. Based on the signal segments before the small resistance is applied, the transient signal segments, and the stable signal segments within the second preset time window, the signal characteristics and signal change trend characteristics of each signal segment, as well as the signal change characteristics of adjacent signal segments, are analyzed within the second preset time window. Based on the signal characteristics and signal change trend characteristics within each signal segment within the first and second preset time windows, as well as the signal change characteristics of adjacent signal segments, these are used as feature vectors and input into a preset identification and classification model to identify whether a ground fault has occurred on each line and the type of ground fault that has occurred. The preset identification and classification model is a model trained by using zero-sequence voltage signals and zero-sequence current signals of historical low-resistance and high-resistance ground faults as training data, and by signal segment identification and signal segment feature input to the model.

[0011] In some implementations, the identification of transient change signals and stationary signals includes: For a signal within a preset time window, the envelope of the signal is obtained through Hilbert transform, and the time-frequency data is obtained through wavelet transform. Based on the envelope stationarity and the frequency component distribution at different times in the time-frequency data, transient and stationary signals can be identified.

[0012] In some implementations, the method for obtaining the signal characteristics of the transient signal time period after a fault and the transient signal segment after applying a small resistor includes: Based on the wavelet packet decomposition of the transient current signal, multiple sub-signals of different frequencies are obtained; For multiple sub-signals of different frequencies, the transverse characteristics of a single frequency sub-signal are analyzed laterally; the longitudinal characteristics of multiple frequency sub-signals at the same time are analyzed longitudinally. The signal features are generated based on the horizontal and vertical features.

[0013] Secondly, a ground fault handling system based on parallel switchable small resistor control of arc suppression coil is provided, the system comprising: The initial control unit of the switch is used to set the arc suppression coil to automatic compensation mode and to disable the small resistance. The signal acquisition unit is used to monitor the neutral point zero-sequence voltage and the zero-sequence current signal of each line in real time. The preliminary analysis unit is used to determine that a ground fault has occurred when the neutral point zero-sequence voltage exceeds the preset upper limit of the normal value of the neutral point zero-sequence voltage. The secondary control unit of the switch is used to switch on a small resistor when the time of the ground fault exceeds a first duration continuously or intermittently meets a first condition multiple times. The first duration is determined based on the range of the self-healing time of the instantaneous ground fault. The ground fault analysis unit is used to analyze whether a ground fault is low-resistance grounding or high-resistance grounding, and the lines that have experienced low-resistance grounding or high-resistance grounding, based on the neutral point zero-sequence voltage and the zero-sequence current signals of each line. The fault handling unit is used to disconnect the faulty line by utilizing the operation of the line protection device.

[0014] This invention discloses a ground fault handling method based on parallel switchable small resistor control of an arc suppression coil, which has the following beneficial effects: By employing a novel neutral point grounding method with a parallel switchable small resistor connected to an arc suppression coil, a novel control strategy for single-phase ground faults is proposed. This achieves automatic elimination of transient faults and rapid elimination of continuous low-resistance ground faults. For intermittent ground faults, a fault time accumulation method is used to ensure timely operation of the line's zero-sequence protection. For continuous high-resistance ground faults, a small resistor is connected, and the line is identified and disconnected in conjunction with a high-resistance fault identification algorithm. This invention enables timely analysis and handling of various ground faults, improving the reliability of single-phase ground fault protection. Attached Figure Description

[0015] Figure 1 This is a schematic flowchart of the ground fault handling method based on parallel switchable small resistor control of arc suppression coil in the embodiments of this application; Figure 2 This is a schematic diagram of the ground fault handling system based on the parallel switchable small resistor control of the arc suppression coil in the embodiments of this application. Detailed Implementation

[0016] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0017] This application provides a ground fault handling method based on parallel switchable small resistor control of arc suppression coil, the method including the following steps: Step 1: Set the arc suppression coil to automatic compensation mode and disable the small resistor; Step 2: Monitor the neutral point zero-sequence voltage and the zero-sequence current signal of each line in real time; Step 3: When the neutral point zero-sequence voltage exceeds the preset upper limit of the normal value of the neutral point zero-sequence voltage, a ground fault is determined to have occurred; Step 4: When the time of the ground fault exceeds the first duration continuously or the first condition is met intermittently and repeatedly, a small resistor is switched on (when the time of the ground fault exceeds the first preset time, it is determined that a permanent ground fault has occurred). The first duration is determined based on the range of self-healing time of instantaneous ground faults. Step 5: Analyze the grounding fault based on the neutral point zero-sequence voltage and the zero-sequence current signal of each line to determine whether it is a low-resistance grounding or a high-resistance grounding, and identify the line where the grounding fault occurred. Step 6: Use the line protection device to disconnect the faulty line.

