A single-phase ground fault section positioning method based on a master station centralized multi-element detection factor comprehensive analysis

Abstract

This invention discloses a method for locating single-phase grounding fault sections using a centralized multi-factor comprehensive analysis at a master station. The method includes: acquiring the topology, terminal register, and operational information of suspected grounding fault lines in the distribution network; constructing alarm signal observers and characteristic quantity observers corresponding to terminal alarm signals, zero-sequence characteristic quantity telemetry, and virtual measurements extracted from fault recording files, respectively; dynamically constructing a fault detection factor set within the fault observers based on site conditions and the grounding fault scenario; using an enhanced multi-factor Bayesian algorithm to calculate the posterior probability of events for the two observers for each line and section; and employing the evidence consistency assumption to correct evidence conflicts caused by missed and false alarms from intelligent terminals; finally, selecting the line, section, and locating the fault section based on the posterior probability of events from the two observers. This invention is flexibly applicable to different automatic distribution terminal configuration schemes in the field and can achieve highly reliable single-phase grounding fault section location.

Description

Technical Field

[0001] This invention relates to the field of distribution network grounding fault assessment and auxiliary decision-making technology, specifically a method for locating single-phase grounding fault sections by comprehensive assessment of multiple detection factors at a central station. Background Technology

[0002] In distribution network faults, single-phase grounding faults account for a high proportion and are difficult to locate. Currently, grounding fault assessment based on distribution automation and digital platforms, using multi-source information integration, has become a hot topic in fault assessment. The advantage of master station-based detection methods lies in acquiring panoramic information and understanding the true upstream and downstream relationships of feeders; however, traditional methods using fixed detection factors suffer from insufficient flexibility and reliability. Furthermore, the limited reliability of existing terminal fault feature detection algorithms and the influence of different neutral grounding methods further reduce the reliability of location. Moreover, the imperfect power flow adaptive function of intelligent terminals leads to missed and false alarms in grounding fault alarms. Existing technologies are difficult to adapt to differentiated terminal configuration scenarios and have not effectively resolved the problem of conflicting evidence.

[0003] Due to differences in the configuration types and coverage of field distribution automation terminals, the traditional master station often fails when using fixed detection factors for horizontal and vertical comparisons because of the asymmetry (absence) of field detection factors. Furthermore, the strategy of using manual assistance for comprehensive judgment of multiple fault characteristics is still difficult to meet the needs of automated fault judgment of distribution cloud master stations. Summary of the Invention

[0004] To overcome the shortcomings of the existing technology, this invention provides a method for locating single-phase grounding fault sections by comprehensively judging multiple detection factors in a centralized main station. It achieves highly reliable single-phase grounding fault section location by enhancing Bayesian posterior probability, evidence consistency correction mechanism, and integrating multiple types of fault feature detection technology.

[0005] Therefore, the present invention adopts the following technical solution: a method for locating single-phase grounding fault sections based on centralized multi-factor comprehensive judgment at a master station, comprising the following steps: S100 acquires single-phase grounding alarm information from distribution terminals and low-current grounding alarm information from substations, uses them as the start signal for fault analysis, and delineates the suspected grounding fault range of the distribution network. S200 acquires the topology of suspected ground fault lines, upstream and downstream terminal connection relationships, terminal ledgers, terminal operation and alarm information, completes virtual measurement calculations for fault recording file recall and feature extraction, and identifies the neutral point grounding mode of the distribution network. S300 constructs alarm signal observers and feature quantity observers according to the division of lines and line sections. Based on the line selection and section selection service scenarios, it dynamically constructs a fault detection factor set and completes the fault detection factor calculation by combining multiple fault detection factors. S400 uses an enhanced multi-factor Bayesian algorithm to calculate the posterior probability of events for two observers in each line and each section. S500: When there is a conflict between the evidence from the alarm signal observer and the feature quantity observer, the evidence consistency assumption is used to correct the signal indication with the highest confidence weight in the alarm signal observer, and to eliminate false alarms and missed alarms at the terminal. S600 selects lines based on the posterior probability comparison strategy of the feature quantity observers; it selects segments based on the arithmetic mean of the posterior probabilities of the two observers, eliminating non-faulty line segments; and finally, based on the remaining chain topology, it verifies the arithmetic mean of the posterior probabilities of the upstream and downstream observers to complete the fault segment location.

[0006] Further, step S300 includes: Step S310: Configure alarm signal observers and characteristic quantity observers for the feeder head, segment and branch nodes according to the feeder segment position.

