A method and system for diagnosing ground fault of an active power distribution network

Abstract

The application discloses a kind of active distribution network grounding fault diagnosis method and system, it is related to power grid fault diagnosis technical field, the method includes steps: obtaining the voltage monitoring data and current monitoring data of active distribution network;Based on the preset zero sequence voltage fluctuation reference, dynamic deviation evaluation is carried out to voltage monitoring data, and suspected ground fault period is screened out;In suspected ground fault period, based on the polarity reversal monitoring point of current monitoring data in distribution network topological relation, fault point is positioned from each monitoring point, and according to the transient zero sequence current attenuation gradient and time sequence position stability of fault point along feeder to two ends, the possible degree value caused by real ground fault in current fault period is calculated;When the possible degree value reaches preset condition, real fault point is screened out from each monitoring point in current fault period, and state early warning information is determined.The application achieves the technical effect of avoiding early warning misjudgment or misjudgment.

Description

Technical Field

[0001] This application relates to the field of power grid fault diagnosis technology, specifically to a grounding fault diagnosis method and system for an active distribution network. Background Technology

[0002] With the widespread integration of distributed power sources, the operating conditions of active distribution networks are complex and variable. Ground fault diagnosis in active distribution networks is a critical link in ensuring power supply reliability and equipment safety. Existing ground fault diagnosis methods mostly rely on fixed thresholds or steady-state components of zero-sequence voltage and current for judgment. When the power grid experiences disturbances such as sudden changes in the output of distributed power sources or large load switching, transient responses ("fault-like" fluctuations) that are highly similar to those of a real single-phase ground fault are generated. Existing methods are unable to effectively distinguish between normal disturbances and real ground faults under the above coupled conditions, which can easily lead to missed or false alarms. Summary of the Invention

[0003] To address the technical problem that many related technologies rely on fixed thresholds or steady-state components of zero-sequence voltage and current for judgment, making it difficult to effectively distinguish between normal disturbances and actual grounding faults under coupled operating conditions, and easily leading to missed or false early warnings, this application provides a grounding fault diagnosis method and system for active distribution networks.

[0004] The specific technical solution adopted is as follows: Acquire voltage and current monitoring data from multiple monitoring points in an active power distribution network; Based on a preset zero-sequence voltage fluctuation benchmark, dynamic deviation assessment is performed on voltage monitoring data to screen out suspected grounding fault periods; During the suspected ground fault period, based on the polarity reversal monitoring points in the distribution network topology of the current monitoring data, the fault point is located from each monitoring point, and the probability value of the current fault period being caused by a real ground fault is calculated according to the transient zero-sequence current decay gradient along the feeder to both ends of the fault point and the temporal position stability of the fault point. When the probability value reaches the preset condition, the actual fault point is selected from each monitoring point in the current fault period, and the status warning information corresponding to the actual fault point is determined.

[0005] In one possible implementation of this application, based on a preset zero-sequence voltage fluctuation benchmark, dynamic deviation assessment is performed on voltage monitoring data to screen out suspected ground fault periods, including: Obtain the time-series curves of voltage monitoring data at each monitoring point within a preset historical reference period, and divide the time-series curves into multiple sliding time windows; For any sliding time window, based on the standard deviation of the voltage monitoring data within the sliding time window and the voltage monitoring data, calculate the first probability value of each moment within the current sliding time window belonging to a suspected ground fault moment; Based on the first probability value, suspected grounding fault periods are selected from the sliding time window.

[0006] In one possible implementation of this application, based on a first probability value, the suspected ground fault period is filtered out from the sliding time window, including: The first probability value is normalized to obtain the normalized probability value; The moment when the normalized probability value is greater than the first preset threshold is marked as a suspected grounding fault moment; The set of suspected ground fault moments that are sequentially linked is taken as the suspected ground fault period.

[0007] In one possible implementation of this application, locating fault points from various monitoring points based on current monitoring data at polarity reversal monitoring points in the distribution network topology includes: For any point within the suspected grounding fault period, sequentially traverse all monitoring points that are in an abnormal state and arranged according to the feeder direction at the current time, and mark the first monitoring point whose current change direction is inconsistent with the previous monitoring point as the fault point.

[0008] In one possible implementation of this application, based on the transient zero-sequence current decay gradient along the feeder from the fault point to both ends and the temporal location stability of the fault point, the probability value of the current fault period being caused by a real ground fault is calculated, including: Based on the transient zero-sequence current decay gradient from the fault point to both ends of the feeder, determine the proportion of upstream monitoring points and the proportion of downstream monitoring points whose zero-sequence current decay gradients from the fault point to both ends of the feeder are in the same direction. The current mutation difference is calculated based on the difference between the zero-sequence current mutation amplitudes of adjacent monitoring points. For any given moment, based on the current change difference, the first probability value, the proportion of upstream monitoring points, and the proportion of downstream monitoring points, a second probability value is calculated to determine if the current moment matches the characteristics of a real ground fault. Based on the temporal location stability of the fault point during the suspected ground fault period and the fluctuation of the second probability value, the probability value that the current fault period is caused by a real ground fault is calculated.