[0018] It should be noted that the ideal value of the neutral point zero-sequence voltage is 0 under fault-free conditions. However, due to slight imbalances or measurement errors, deviations may occur. For example, in this embodiment, the normal range of the neutral point zero-sequence voltage can be estimated based on a large number of actual measurements under fault-free conditions. For instance, the upper limit of the normal range is 15% of the phase voltage. If it is lower than the preset upper limit of the normal value of the neutral point zero-sequence voltage, it is considered to be within the normal range. If it exceeds the preset upper limit of the normal value of the neutral point zero-sequence voltage, it is initially considered to be a ground fault. In this embodiment, a novel neutral point grounding method with a parallel switchable small resistor connected to the arc suppression coil is proposed. This method provides a novel control strategy for single-phase ground faults, achieving automatic elimination of transient faults and rapid elimination of continuous low-resistance ground faults. Compared to the arc suppression coil grounding method, it can quickly isolate faults; compared to the small resistance grounding method, it significantly improves power supply reliability. It is understood that the parallel switchable small resistor grounding system of the arc suppression coil in this application embodiment achieves the following: (1) In the event of an instantaneous grounding, the arc suppression coil automatically compensates, the small resistor is not activated, and 90% of instantaneous grounding can be automatically eliminated. (2) When the grounding fault continues for more than a first duration, such as 3~10s, it is determined to be a permanent grounding fault, the small resistor is activated, and the line protection device is activated to disconnect the faulty line. (3) When the grounding fault intermittently and repeatedly meets the first condition, the small resistor is activated. This situation is for intermittent arcing grounding. The small resistor is activated according to the time accumulation algorithm, and the line protection device is activated to disconnect the faulty line. For intermittent grounding faults, the fault time accumulation method is adopted. When multiple consecutive, short-term, intermittent groundings exceeding the set time threshold occur within a certain period of time, the small resistor control strategy should no longer wait for the arc suppression coil delay, but immediately activate the small resistor so that the zero-sequence protection of the line can operate. (4) When there is a continuous high-resistance grounding, the small resistor is activated, and the line is determined and disconnected in conjunction with the high-resistance fault identification algorithm. The reliability of single-phase grounding protection is improved in this application embodiment.

[0019] Among them, the intermittent, discrete accumulation of ground faults over time satisfies the first condition, specifically including: (A1) Obtain the fault status at the current time t. If a ground fault occurs, generate a scaling time window starting from the current time. The scaling time window has the current time as the fixed left end and the right end of the time window is continuously updated as time is updated. The latest time is used as the right end of the scaling time window. (A2) Obtain the fault status at time t+1, and use the latest time as the right end of the scaling time window. Record the initial value of L as 0. If no ground fault occurs at that time, the value of L is incremented by 1. If a ground fault occurs at the current time, the value of L is reset to 0. (A3) If the value of L does not exceed the third preset duration at the current time, proceed to steps (A4)-(A5); otherwise, proceed to step (A6). (A4) If the current scaling time window is greater than the first preset window length and less than the second preset window length, then obtain the continuous duration of a single ground fault occurrence and the interval duration between two adjacent ground fault occurrences within the current scaling time window. Based on the interval duration between two adjacent ground fault occurrences, determine the mean and variance of the interval duration between two adjacent ground fault occurrences. Based on the mean and variance of the interval duration between two adjacent ground fault occurrences, update the second preset window length. (A5) Within the updated second preset window length, if the proportion of the cumulative duration of intermittent and discrete ground faults exceeds the second preset threshold, then a small resistor is activated; otherwise, a fixed time window is generated, with the latest time as the right end of the fixed time window and the first preset window length as the left end of the fixed time window; if the proportion of the cumulative duration of intermittent and discrete ground faults exceeds the second preset threshold within the fixed time window, then a small resistor is activated. (A6) When the value of L exceeds the third preset duration at the current time, eliminate the scaling time window and proceed to steps (A1)-(A5) to analyze the next time.