[0007] Furthermore, step S300 also includes: Step S320: The telemetry data, teleindication data, and virtual measurements obtained from the analysis of fault recording files of the power distribution cloud master station are modeled as fault detection factors, and the fault detection factors are dynamically configured into two observers according to the jurisdiction of the section.

[0008] Furthermore, step S300 also includes: Step S330: Calculate the output result of each fault detection factor (where "1" indicates that a grounding event was detected and "0" indicates that no grounding event was detected).

[0009] Further, step S400 includes: Step S410: In the alarm signal observer and the feature quantity observer, it is assumed that a fault detection factor set with N fault detection factors is configured. ,in ; For each fault detection factor Preset performance parameters, including hit rate underreporting rate False alarm rate and correct rejection rate ; Define the confidence weight of the detection factor ,in This indicates that the fault detection factor is effective.

[0010] Furthermore, step S400 also includes: Step S420: Based on the independence of fault detection factors, calculate the weighted likelihood of event occurrence and the weighted likelihood of event non-occurrence. The likelihood is the weighted product of the output probabilities of individual fault detection factors within the observer. Logarithmic transformation is used to optimize the likelihood value calculation. Step S430: Calculate the weighted evidence factor based on the weighted likelihood of the event occurring and the weighted likelihood of the event not occurring, and dynamically update the prior probability of the event occurring according to the alarm signal indication. Step S440: Based on the weighted likelihood, weighted evidence factor, and dynamically updated prior probability, the enhanced multi-factor Bayesian algorithm is used to calculate the posterior probability of events for the two observers of each line and each section.

[0011] Further, step S500 includes: S510, Missed Report Correction: Traverse the observer set of each segment along the feeder, if... If the value is greater than the first preset threshold, it is determined that the alarm signal observer has missed an alarm. The alarm signal with the highest confidence weight in the alarm signal observer is corrected to "1", and steps S430-S440 are executed again. in, For characteristic quantity observer The posterior probability of the event, Alarm signal observer The posterior probability of the event.

[0012] Furthermore, step S500 also includes: Step S520, False Alarm Correction: Traverse the observer set of each segment along the feeder, if If the current centroid frequency is greater than the first preset threshold and the zero mode is traced from downstream to upstream according to the upstream and downstream relationship of the feeder section, then it is determined that the alarm signal observer has a false alarm. The alarm signal with the highest confidence weight in the alarm signal observer is corrected to "0", and steps S430-S440 are re-executed.

[0013] Furthermore, step S500 also includes: Step S530, Fault Channel Correction: According to the upstream and downstream relationship of the feeder section, trace from downstream to upstream. If the value of the posterior probability of the downstream alarm signal observer event minus the posterior probability of the upstream alarm signal observer event is greater than the first preset threshold, then based on the ground fault characteristic continuity assumption, it is determined that the upstream alarm signal observer has missed alarm reporting. The alarm signal with the highest confidence weight in the upstream alarm signal observer is corrected to "1", and steps S430-S440 are re-executed.

[0014] Further, step S600 includes: Step S610, Execution of line selection strategy: In the distribution network suspected grounding fault range defined in step S100, when there are ≥2 suspected grounding faults on feeders, the event posterior probability comparison of the pairwise line characteristic quantity observers is adopted. If the value of the event posterior probability of the first line minus the event posterior probability of the second line is greater than the second preset threshold, then the fault probability of the first line is determined to be high. Step S620, execution of the segment selection strategy: Based on the upstream and downstream relationships of the suspected faulty line segment constructed in step S200, traverse the first-level branch nodes of the line. If the arithmetic mean of the event posterior probability of the alarm signal observer and the event posterior probability of the feature quantity observer is less than the third preset threshold, it is determined that the fault did not occur in that branch. Traverse the remaining feeder chain segments from downstream to upstream. If the arithmetic mean of the event posterior probability of the downstream segment is less than the third preset threshold, it is determined that the fault did not occur downstream of that segment. Step S630, final location of the faulty section: Traverse each segment of the remaining chain structure from downstream to upstream, and check whether each segment meets the following condition: the arithmetic mean of the posterior probability of the alarm signal observer event and the posterior probability of the feature quantity observer event is greater than the fourth preset threshold. After the check is passed, the last segment of the remaining suspected faulty line chain structure is determined to be the faulty segment.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention constructs a set of multi-factor fault detection factors that adapt to the topology and terminal configuration, accumulates the posterior probabilities of multiple detection factors, and combines alarm signal correction under key evidence conflict and other technical means to form a comprehensive probabilistic auxiliary positioning of dynamic multi-factor fault detection factors, thereby achieving the technical effect of adaptive grounding fault section positioning in the field. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of the single-phase grounding fault section location method of the present invention; Figure 2 This is a schematic diagram of the fault assessment start conditions and signal triggering mechanism of the present invention; Figure 3 This is a schematic diagram illustrating the scenario of dividing the suspected faulty line range according to the present invention; Figure 4 This is a schematic diagram showing the typical feeder terminal deployment and grounding point location in an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] Example 1 Currently, the selection of distribution intelligent terminals (FTU, DTU, transient waveform-based fault indicators, and transient characteristic-based fault indicators) varies significantly across China, and the coverage of these terminals is closely related to local needs and the level of technological and economic development. Furthermore, some provinces have implemented fault waveform file recall and calculated fault characteristics using transient signal methods (such as the first half-wave method, zero-mode polarity method, frequency domain centroid method, S-transform method, and wavelet transform method) for fault line selection and location. However, different types of terminals and different fault detection factors in the distribution network cannot achieve 100% accuracy in identifying single-phase grounding faults. Simultaneously, factors such as the neutral grounding type of the distribution system and the backflow of power flow significantly affect the accuracy of fault detection, especially for FTUs and DTUs with adaptive power flow, which exhibit certain false alarms and missed alarms.