[0009] In one possible implementation of this application, based on the temporal location stability of the fault point during the suspected ground fault period and the fluctuation degree of the second probability value, the probability value that the current fault period is caused by a real ground fault is calculated, including: If the elements in the upstream monitoring point sequence corresponding to any two times are the same, then the two times are merged into a sub-time period, and the first proportion of the sub-time period in the suspected ground fault time period is determined. Obtain the number of times the location of the fault point changes during the suspected ground fault period; The second probability value at each moment during the suspected ground fault period is subjected to first-order difference and summation to obtain the first-order difference sum, which is used to characterize the fluctuation of the second probability value. Based on the number of location changes, the first proportion, and the sum of the first-order differences, the probability that the current fault period is caused by a real ground fault is calculated.

[0010] In one possible implementation of this application, when the probability value reaches a preset condition, the actual fault point is selected from each monitoring point in the current fault period, including: The probability values ​​are normalized to obtain normalized probability values; If the normalized probability value is greater than the second preset threshold, then the current suspected ground fault period is taken as the period corresponding to the actual fault point, and the monitoring point marked as the fault point most frequently is extracted from the current suspected ground fault period as the actual fault point.

[0011] In one possible implementation of this application, determining the status warning information corresponding to the actual fault point includes: Based on the temporal similarity of transient pulses of the actual fault point in temporally adjacent suspected ground fault periods, multiple recurrence periods corresponding to the suspected ground fault periods are determined. Based on the probability value of each reproduction period and the degree of zero-sequence voltage recovery, the warning priority of the actual fault point is determined, and the status warning information corresponding to the warning priority is output.

[0012] In one possible implementation of this application, the warning priority of the actual fault point is determined based on the probability value of each reproduction period and the degree of zero-sequence voltage recovery, including: The degree of zero-sequence voltage recovery is determined based on the degree of change of the first probability value during the reproduction period; The warning priority for the actual fault point is determined based on the probability value of each recurrence period, the frequency of occurrence of the recurrence period, and the duration of the recurrence period.

[0013] To achieve the above objectives, a ground fault diagnosis system for an active power distribution network is also provided. The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the above-mentioned ground fault diagnosis methods for an active power distribution network.

[0014] This application has, but is not limited to, the following technical effects: By acquiring voltage and current monitoring data from multiple monitoring points in the active power distribution network, and performing dynamic deviation assessment on the voltage monitoring data based on a preset zero-sequence voltage fluctuation benchmark, suspected ground fault periods are screened out. Based on the polarity reversal monitoring points within the suspected ground fault periods, the fault point is located from each monitoring point. Furthermore, based on the transient zero-sequence current attenuation gradient along the feeder from the fault point to both ends and the temporal position stability of the fault point, the probability value of the current fault period being caused by a real ground fault is determined. Then, when the probability value reaches a preset condition, the real fault point is screened from each monitoring point in the current fault period, and the corresponding status warning information is determined. In this application, by using the physical characteristics of "polarity reversal and bidirectional attenuation stability" of the real fault, the boundary is successfully distinguished from the unidirectional and unstable attenuation disturbance caused by "large load switching / sudden change in power output". The probability value of the current fault period being caused by a real ground fault is calculated. By comparing the probability value with the preset condition, the transient response corresponding to the real fault and normal disturbance is distinguished, thereby avoiding the occurrence of missed or misjudged warnings. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the first embodiment of the grounding fault diagnosis method for active power distribution networks according to this application. Figure 2 This is a schematic diagram of the overall implementation process of the active power distribution network grounding fault diagnosis method of this application. Detailed Implementation

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

[0017] This application provides a ground fault diagnosis method for an active power distribution network. In the first embodiment of the ground fault diagnosis method for an active power distribution network, refer to... Figure 1 The overall implementation flowchart of this application is as follows: Figure 2 As shown, the method includes: Step S10: Obtain voltage monitoring data and current monitoring data from multiple monitoring points in the active power distribution network.

[0018] As an example, the ground fault diagnosis method for active distribution networks can also be applied to the ground fault diagnosis system of active distribution networks. By using monitoring terminals deployed along the distribution network, the monitoring points in the active distribution network are monitored synchronously, and voltage monitoring data and current monitoring data of multiple monitoring points are collected. The voltage monitoring data includes three-phase voltage and zero-sequence voltage, and the current monitoring data includes three-phase current and zero-sequence current. After the data is collected, it is recorded synchronously according to a unified timestamp.