[0020] In this embodiment, a scaling time window is generated when the first ground fault occurs. Based on the distribution pattern of each ground fault occurring within the subsequent first preset window length, an updated second preset window length is flexibly determined. If the proportion of intermittent, discrete, cumulative ground fault occurrences within the second preset window length exceeds a second preset threshold, a small resistor is applied. It can be understood that if the mean and variance of the intervals between two adjacent ground fault occurrences within the first preset window length indicate that ground faults occur frequently and are unstable, with significant fluctuations, a shorter judgment period can be used to update the second preset window length to a third preset window length smaller than the original predetermined second preset window length. This third preset window length can be the length from the fixed left end of the scaling time window to the current time. If the mean and variance of the intervals between two adjacent ground fault occurrences within the first preset window length indicate relatively stable fluctuations in ground fault occurrence time, a longer judgment period can be used to update the second preset window length to the original predetermined second preset window length. Here, the first preset window length is less than the third preset window length, which is less than the second preset window length. In step (A6), if the value of L at the current moment exceeds the third preset duration, the scaling time window is eliminated, and the analysis of (A1)-(A5) is re-entered from the next moment, that is, the fixed left end position of the scaling time window is determined again.

[0021] In one embodiment, step 3 above, where a ground fault is determined to have occurred when the neutral point zero-sequence voltage exceeds the preset upper limit of the normal value of the neutral point zero-sequence voltage, further includes: If the zero-sequence voltage of the neutral point does not exceed the preset upper limit of the normal value of the zero-sequence voltage of the neutral point, but the zero-sequence voltage signal changes abruptly and the zero-sequence current signal of the line changes abruptly, a ground fault is determined to have occurred.

[0022] In this embodiment, it is considered that under high-resistance grounding fault conditions, the neutral point zero-sequence voltage is relatively small, and the zero-sequence current of each line is also relatively small, which may be less than the preset grounding fault zero-sequence voltage threshold (i.e., the preset upper limit of the normal value of the neutral point zero-sequence voltage), resulting in the grounding fault being unable to be identified and cleared, causing a continuous impact on the system. In this embodiment, to avoid omitting high-resistance grounding faults, the changes in the neutral point zero-sequence voltage and the changes in the zero-sequence current signal on the line are combined as the initial basis for determining the occurrence of grounding faults. When the neutral point zero-sequence voltage does not exceed the preset upper limit of the normal value of the neutral point zero-sequence voltage, but the zero-sequence voltage signal changes abruptly and there is a change in the zero-sequence current signal of the line, a grounding fault is determined to have occurred. It can be understood that in this embodiment, the arc suppression coil grounding is in automatic compensation mode and the small resistor is not engaged. When a fault occurs, in the circuit system of the faulty circuit, non-faulty line, and arc suppression coil grounding line, the zero-sequence voltage will rise and gradually stabilize immediately after the fault occurs (e.g., within 200 microseconds after the fault occurs), and the zero-sequence current will oscillate and change and will reach a stable state after a certain moment. Therefore, capturing these abrupt changes in zero-sequence voltage and zero-sequence current signals can be used to assist in determining the occurrence of ground faults. This avoids missing ground faults, especially high-resistance ground faults, by relying solely on zero-sequence voltage or zero-sequence current thresholds.

[0023] In one embodiment, step 5 above, which involves analyzing the ground fault based on the neutral point zero-sequence voltage and the zero-sequence current signals of each line to determine whether it is a low-resistance ground fault or a high-resistance ground fault, and identifying the line experiencing either a low-resistance ground fault or a high-resistance ground fault, includes: (31) Analyze the magnitude of the zero-sequence current of each line and analyze whether there is a line with a zero-sequence current greater than the first preset current threshold. If there is, proceed to step (32); otherwise, proceed to step (33). (32) For a line with a zero-sequence current greater than the first preset current threshold, determine that a low-resistance grounding has occurred and determine the location of the low-resistance fault line where the low-resistance grounding has occurred; for a low-resistance fault line, use the line protection device to disconnect the fault line. (33) For lines with zero-sequence current less than the first preset current threshold, they are identified as candidate high-resistance ground fault lines; whether a ground fault still occurs is determined based on whether the neutral point zero-sequence voltage exceeds the preset upper limit of the normal value of the neutral point zero-sequence voltage; if so, proceed to step (34), otherwise proceed to step (35). (34) Analyze the candidate high-resistance ground fault lines, determine the location of the high-resistance ground fault lines, and use the line protection device to disconnect the fault lines. (35) Grounding fault eliminated.

[0024] In this embodiment, considering that a low-resistance grounding fault occurs when the zero-sequence current value of the line is relatively large, a low-resistance grounding fault is determined based on the line whose zero-sequence current is greater than a first preset current threshold. Considering that a line whose zero-sequence current is less than the first preset current threshold may have a high-resistance grounding fault, a high-resistance grounding fault analysis is performed based on the neutral point zero-sequence voltage and each line to further identify and determine the line where the high-resistance grounding fault occurs, thus reducing the grounding fault missed detection rate.