[0020] Based on this, this embodiment proposes an enhanced Bayesian posterior probability and evidence consistency hypothesis based on the integration of various types of distribution network terminals and various fault detection factors. The proposed method is decoupled from the distribution automation level and fault detection factor selection in various regions, and can be adapted to the actual field conditions in various regions. It can integrate multiple types of feature detection technologies and has the advantages of strong adaptability and high reliability.

[0021] This embodiment provides a method for locating single-phase grounding fault sections based on a centralized multi-factor detection system at a master station. It aims to flexibly adapt to the configuration schemes of distribution automation terminals in various regions and the feature value calculation capabilities of fault recording files at provincial master stations. By integrating multiple feature detection technologies such as enhanced Bayesian comprehensive probability and evidence conflict correction, it achieves automated and highly reliable location of single-phase grounding fault sections in the distribution network.

[0022] A method for locating single-phase grounding fault sections based on comprehensive analysis of multiple fault detection factors from a distribution cloud master station, such as... Figure 1 As shown, the steps are as follows: S100 acquires single-phase grounding alarm information from distribution terminals and low-current grounding alarm information from substations, uses them as the start signal for fault analysis, and delineates the suspected grounding fault range of the distribution network. S200 acquires the topology of suspected ground fault lines, upstream and downstream terminal connection relationships (adjacent matrix), terminal ledger, terminal operation and alarm information, completes virtual measurement calculations for fault recording file recall and feature extraction, and identifies the neutral point grounding mode of the distribution network. S300, based on line and line section division, constructs an alarm signal observer. and characteristic quantity observer Based on the route selection and segment selection business scenarios, and combined with multiple fault detection factors, a fault detection factor set is dynamically constructed. And complete the calculation of fault detection factors; S400 uses an enhanced multi-factor Bayesian algorithm to calculate the posterior probability of events for two observers in each line and each section. S500, when alarm signal observer With characteristic quantity observer When there is conflict of evidence, the alarm signal observer is modified using the evidence consistency assumption. The signal with the highest confidence weight is used to eliminate false alarms and missed alarms from the terminal. S600, based on characteristic quantity observer Line selection is performed using a posterior probability comparison strategy; based on two observers (i.e., alarm signal observers)... and characteristic quantity observer The arithmetic mean of the posterior probabilities of the events is used to select segments and eliminate non-faulty line segments. Finally, based on the remaining chain topology, the arithmetic mean of the posterior probabilities of the events of the upstream and downstream observers is checked to complete the fault segment location.

[0023] Specifically, step S300 includes the following sub-steps: Step S310: Configure alarm signal observers for the feeder start-up, segment, and branch nodes according to the feeder segment location. and characteristic quantity observer .

[0024] Step S320: Model the telemetry and teleindication data of the distribution cloud master station and the virtual measurements obtained by analyzing the fault waveform files as fault detection factors; dynamically configure the fault detection factors to the two observers according to the jurisdiction of the section.

[0025] Step S330: Calculate the output result of each fault detection factor (where "1" indicates that a grounding event was detected, and "0" indicates that no grounding event was detected).

[0026] Specifically, step S400 includes the following sub-steps: Step S410: Assume that the observer is configured with A set of fault detection factors ,in For the detection factors in the observer The preset detection factor performance parameters are defined; the detection accuracy is described by conditional probability (the value is either an empirical value or a multi-event posterior correction value), specifically defined as follows: The probability that an event occurs and a correct alert is issued (hit rate); The probability that an event occurs but no alarm is triggered (false alarm rate); The probability of a false alarm being triggered even though the event has not occurred (false alarm rate); The probability that the event did not occur and no alert was issued (correct rejection rate).