[0019] Furthermore, the switching times of the distribution network's operating conditions are recorded synchronously, including the times of distributed power generation grid connection and disconnection, capacitor bank switching, network reconfiguration operations, and large-capacity load start-up and shutdown, aligned with the monitoring data by timestamp. Based on the actual route and branch structure of the distribution network, monitoring points are deployed at the start and end points of each line section, forming a monitoring sequence along the line. The upstream and downstream connections between each monitoring point, the line section numbers and spatial distances between adjacent monitoring points are recorded, establishing a spatial topology table of monitoring points.

[0020] As an example, the collected multi-source data are uniformly aligned to the same time coordinate system, and missing values ​​are interpolated linearly. Zero-sequence voltage and zero-sequence current are segmented and associated according to monitoring point number and time sequence, which serve as the basic input data for subsequent analysis.

[0021] It should be noted that in active distribution networks, the monitoring data of zero-sequence voltage anomalies caused by single-phase grounding faults are similar to those of "fault-like" fluctuations in normal processes. For example, fault-like disturbances such as sudden changes in distributed power output and large load switching are highly similar in signal performance. Furthermore, intermittent arc grounding can repeatedly generate transient pulses, making it easy for the actual fault risk to be masked by short-term disturbances. This application distinguishes between actual grounding faults and operating condition fluctuations based on the stable distribution of the zero-sequence current of an actual grounding fault along the abnormal change direction of the feeder during abnormal periods, and the synchronous abnormal response of upstream monitoring points. This identifies the actual fault point and determines the early warning priority of actual grounding faults in active distribution networks by observing the abnormal evolution of multiple transient pulses at the same fault point.

[0022] Step S20: Based on the preset zero-sequence voltage fluctuation benchmark, perform dynamic deviation assessment on the voltage monitoring data to screen out suspected grounding fault periods.

[0023] As an example, when a ground fault occurs, an abnormal conductive path is formed between the fault point and the ground, causing the three-phase balance of the system to be disrupted and the zero-sequence voltage to rise abnormally. Therefore, the time periods suspected of ground faults are obtained by screening the times of abnormal zero-sequence voltage at various monitoring points in the active distribution network over a historical period.

[0024] Among them, the preset zero-sequence voltage fluctuation benchmark is a benchmark condition determined based on historical approximate zero-sequence voltage fluctuation under distributed power source conditions. By comparing with this benchmark, the time period suspected of grounding fault is screened out, that is, the suspected grounding fault time period is determined.

[0025] Step S20 further includes steps S21 to S23: Step S21: Obtain the time-series curve of voltage monitoring data of each monitoring point within a preset historical reference period, and divide the time-series curve into multiple sliding time windows.

[0026] As an example, the load and output of distributed generation in active distribution networks often exhibit seasonal differences. Therefore, short-term monitoring data from a single monitoring point (in this embodiment, historical data from the past month is used as short-term monitoring data) are analyzed as historical reference data. Specifically, since the load varies greatly between working days and non-working days, the monitoring cycle is coded and reference data under similar operating conditions is selected: one week is a reference cycle and one day is a monitoring cycle, and the monitoring cycles within each reference cycle are numbered according to the same time sequence.

[0027] The historical reference period for the period to be measured is the monitoring period with the same number.

[0028] The time-series curves of zero-sequence voltage measured at a single monitoring point over all historical reference periods are obtained, and the time-series curves are segmented at multiple extreme points. The average duration of the segmented time periods is used as the sliding time window size for the current monitoring period.

[0029] Step S22: For any sliding time window, based on the standard deviation of the voltage monitoring data within the sliding time window and the voltage monitoring data, calculate the first probability value of each moment within the current sliding time window belonging to a suspected ground fault moment.

[0030] As an example, the zero-sequence voltage fluctuation state varies under different operating conditions in an active distribution network. If, within a certain sliding time window (e.g., m), the fluctuation of the zero-sequence voltage is greater than that of the same period in the historical reference period, and at a certain moment (e.g., time t), the deviation from the overall zero-sequence voltage state of that period is significant, then the probability that time t is a suspected ground fault moment is higher. Based on this, the first probability value of time t within the current sliding time window m belonging to a suspected ground fault moment is calculated. : In the formula, Let m be the standard deviation of the zero-sequence voltage over the sliding time window. This represents the mean standard deviation of the zero-sequence voltage for the same time period across all sliding time windows (m) within the historical reference period. Let be the value of the zero-sequence voltage measured at the current monitoring point at time t. Let be the mean of the zero-sequence voltage within the sliding time window m. For example, a very small positive number. This is used to prevent the denominator from being zero.

[0031] Step S23, based on the first probability value, filters out suspected grounding fault time periods from the sliding time window, specifically including: The first probability value is normalized to obtain the normalized probability value.

[0032] The moment when the normalized probability value is greater than the first preset threshold is marked as a suspected grounding fault moment.

[0033] The set of suspected ground fault moments that are sequentially linked is taken as the suspected ground fault period.

[0034] As an example, the max-min normalization method is used to... Normalized to the (0,1) interval, the maximum and minimum values ​​are extracted using the first probability value of each sliding time window. The first preset threshold can be 0.5, and its value is determined based on the median decision method (or equal probability segmentation).