[0025] In one implementation, step (34) above involves analyzing candidate high-resistance ground fault lines, including: Step 3401: Obtain the angle between the zero-sequence current and the zero-sequence voltage of each line before the small resistor is applied, and record it as the first angle. Step 3402: Obtain the angle between the zero-sequence current and the zero-sequence voltage of each line after the small resistor is connected, and record it as the second angle. Step 3403: For the line with the largest difference between the first included angle and the second included angle, analyze whether it is the line with the largest zero-sequence current among all lines. If so, it is determined to be a high-resistance grounding fault line.

[0026] In this embodiment, the change in phase of the faulty line after the small resistor is applied, and the influence of the resistive current of the small resistor on the magnitude of the faulty line current, are utilized. Before the small resistor is applied, the angle between the zero-sequence current and zero-sequence voltage of the faulty line is approximately 90 degrees. After the small resistor is applied, the angle between the zero-sequence current and zero-sequence voltage of the faulty line is approximately 180 degrees, while the angle between the zero-sequence current and zero-sequence voltage of the non-faulty line remains approximately 90 degrees. For the faulty line, the difference between the first angle and the second angle is significantly greater than that for the non-faulty line.

[0027] In one implementation, step (34) above involves analyzing candidate high-resistance ground fault lines, including: Step 3411: Obtain the phase data of zero-sequence current signal and zero-sequence voltage signal on each line after the small resistor is put into operation. Based on the phase data of zero-sequence current signal and zero-sequence voltage signal on each line, perform cluster analysis to obtain the set of normal lines and the set of candidate lines for high-resistance grounding faults. Step 3412: Based on the zero-sequence current signal of the candidate lines in the high-resistance grounding fault candidate line set, perform identification analysis to determine the line where the high-resistance grounding fault occurred.

[0028] In this embodiment, considering that the phase difference between the zero-sequence current and zero-sequence voltage of non-faulty lines and faulty lines is significantly different after the small resistor is put into operation, cluster analysis of the phase data of the zero-sequence current signal and zero-sequence voltage signal on the line is first performed. This can preliminarily determine the zero-sequence current signal and zero-sequence voltage signal that may be faulty lines. Furthermore, based on the zero-sequence current signal of the candidate lines in the high-resistance grounding fault candidate line set, signal feature identification analysis is performed to determine the line where the high-resistance grounding fault occurred.

[0029] Specifically, in step 3412 above, based on the zero-sequence current signal of the candidate lines in the high-resistance grounding fault candidate line set, identification and analysis are performed to determine the line where the high-resistance grounding fault has occurred, including: Step 34121: Based on the zero-sequence current signal of the candidate line, wavelet packet decomposition is performed to obtain multiple sub-signals of different frequencies. Step 34122: Discretely sample the sub-signals of different frequencies to obtain signal values ​​of different frequencies at different times, and obtain the frequency-time signal matrix; Step 34123: Based on the frequency-time signal matrix, input convolutional network to extract deep features to obtain the first feature; simultaneously, input attention-based encoding network based on the frequency-time signal matrix to obtain the second feature. The convolutional network includes three residual structure convolutional units, the input of each convolutional unit is directly connected to the output of the convolutional unit, and each convolutional unit includes a convolutional layer, a normalization layer, an activation function layer, another convolutional layer, and a normalization layer in sequence. The attention-based encoding network is based on the encoder of the Transformer model and includes a self-attention layer and a feedforward network in sequence. Step 34124: Based on the first feature and the second feature, the data is input into the classification layer for identification and analysis to determine whether a high-resistance grounding fault has occurred.

[0030] In this embodiment, wavelet packet decomposition of the zero-sequence current signal is used to obtain the time-frequency signal matrix of the zero-sequence current signal. At the same time, dual feature extraction is performed based on convolutional network and attention mechanism-based coding network to improve the extraction and identification capability of zero-sequence current features of high-resistance ground fault lines.

[0031] In one embodiment, step 5 above, based on the neutral point zero-sequence voltage and the zero-sequence current signal of each line, analyzes whether the ground fault is a low-resistance ground fault or a high-resistance ground fault, and identifies the lines experiencing low-resistance ground faults or high-resistance ground faults, including: Step 501: Obtain the zero-sequence voltage signal and zero-sequence current signal within a first preset time window before and after the fault occurs, and denoted as the first zero-sequence voltage signal and the first zero-sequence current signal. It can be understood that the first preset time window includes the signal for a period of time before the fault and the signal for a period of time after the fault, and includes the signal of the transient process gradually changing into a steady state process after the fault. Thus, the signal for a period of time before the fault, the transient change signal after the fault, and the steady state change signal can be obtained.

[0032] Step 502: Obtain the zero-sequence voltage signal and zero-sequence current signal within a second preset time window before and after the small resistor is connected, denoted as the second zero-sequence voltage signal and the second zero-sequence current signal. It can be understood that the second preset time window includes the signal for a period of time before the small resistor is connected and the signal for a period of time after the small resistor is connected, and also includes the signal of the transient process gradually changing to the steady-state process after the small resistor is connected. Thus, the signal for a period of time before the small resistor is connected, the transient change signal after the small resistor is connected, and the steady-state change signal can be obtained.

[0033] Step 503: Based on the signal within the first preset time window, identify transient change signals and stable signals, and obtain the time period of transient change signals after the fault, the time period of signals before the fault, and the time period of stable signals after the fault. The transient change signals represent change signals without a fixed period, including the gradually increasing change signal of zero-sequence voltage after the fault and the oscillating change signal of zero-sequence current after the fault. Step 504: Based on the signal time period before the fault, the transient change signal time period after the fault, and the steady-state signal time period after the fault within the first preset time window, analyze the signal characteristics and signal change trend characteristics of each signal segment and the signal change characteristics of adjacent signal segments within the first preset time window. The signal change trend characteristics within each signal segment include features used to characterize the change trend, such as signal fluctuation, smoothness, and increase. The signal change characteristics of adjacent new signal segments include the growth value of the average signal amplitude of the last preset time period of the previous signal segment and the average signal amplitude of the first preset time period of the next signal segment. Step 505: Based on the signal within the second preset time window, identify transient change signals and stable sinusoidal wave signals, and obtain the signal segment before the small resistor is connected, the transient change signal segment, and the stable signal segment. The transient change signal represents a change signal without a fixed period, including the decrease in zero-sequence voltage and the increase in zero-sequence current after the small resistor is connected. Step 506: Based on the signal segments before the small resistance is applied, the transient signal segments, and the stable signal segments within the second preset time window, analyze the signal characteristics and signal change trend characteristics of each signal segment, as well as the signal change characteristics of adjacent signal segments within the second preset time window. Step 507: Based on the signal characteristics and signal change trend characteristics within each signal segment within the first and second preset time windows, and the signal change characteristics of adjacent signal segments, these are used as feature vectors and input into a preset identification and classification model to identify whether a ground fault has occurred on each line and the type of ground fault. The preset identification and classification model is based on training data using zero-sequence voltage and zero-sequence current signals of historical low-resistance and high-resistance ground faults. After signal segment identification and signal segment feature analysis, the signal segment features are input into the model and trained. It can be understood that the training data is based on zero-sequence voltage and zero-sequence current signals of historical low-resistance and high-resistance ground faults. After the training data is processed through signal segment identification and signal segment feature acquisition as in steps 501-505, it is input into the model. Low-resistance and high-resistance ground fault types are used as labels. After training the model, the trained model is the preset identification and classification model.

[0034] In this embodiment, low-resistance and high-resistance fault lines are better identified and analyzed. This is achieved by fully utilizing the signal change characteristics at the instant of the fault, the steady-state characteristics over a period after the fault, the impact of the small resistor on the zero-sequence voltage and zero-sequence current signals of each line before and after its application, and the steady-state characteristics after the small resistor is applied, thus avoiding misjudgments and missed judgments caused by a single feature. It is understood that at the instant of a fault, whether it is a low-resistance or high-resistance fault, the zero-sequence current and zero-sequence voltage on each line, including both faulty and non-faulty lines, will change. The difference between low-resistance and high-resistance faults lies in the magnitude of these changes (including the magnitude of the amplitude change and the magnitude of the phase angle change). After a brief transient change following the fault, a steady state gradually forms. The application of a small resistor also produces distinct changes in both faulty and non-faulty lines. In this embodiment, the signal change characteristics at the instant of a fault, the steady-state characteristics over a period of time after the fault, the influence of the small resistor on the zero-sequence voltage and zero-sequence current signals of each line before and after its application, and the steady-state characteristics over a period of time after the small resistor is applied are combined. If a feature in one dimension causes a measurement error that affects the judgment of the fault type and location, it does not affect the judgment of the fault type and location by the features in the remaining dimensions. This reduces the error of single-dimensional feature measurement error on the fault analysis results. For example, for a low-resistance grounding fault, after the fault occurs, the zero-sequence voltage of the faulty line will gradually rise and stabilize at a relatively large stable value, such as 70% phase voltage. The zero-sequence current of the faulty line will undergo a significant amplitude oscillation process and gradually stabilize at a relatively small zero-sequence current value (the residual current value after arc suppression coil compensation). The oscillation process of the zero-sequence current of the non-faulty line and the circuit phase are different from those of the faulty line. After the small resistor is applied, the zero-sequence voltage of the faulty line will gradually decrease slightly and stabilize, and the zero-sequence current of the faulty line will gradually increase significantly and stabilize, and the current direction is different from that of the non-faulty line. The zero-sequence voltage and zero-sequence current of non-faulty lines will both decrease slightly.