[0035] As an example, after normalizing the first probability value, if the normalized probability value at time t within a certain sliding window is greater than 0.5, then that time is marked as a suspected ground fault time. The set of suspected ground fault times that are sequentially connected is denoted as a suspected ground fault time period, and the corresponding monitoring point is marked as being in an abnormal state during that time period.

[0036] Step S30: During the suspected ground fault period, based on the polarity reversal monitoring points in the distribution network topology using current monitoring data, locate the fault point from each monitoring point, and calculate the probability that the current fault period is caused by a real ground fault based on the transient zero-sequence current decay gradient along the feeder to both ends of the fault point and the temporal position stability of the fault point.

[0037] As an example, a suspected ground fault period could be caused by other fluctuations or by a real fault. The probability value indicates the likelihood that the current fault period is caused by a real ground fault; the higher the probability value, the greater the probability that the current fault period is caused by a real ground fault. The fault point is the initial fault point determined based on the direction of the current change.

[0038] It should be noted that active distribution networks are usually subject to normal operating condition disturbances, such as sudden changes in the output of distributed power sources, large load switching, or capacitor switching, which may trigger transient responses in the zero-sequence loop, making its zero-sequence voltage fluctuation characteristics similar to those of ground faults.

[0039] Therefore, it is necessary to distinguish between transient grounding faults and operational disturbances in real active distribution networks by combining the actual transient evolution direction and distribution range stability of the grounding fault. Specifically, the transient zero-sequence current mutation amplitude of a real fault usually decays from the fault point to both ends of the feeder, with a stable bidirectional decay gradient, and the location of the fault point is relatively stable within the abnormal section; while the disturbance source of a fault-like disturbance is usually at the bus or the point of access of distributed power sources, the amplitude decay is unstable, and the fault point may usually drift with the operational conditions. This is used to distinguish whether the current fault period is caused by a real grounding fault or by normal fluctuations.

[0040] Among them, step S30, which involves locating the fault point from each monitoring point based on the polarity reversal monitoring point in the distribution network topology using current monitoring data, includes: For any point within the suspected grounding fault period, sequentially traverse all monitoring points that are in an abnormal state and arranged according to the feeder direction at the current time, and mark the first monitoring point whose current change direction is inconsistent with the previous monitoring point as the fault point.

[0041] As an example, taking a specific moment within a single suspected ground fault period (e.g., j), for instance, at time t, all monitoring points in abnormal states are arranged along the feeder direction. Each monitoring point is traversed sequentially, and the first monitoring point whose current abrupt change direction differs from the preceding monitoring point is marked as the fault point. In practice, the branch characteristics of the distribution network are considered as supplementary criteria: if, after traversing all marked abnormal monitoring points within the current line, the abrupt change direction of the transient zero-sequence current at each point remains consistent without any turning point, then the abnormal fluctuation is determined to converge at the end. In this case, the last monitoring point in the entire monitoring sequence is marked as the fault point.

[0042] The step S30, which calculates the probability that the current fault period is caused by a real ground fault based on the transient zero-sequence current attenuation gradient along the feeder from the fault point to both ends and the temporal location stability of the fault point, includes: Based on the transient zero-sequence current decay gradient from the fault point to both ends of the feeder, determine the proportion of upstream monitoring points and the proportion of downstream monitoring points whose zero-sequence current decay gradients from the fault point to both ends of the feeder are in the same direction.

[0043] As an example, the preceding monitoring point of the fault point is marked as the upstream monitoring point, and all monitoring points along the feeder towards the load end after the fault point are uniformly recorded as downstream monitoring points, thus obtaining the upstream monitoring point sequence, denoted as... .

[0044] As an example, let's denote the percentage of the total number of upstream monitoring points where the maximum number of consecutively occurring zero-sequence current abrupt amplitude decay gradients are aligned with the direction of the gradients during the upstream traversal as the upstream monitoring point percentage. Similarly, during downstream traversal, the percentage of the largest number of monitoring points with consecutive occurrences and consistent decay gradient directions of zero-sequence current abrupt amplitudes is obtained out of all downstream monitoring points and denoted as the percentage of downstream monitoring points. If the currently marked fault point is located at the end of the overall monitoring sequence, since there is no physical set of downstream monitoring points, this condition will not perform a division comparison operation on the total number of downstream points, and the constant of the proportion of downstream monitoring points in this case will be directly assigned to 0.

[0045] The current mutation difference is calculated based on the difference between the zero-sequence current mutation amplitudes of adjacent monitoring points.

[0046] For any given moment, based on the current change difference, the first probability value, the proportion of upstream monitoring points, and the proportion of downstream monitoring points, a second probability value is calculated to determine if the current moment matches the characteristics of a real ground fault.