[0035] In one embodiment, identifying transient change signals and stationary signals in steps 503 and 505 above includes: Step 5031: For a signal within a preset time window, obtain the envelope of the signal through Hilbert transform and obtain time-frequency data through wavelet transform; Step 5032: Based on the envelope stationarity and the frequency component distribution at different times in the time-frequency data, identify transient change signals and stationary signals.

[0036] In this embodiment of the application, transient signals and stationary signals are determined based on the envelope and frequency distribution in the signal. It can be understood that the envelope of the transient signal is less stable and has more high-frequency components.

[0037] In one embodiment, the method for obtaining the signal characteristics of the transient change signal time period after the fault and the transient change signal segment after the small resistor is applied in steps 504 and 506 above includes: Step 5041: Based on the transient current signal, multiple sub-signals of different frequencies are obtained through wavelet packet decomposition. Step 5042: For multiple sub-signals of different frequencies, analyze the horizontal characteristics of a single frequency sub-signal in the horizontal direction; and analyze the vertical characteristics of multiple frequency sub-signals at the same time in the vertical direction. Step 5043: Generate the signal features based on the horizontal and vertical features.

[0038] In this embodiment of the application, the lateral analysis of the lateral characteristics of a sub-signal at a single frequency includes: acquiring the signal energy of the same sub-signal over the entire time period (i.e., the integral or summation of the sub-signal over the entire time period), as the first lateral characteristic; and calculating, for the same sub-signal, the proportion P of the signal energy at each moment to the total signal energy, based on... As the second lateral feature, among which This represents the proportion of the signal energy of the i-th sub-signal at time k to the total signal energy. Longitudinal analysis of the longitudinal characteristics of sub-signals at multiple frequencies within the same time period includes: simultaneously aligning and dividing the sub-signals into multiple small window sub-signals, obtaining P, which represents the proportion of the energy of each sub-signal within the small window to the total signal energy of all sub-signals within the same time period, based on... As a vertical feature, among which This represents the percentage of the energy of the i-th sub-signal within the j-th small window relative to the total energy of all sub-signals within the j-th small window.

[0039] In one embodiment, the feature vector in step 507 above further includes feature data characterizing the signal phase within each signal segment.

[0040] This application provides a ground fault handling system based on parallel switchable small resistor control of arc suppression coil, the system comprising: The initial control unit of the switch is used to set the arc suppression coil to automatic compensation mode and to disable the small resistance. The signal acquisition unit is used to monitor the neutral point zero-sequence voltage and the zero-sequence current signal of each line in real time. The preliminary analysis unit is used to determine that a ground fault has occurred when the neutral point zero-sequence voltage exceeds the preset upper limit of the normal value of the neutral point zero-sequence voltage. The secondary control unit of the switch is used to switch on a small resistor when the time of the ground fault exceeds a first duration continuously or intermittently meets a first condition multiple times. The first duration is determined based on the range of the self-healing time of the instantaneous ground fault. The ground fault analysis unit is used to analyze whether a ground fault is low-resistance grounding or high-resistance grounding, and the lines that have experienced low-resistance grounding or high-resistance grounding, based on the neutral point zero-sequence voltage and the zero-sequence current signals of each line. The fault handling unit is used to disconnect the faulty line by utilizing the operation of the line protection device.

[0041] Specific limitations regarding the ground fault handling system based on parallel switchable small resistor control of arc suppression coils can be found in the limitations of the ground fault handling method based on parallel switchable small resistor control of arc suppression coils mentioned above, and will not be repeated here. Each unit in the aforementioned ground fault handling system based on parallel switchable small resistor control of arc suppression coils can be implemented entirely or partially through software, hardware, or a combination thereof. These units can be embedded in hardware or independently of the processor in a computer device, or stored in software in the memory of a computer device, so that the processor can call and execute the corresponding operations of each unit.

[0042] This invention is not limited to the specific embodiments described above. Any modifications made by those skilled in the art based on the above concept without creative effort are within the scope of protection of this invention.

Claims

1. A method for handling ground faults based on arc suppression coil in parallel with a controllable small resistance, characterized in that, The method comprises the following steps: setting the arc suppression coil in an automatic compensation mode and not putting in a small resistance; monitoring the neutral point zero sequence voltage and the zero sequence current signals of each line in real time; determining that a ground fault occurs when the neutral point zero sequence voltage exceeds a preset upper limit of a normal value of the neutral point zero sequence voltage; putting in the small resistance when the time of the ground fault continuously exceeds a first time length or intermittently meets a first condition multiple times, and the first time length is determined based on a range of instantaneous ground fault self-healing time length; analyzing whether the ground fault is a low resistance ground or a high resistance ground and a line in which the low resistance ground or the high resistance ground occurs based on the neutral point zero sequence voltage and the zero sequence current signals of each line; and cutting off the fault line by using a line protection device.

2. The method of claim 1, wherein the method is characterized by, The step of determining that a ground fault occurs when the neutral point zero sequence voltage exceeds a preset upper limit of a normal value of the neutral point zero sequence voltage further comprises: determining that a ground fault occurs when the neutral point zero sequence voltage does not exceed the preset upper limit of the normal value of the neutral point zero sequence voltage, but a sudden change occurs in the zero sequence voltage signal and a sudden change occurs in the zero sequence current signal of a line.

3. The method according to claim 2, wherein the arc-suppression coil parallel- connectable small-resistance control based ground fault processing method is characterized by, The step of analyzing whether the ground fault is a low resistance ground or a high resistance ground and a line in which the low resistance ground or the high resistance ground occurs based on the neutral point zero sequence voltage and the zero sequence current signals of each line comprises: (31) analyzing the zero sequence current of each line, analyzing whether there is a line in which the zero sequence current is greater than a first preset current threshold, if there is, proceeding to step (32), otherwise, proceeding to step (33); (32) for the line in which the zero sequence current is greater than the first preset current threshold, determining that a low resistance ground occurs, and determining the position of the low resistance fault line in which the low resistance ground occurs; for the low resistance fault line, cutting off the fault line by using a line protection device; (33) for the line in which the zero sequence current is less than the first preset current threshold, determining that it is a candidate high resistance ground fault line; determining whether a ground fault still occurs based on whether the neutral point zero sequence voltage exceeds a preset upper limit of a normal value of the neutral point zero sequence voltage; if yes, proceeding to step (34), otherwise, proceeding to step (35); (34) analyzing the candidate high resistance ground fault line to determine the position of the high resistance ground fault line, and cutting off the fault line by using a line protection device; (35) eliminating the ground fault.

4. The method according to claim 3, wherein the arc-suppression coil parallel- connectable small-resistance control type ground fault processing method is characterized by, The step of analyzing the candidate high resistance ground fault line in step (34) comprises: obtaining the angle between the zero sequence current and the zero sequence voltage of each line before the small resistance is put in, denoted as a first angle; obtaining the angle between the zero sequence current and the zero sequence voltage of each line after the small resistance is put in, denoted as a second angle; for the line in which the first angle and the second angle differ the most, analyzing whether it is the line in which the zero sequence current is the largest among all lines, if yes, determining that it is a high resistance ground fault line.

5. The method of claim 3, wherein the method is characterized by, The step of analyzing the candidate high resistance ground fault line in step (34) comprises: obtaining the phase data of the zero sequence current signal and the zero sequence voltage signal of each line after the small resistance is put in, and performing clustering analysis based on the phase data of the zero sequence current signal and the zero sequence voltage signal of each line to obtain a normal line set and a high resistance ground fault candidate line set. Based on the zero-sequence current signal of the candidate lines in the high-resistance grounding fault candidate line set, identification and analysis are performed to determine the line where the high-resistance grounding fault has occurred.

6. The method of claim 5, wherein the method is characterized by, The process of identifying and analyzing the zero-sequence current signals of candidate lines in the high-resistance grounding fault candidate line set to determine the line experiencing the high-resistance grounding fault includes: The zero-sequence current signal based on the candidate line is decomposed by wavelet packet to obtain multiple sub-signals of different frequencies. The sub-signals of different frequencies are discretely sampled to obtain signal values ​​of different frequencies at different times, and the frequency-time signal matrix is ​​obtained. The first feature is obtained by inputting a frequency-time signal matrix into a convolutional network to extract deep features. Simultaneously, the second feature is obtained by inputting the frequency-time signal matrix into an attention-based encoding network. The convolutional network comprises three residual convolutional units, with the input of each unit directly connected to its output. Each convolutional unit sequentially includes a convolutional layer, a normalization layer, an activation function layer, another convolutional layer, and a normalization layer. The attention-based encoding network is constructed based on a Transformer model encoder and sequentially includes a self-attention layer and a feedforward network. The first and second features are concatenated and then input into the classification layer for identification and analysis to determine whether a high-resistance grounding fault has occurred.