[0047] As an example, the difference between the zero-sequence current mutation amplitudes of two adjacent monitoring points is calculated along the traversal order, and this difference is denoted as the current mutation difference d. Before calculating the second probability value, d is detected. If the zero-sequence current mutation amplitudes of multiple adjacent monitoring points are zero, then the zero-sequence current mutation amplitude does not decay, that is, there is no fault, and the subsequent calculation of the second probability value is not performed.

[0048] As an example, if monitoring points are traversed from the fault point to both sides, and if traversing upstream or downstream, the attenuation gradient direction of the transient zero-sequence current abrupt change amplitude at each monitoring point is relatively consistent, the attenuation amplitude is relatively gentle (the average value of the amplitude difference is small), and the first probability value of all monitoring points at that moment within their corresponding time window is... If both values ​​are relatively large, then the time at that moment matches the characteristics of a real grounding fault; otherwise, it is a normal fluctuation.

[0049] As an example, taking time t as an example, the second probability value that conforms to the characteristics of a real ground fault at time t is... for: In the formula, Let be the mean of all current abrupt changes during the upstream and downstream traversal at time t. For all monitoring points at time t, within their corresponding sliding time windows mean This indicates the percentage of upstream monitoring points. This indicates the percentage of downstream monitoring points.

[0050] Based on the temporal location stability of the fault point during the suspected ground fault period and the fluctuation of the second probability value, the probability value that the current fault period is caused by a real ground fault is calculated, specifically including: If the elements in the upstream monitoring point sequence corresponding to any two times are the same, then the two times are merged into a sub-time period, and the first proportion of the sub-time period in the suspected ground fault time period is determined.

[0051] Obtain the number of times the location of the fault point changes during the suspected grounding fault period.

[0052] As an example, based on the above analysis, the fault point of a real grounding fault is relatively stable, and the upstream monitoring point of the fault point usually responds stably to the fault abnormally (i.e., at each moment within that period). The similarity of the elements in the middle should be high), and the corresponding values ​​at each time step should be high. If the situation should evolve steadily, then the probability that the anomaly during this period is dominated by a real ground fault is relatively high. Based on this, we calculate the probability that the current fault period is caused by a real ground fault.

[0053] As an example, if the upstream monitoring point sequences at time t and t+1 are completely identical, they are merged into a sub-time period (i.e., the two times have the same upstream monitoring point sequence), and the proportion of the sub-time period's duration in the total duration of the suspected ground fault period is taken as the first proportion.

[0054] The number of location changes can be obtained as follows: within the current suspected ground fault period j, extract the fault point location identifiers obtained from each sampling time according to the time sequence, and compare the fault point location identifiers of two adjacent sampling times (t and t+1). If the fault point location identifiers of the two consecutive times are different, it is recorded as a change. The total number of such location changes between adjacent times within the entire period j is the number of location changes.

[0055] The second probability value at each moment during the suspected ground fault period is subjected to first-order difference and summation to obtain the first-order difference sum, which is used to characterize the fluctuation of the second probability value.

[0056] Based on the number of location changes, the first proportion, and the sum of the first-order differences, the probability that the current fault period is caused by a real ground fault is calculated.

[0057] As an example, the first-order difference summation is the second probability value during the suspected ground fault period. The sum of the absolute values ​​of all first-order differences in the corresponding time series is used to represent the degree of fluctuation of the second probability value.

[0058] As an example, taking the j-th suspected ground fault period as an example, the probability value... The calculation method is as follows: In the formula, The first percentage corresponding to each sub-period within the current time period. The maximum value indicates less disturbance, representing the percentage of time during which the upstream response is most stable. This represents the number of times the fault point changes between adjacent times within the time period j, i.e., the number of location changes. The smaller the value, the more stable the location. During this period The sum of the absolute values ​​of all first-order differences of a time series is the first-order difference sum. Here, the subscript t indicates that the parameter is calculated based on a time series in the time domain, the subscript d indicates that the parameter has undergone difference operations, and the subscript s indicates that the parameter has finally undergone a summation operation. The balance coefficient is typically set to 1. Real-time monitoring data of the distribution network inevitably contains background noise, measurement errors, and minor fluctuations in the power grid. There will inevitably be slight differences at different times. It cannot be absolutely zero.

[0059] Step S40: When the probability value reaches the preset condition, the actual fault point is selected from each monitoring point in the current fault period, and the status warning information corresponding to the actual fault point is determined.

[0060] As an example, the preset condition can be that the normalized probability value is greater than the second preset threshold. First, the time period corresponding to the real fault point is filtered out, then the real fault point is determined from the time period, and then the status warning information corresponding to the real fault point is output.

[0061] Step S40 includes: The probability values ​​are normalized to obtain normalized probability values.

[0062] If the normalized probability value is greater than the second preset threshold, then the current suspected ground fault period is taken as the period corresponding to the actual fault point, and the monitoring point marked as the fault point most frequently is extracted from the current suspected ground fault period as the actual fault point.

[0063] As an example, one way to normalize the probability values ​​is through max-min normalization, which normalizes the probability values ​​to the (0,1) interval. The maximum and minimum values ​​can be extracted from multiple probability values ​​within a historical time period. The normalized probability values ​​are denoted as... The second preset threshold can be 0.5, 0.6, or other values, and can be adjusted according to the specific situation.

[0064] As an example, A time period with a value greater than 0.5 is considered the time period corresponding to the actual fault point. Specifically, 0.5 is the mathematical median of the normalized interval from 0 to 1, serving as the benchmark threshold for determining the characteristic tendency. A calculation result greater than 0.5 indicates that the characteristic behavior of the current time period tends to be a true ground fault. A calculation result less than 0.5 indicates that the characteristic behavior tends to be a normal operating condition disturbance. This threshold serves as an initial benchmark reference for initially balancing the probability of missed detection and the probability of false detection. In specific implementation scenarios, this threshold can be calibrated and adjusted based on the actual false alarm rate and characteristic background noise in the historical operating data of the active distribution network. If the on-site safety requirements are extremely high, the threshold can be lowered to 0.4; if there is excessive on-site interference requiring frequent voltage drop alarms, the threshold can be raised to 0.6.

[0065] As an example, a suspected ground fault period includes multiple sampling times. The monitoring point that is marked as a fault point the most times during that period is taken as the actual fault point for the suspected ground fault period.

[0066] The step of determining the status warning information corresponding to the actual fault point includes steps S41 to S42: Step S41: Based on the temporal similarity of transient pulses of the actual fault point in the temporally adjacent suspected ground fault periods, determine multiple recurrence periods corresponding to the suspected ground fault periods.

[0067] As an example, in active distribution network grounding faults, there may be multiple occurrences of the same grounding fault. For instance, repeated arc reignition may cause multiple transient pulses to be generated at the same fault point. Therefore, it is necessary to analyze the risk index of the current active distribution network to normal operation by combining the evolution trend of the multiple transient pulses.

[0068] Among them, if there are many abnormal transient pulses at the same fault point, the duration is long, and each transient pulse does not have the characteristic of self-extinguishing with zero-sequence voltage drop, then the early warning priority of the current active distribution network grounding fault is higher.

[0069] As an example, consider two time periods that are adjacent to each other and suspected of being ground faults (e.g., time periods j and j). If the fault points in the two time periods are the same, the direction of the current change is the same, and they closely approximate a real ground fault anomaly, with a short interval between them, then the two suspected ground fault periods are likely to be approximate transient pulses. Based on this, calculate j and j for the two time periods. Similarity values ​​between : In the formula, For the two time periods j and above The mean of the normalized probability values, For two time periods j and The difference between the normalized possible values, The interval between the two time periods mentioned above is defined as j and j', which are two independent time intervals determined to be abnormal. Due to intermittent arc faults or the reclosing action of the protection device, there will be a recovery interval between abnormal pulses. Greater than zero, The preset maximum repeating pulse time interval threshold can be set from 1 second to 3 seconds, and the specific value can be set according to the setting parameters of the target power grid protection device. For example, a very small positive number. This is used to prevent the denominator from being zero.

[0070] The label value indicates that if the fault point is the same in two time periods (the monitoring point marked as the fault point most times in each time period is taken as the fault point in that time period), and the direction of the current change is the same (the direction of the current change is taken as the direction of the current change in each time period), then the fault point is the same in two time periods. The direction of the current change in the upstream monitoring point sequence corresponding to the maximum value is the direction of the current change in that time period. It is 1 if it is true, otherwise it is 0.

[0071] Based on the maximum-minimum normalization method Normalize to (0,1). If the normalization result is greater than 0.5, then the time period is... The recurring time period marked as time period j is analyzed again using the same steps until no new recurring time period is found. Before normalization, if the absolute value of the difference between the probability values ​​of two time periods approaches 0 (e.g., less than the set tolerance threshold), the feature is directly determined to be highly consistent, and the corresponding time period is marked as a recurring time period. This value is not substituted into the normalization process of the remaining array. For groups that do not trigger the above conditions, they are normalized to (0,1) based on the max-min normalization method. Here, 0.5 is an empirical decision boundary, and its value is based on the median decision method (or equal probability segmentation). After normalizing the continuous variable to the (0,1) interval, 0 represents absolutely impossible (completely unrelated), and 1 represents absolutely possible (completely recurring). 0.5 is exactly the midpoint of the interval, which means that when the similarity feature of the target time period crosses the "compromise point between dissimilarity and similarity", it tends to be judged as a related impulse. The threshold of 0.5 is the baseline reference value. In practical applications, it can be finely adjusted between 0.4 and 0.7 according to the sensitivity requirements of the distribution network.

[0072] Step S42: Based on the probability value of each reproduction period and the degree of zero-sequence voltage recovery, determine the warning priority of the actual fault point, and output the status warning information corresponding to the warning priority.