7. The method of claim 2, wherein the method is characterized by, The analysis based on the neutral point zero-sequence voltage and the zero-sequence current signals of each line indicates that the ground fault is either a low-resistance ground fault or a high-resistance ground fault, and the lines experiencing either low-resistance ground faults or high-resistance ground faults include: When the fault occurs, the zero-sequence voltage signal and the zero-sequence current signal are acquired within the first preset time window before and after the fault, and are denoted as the first zero-sequence voltage signal and the first zero-sequence current signal. When the small resistor is connected, the zero-sequence voltage signal and the zero-sequence current signal are obtained within a second preset time window before and after the connection, and are denoted as the second zero-sequence voltage signal and the second zero-sequence current signal. Based on the signal within the first preset time window, transient change signals and stationary signals are identified, and the time period of transient change signal after fault, the time period of signal before fault, and the time period of stationary signal after fault are obtained. The transient change signal represents a change signal without a fixed period, including the gradually increasing change signal of zero-sequence voltage after fault and the oscillating change signal of zero-sequence current after fault. Based on the signal time period before the fault, the transient change signal time period after the fault, and the steady-state signal time period after the fault within the first preset time window, the signal characteristics and signal change trend characteristics of each signal segment, as well as the signal change characteristics of adjacent signal segments, are analyzed within the first preset time window. Based on the signal within the second preset time window, transient change signals and stable sinusoidal wave signals are identified, and the signal segment before the small resistor is put into operation, the transient change signal segment, and the stable signal segment are obtained. The transient change signal represents a change signal without a fixed period, including the decrease of the zero-sequence voltage after the small resistor is put into operation and the increase of the zero-sequence current after the small resistor is put into operation. The signal features and signal change trend features in each signal segment and the signal change features of adjacent signal segments in the second preset time length time window are analyzed based on the signal segments before the small resistance input, the transient change signal segments and the smooth signal segments in the second preset time length time window. The signal features and signal change trend features in each signal segment and the signal change features of adjacent signal segments in the first preset time length time window and the second preset time length time window are taken as feature vectors and input into a preset recognition classification model to identify whether each line has a ground fault and the type of the ground fault, and the preset recognition classification model is a trained model obtained through a model training process based on the zero sequence voltage signals and the zero sequence current signals of historical low resistance ground faults and high resistance ground faults as training data and through signal segment recognition and signal segment feature input models.

8. The method according to claim 7, wherein the arc-extinguishing coil parallel- connectable small resistance control is a method for handling ground fault, characterized in that, The identification of the transient change signal and the smooth signal includes: For the signals in the preset time length time window, the envelope lines of the signals are obtained through Hilbert transform, and time-frequency data are obtained through wavelet transform; The transient change signal and the smooth signal are identified based on the envelope line smoothness and the frequency component distribution at different times in the time-frequency data.

9. The method of claim 7, wherein the method is characterized by, The signal feature acquisition method of the post-fault transient change signal time period and the transient change signal segment after the small resistance is input includes: Based on the transient current signal, a plurality of sub-signals of different frequencies are obtained through wavelet packet decomposition; The horizontal features of the sub-signals of a single frequency are analyzed horizontally, and the vertical features of the sub-signals of a plurality of frequencies at the same time are analyzed vertically; The signal features are generated based on the horizontal features and the vertical features.

10. A system for handling ground faults based on an arc suppression coil in parallel with a controllable small resistance, characterized in that, It includes: A switch preliminary control unit for setting the arc-extinguishing coil in an automatic compensation working mode and the small resistance not being input; A signal acquisition unit for monitoring the neutral point zero sequence voltage and the zero sequence current signals of each line in real time; A preliminary analysis unit for determining that a ground fault occurs when the neutral point zero sequence voltage exceeds a preset upper limit value of the normal neutral point zero sequence voltage; A switch secondary control unit for inputting the small resistance when the time of the ground fault continuously exceeds a first time length or intermittently meets a first condition a plurality of times, and the first time length is determined based on the instantaneous ground fault self-healing time length range; A ground fault analysis unit for analyzing whether the ground fault is a low resistance ground fault or a high resistance ground fault and the line in which the low resistance ground fault or the high resistance ground fault occurs based on the neutral point zero sequence voltage and the zero sequence current signals of each line; A fault processing unit for cutting off the fault line by using a line protection device.

Images