[0073] As an example, if the current time (e.g., t) is in an abnormal state, and the suspected ground fault period has many recurring periods in historical data, the duration of each recurring period is relatively long, and the recovery degree of zero-sequence voltage anomaly is relatively small (the probability recovery degree of each time in the period being a suspected ground fault time is relatively poor), and the probability that it is the dominant anomaly of a real ground fault is relatively high, then the ground fault warning priority of the current period is relatively high. Based on this, the warning priority of the real fault point is calculated.

[0074] Step S42 includes: The degree of zero-sequence voltage recovery is determined based on the degree of change in the first probability value during the reproduction period.

[0075] The warning priority for the actual fault point is determined based on the probability value of each recurrence period, the frequency of occurrence of the recurrence period, and the duration of the recurrence period.

[0076] As an example, let the degree of zero-sequence voltage recovery in a single recurrence period (e.g., recurrence period j) be denoted as: in, The first probability value within this time period The maximum value, The first probability value corresponding to the end time of this period. , for The time remaining until the end of the time period is calculated, wherein when the maximum value time is at the end of the time period, and the time remaining until the end of the time period is 0 or less than the set reference protection time, the time is forcibly limited to the reference protection time, and the calculation is performed accordingly.

[0077] As an example, for time t in any recurrence period, the warning priority is... The calculation formula can be: In the formula, Let be the number of recurrence intervals in the time interval containing time t. The probability value corresponding to each recurrence period The mean, The total duration of each recurrence period. For each recurrence period Mean.

[0078] Furthermore, based on the max-min normalization method, Normalize to (0,1), and denote the normalization result as... ,Will The data is transmitted to the output unit, which then outputs the corresponding warning information. when When the fault location is marked at that time, a red warning message is output, along with the corresponding upstream monitoring point sequence and the location of the monitoring point in an abnormal state during the recurrence period.

[0079] when At that time, an orange alert will be issued, along with the corresponding sequence of upstream monitoring points.

[0080] when At that time, only the corresponding ground fault warning priority is recorded.

[0081] Among them, 0.7 is the high-risk threshold, indicating that transient pulses are frequently triggered at the same location and the zero-sequence voltage shows no recovery trend, corresponding to persistent grounding or high-frequency reignition faults, requiring an immediate red alert to guide isolation; 0.4 is the attention threshold, indicating the presence of intermittent grounding pulses that have not completely deteriorated, corresponding to unstable potential hazards, requiring an orange alert to prompt inspection; values ​​below 0.4 indicate transient contact disturbances that have already decayed on their own, and are only recorded in the logbook. Specific values ​​are determined by the implementation personnel based on the historical defect elimination records and fault tolerance rate of the target distribution network.

[0082] This application provides a ground fault diagnosis method for an active distribution network. It acquires voltage and current monitoring data from multiple monitoring points in the active distribution network, performs dynamic deviation assessment on the voltage monitoring data based on a preset zero-sequence voltage fluctuation benchmark, and filters out suspected ground fault periods. Based on the polarity reversal monitoring points within the suspected ground fault period, it locates the fault point from each monitoring point. Then, based on the transient zero-sequence current attenuation gradient along the feeder from the fault point to both ends and the temporal position stability of the fault point, it determines the probability that the current fault period is caused by a real ground fault. Furthermore, when the probability reaches a preset condition, it filters out the real fault point from each monitoring point in the current fault period and determines the corresponding state warning information. In this application, the physical characteristics of "polarity reversal and bidirectional attenuation stability" of a real fault are successfully distinguished from the unidirectional, unstable attenuation disturbance caused by "large load switching / sudden power output changes." The probability of the current fault period being caused by a real ground fault is calculated, and the probability is compared with preset conditions to differentiate the transient response corresponding to a real fault from that of a normal disturbance, thereby avoiding missed or misjudged warnings.

[0083] This application also provides a ground fault diagnosis system for an active power distribution network. The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the above-described methods for diagnosing ground faults in an active power distribution network.

[0084] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0085] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0086] The above are merely preferred embodiments of this application and do not limit the scope of this application. Any equivalent structural or procedural transformations made based on the description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the scope of protection of this application.

Claims

1. A method for diagnosing grounding faults in an active power distribution network, characterized in that, The method includes: Acquire voltage and current monitoring data from multiple monitoring points in an active power distribution network; Based on a preset zero-sequence voltage fluctuation benchmark, the voltage monitoring data is dynamically evaluated to identify suspected grounding fault periods. During the suspected ground fault period, based on the polarity reversal monitoring points in the distribution network topology of the current monitoring data, the fault point is located from each monitoring point, and the probability value of the current fault period being caused by a real ground fault is calculated according to the transient zero-sequence current decay gradient of the fault point along the feeder to both ends and the temporal position stability of the fault point. When the probability value reaches the preset condition, the actual fault point is selected from each monitoring point in the current fault period, and the status warning information corresponding to the actual fault point is determined.

2. The method for diagnosing grounding faults in an active power distribution network as described in claim 1, characterized in that, The process of dynamically evaluating the voltage monitoring data based on a preset zero-sequence voltage fluctuation benchmark to identify potential grounding fault periods includes: Obtain the time-series curves of voltage monitoring data at each monitoring point within a preset historical reference period, and divide the time-series curves into multiple sliding time windows; For any sliding time window, based on the standard deviation of the voltage monitoring data within the sliding time window and the voltage monitoring data, calculate the first probability value of each moment within the current sliding time window belonging to a suspected ground fault moment; Based on the first probability value, suspected grounding fault periods are selected from the sliding time window.

3. The method for diagnosing grounding faults in an active power distribution network as described in claim 2, characterized in that, The step of filtering out suspected grounding fault periods from the sliding time window based on the first probability value includes: The first probability value is normalized to obtain a normalized probability value; The moment when the normalized probability value is greater than the first preset threshold is marked as a suspected grounding fault moment; The set of suspected ground fault moments that are sequentially linked is taken as the suspected ground fault period.

4. The method for diagnosing grounding faults in an active power distribution network as described in claim 1, characterized in that, The polarity reversal monitoring point in the distribution network topology based on the current monitoring data, locating the fault point from each monitoring point, includes: For any moment within the suspected grounding fault period, sequentially traverse all monitoring points that are in an abnormal state and arranged according to the feeder direction at the current moment, and mark the first monitoring point whose current change direction is inconsistent with the previous monitoring point as the fault point.

5. The method for diagnosing grounding faults in an active power distribution network as described in claim 2, characterized in that, The step of calculating the probability that the current fault period is caused by a real ground fault based on the transient zero-sequence current decay gradient along the feeder from the fault point to both ends and the temporal position stability of the fault point includes: Based on the transient zero-sequence current decay gradient from the fault point to both ends of the feeder, determine the proportion of upstream monitoring points and the proportion of downstream monitoring points whose zero-sequence current decay gradients from the fault point to both ends of the feeder are in the same direction. The current mutation difference is calculated based on the difference between the zero-sequence current mutation amplitudes of adjacent monitoring points. For any given moment, based on the current change difference, the first probability value, the proportion of upstream monitoring points, and the proportion of downstream monitoring points, a second probability value is calculated to indicate that the current moment conforms to the characteristics of a real ground fault. Based on the temporal location stability of the fault point during the suspected ground fault period and the fluctuation of the second probability value, the probability value that the current fault period is caused by a real ground fault is calculated.

6. The method for diagnosing grounding faults in an active power distribution network as described in claim 5, characterized in that, The calculation of the probability that the current fault period is caused by a real ground fault, based on the temporal location stability of the fault point during the suspected ground fault period and the fluctuation of the second probability value, includes: If the elements in the upstream monitoring point sequence corresponding to any two times are the same, then the two times are merged into a sub-time period, and the first proportion of the sub-time period in the suspected ground fault time period is determined. Obtain the number of times the location of the fault point changes during the suspected ground fault period; The second probability value at each moment during the suspected ground fault period is subjected to first-order difference and summation to obtain the first-order difference sum, which is used to characterize the fluctuation of the second probability value. Based on the number of location changes, the first proportion, and the sum of the first-order differences, the probability value that the current fault period is caused by a real ground fault is calculated.

7. The method for diagnosing grounding faults in an active power distribution network as described in claim 1, characterized in that, When the probability value reaches a preset condition, the actual fault point is selected from each monitoring point in the current fault period, including: The possible probability values ​​are normalized to obtain normalized possible probability values; If the normalized probability value is greater than the second preset threshold, then the current suspected ground fault period is taken as the period corresponding to the actual fault point, and the monitoring point marked as the fault point most frequently is extracted from the current suspected ground fault period as the actual fault point.

8. The method for diagnosing grounding faults in an active distribution network as described in claim 1, characterized in that, The status warning information corresponding to the actual fault point includes: Based on the temporal similarity of transient pulses of the actual fault points in temporally adjacent suspected ground fault periods, multiple recurrence periods corresponding to the suspected ground fault periods are determined; Based on the probability value of each reproduction period and the degree of zero-sequence voltage recovery, the warning priority of the actual fault point is determined, and the status warning information corresponding to the warning priority is output.

9. The method for diagnosing grounding faults in an active power distribution network as described in claim 8, characterized in that, The step of determining the early warning priority of the actual fault point based on the probability value of each reproduction period and the degree of zero-sequence voltage recovery includes: The degree of zero-sequence voltage recovery is determined based on the degree of change of the first probability value during the reproduction period; The warning priority of the actual fault point is determined based on the probability value of each recurrence period, the number of times the recurrence period occurs, and the duration of the recurrence period.

10. A grounding fault diagnosis system for an active power distribution network, characterized in that, The system includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the method as described in any one of claims 1 to 9.

Images