[0027] Note: , Furthermore, the reliability of different fault detection factors is independent.

[0028] To reflect the reliability of different fault detection factors, a resolution weight is defined (to enhance the influence of the dominant factor). This indicates that the fault detection factor is effective (its resolution is higher than that of randomness). The larger the value, the higher the reliability of the fault detection factor. Its calculation formula is: (1).

[0029] Step S420: Calculate the multi-factor weighted likelihood value and Based on the independence of fault detection factors, the likelihood is the product of the output probabilities of individual fault detection factors within the observer; to avoid numerical underflow, a logarithmic transformation is used to optimize the likelihood value.

[0030] Key design principles of multi-factor weighted likelihood include: ① "Positive superposition"—the probability of an event should monotonically increase as the number of test factors with an output of 1 increases; ② "Negative penalty"—the test factors with an output of 0 should be used as inhibitory factors to reduce the probability of an event, rather than being directly eliminated.

[0031] Weighted likelihood value when the event occurs : First, calculate the log-likelihood of the weighted likelihood of the above events occurring: (2) Reconstruct the weighted likelihood of the event occurrence: (3) Weighted likelihood value when the event has not occurred : First, calculate the log-likelihood of the weighted likelihood that the above events have not occurred: (4) No weighted likelihood value was obtained when the event was restored: (5) Step S430: Calculate the weighted evidence factor The calculation formula is: (6) in, The prior probability of the event occurring. The prior probability of the event not occurring is given by the dominant evidence: when the alarm signal observer... A certain detection factor in hour, =0.7, The same applies to the opposite; and the system is dynamically updated based on changes (corrections) indicated by alarm signals. and The value of .

[0032] Step S440: Calculate the posterior probability of events for the observers using the enhanced multi-factor Bayesian algorithm: The posterior probability of an event occurring under multiple detection factors: (7) The posterior probability that an event has not occurred under multiple detection factors: (8) Specifically, step S500 includes the following sub-steps: Step S510, Missed Report Correction: Traverse the observer set of each segment along the feeder, if (in For example, a value of 0.8 indicates that no ground fault alarm signals were reported within the section, but high-confidence ground fault characteristic telemetry was detected. Based on the assumption of consistent ground fault evidence, it is determined that the FTU / DTU has experienced alarm "missed reporting" (missed reporting of downstream faults under reverse power flow). In this case, the alarm signal observer will be... The alarm signal with the highest confidence weight is corrected to "1", and steps S430~S440 are re-executed.

[0033] Step S520, False Alarm Correction: Traverse the observer set of each segment along the feeder, if The display section reported a ground fault alarm signal but no ground fault characteristic telemetry was detected. Furthermore, when the zero-mode signal was traced from downstream to upstream according to the upstream-downstream relationship of the feeder section, the current centroid frequency... If the indication is "0", then based on the assumption of consistent evidence for ground faults, it is determined that the FTU / DTU has issued a false alarm (false alarm of upstream fault under reverse power flow). In this case, the alarm signal observer will be activated. The alarm signal with the highest confidence weight is corrected to "0", and steps S430~S440 are re-executed.

[0034] Step S530, Fault Path Correction: Trace upstream from downstream according to the upstream and downstream relationship of the feeder section. If the condition is met... Based on the assumption of ground fault characteristic continuity, the upstream alarm signal observer is determined. An alarm was missed; at this time, the upstream alarm signal observer will be activated. The alarm signal with the highest confidence weight is corrected to "1", and steps S430~S440 are re-executed.

[0035] It should be noted that this invention does not mandate the use of assignment to correct the posterior probability of observer events. In a preferred embodiment, a conflict evidence smoothing formula can be used to mitigate evidence conflicts, as follows: , , Note: This represents the likelihood value of "randomly guessing the detection factor".

[0036] Specifically, step S600 includes the following sub-steps: Step S610, execution of the line selection strategy: the suspected grounding fault range defined in step S100; when there are ≥2 suspected grounding faults on feeders, the pairwise line characteristic quantity observers are used. The posterior probability of the event is compared, if the following conditions are met. (in If the value is, for example, 0.7, then it is determined that... The probability of line failure is high.

[0037] It should be noted that this invention does not mandate the use of a feature quantity observer. Post-event probability comparison method: If the distribution automation master station can recall the fault waveform files of the first switches of two lines and perform zero-mode current polarity calculation (or zero-mode current waveform similarity comparison), then the line selection can be directly performed using the differentiated indication of detection factors such as zero-mode current polarity and zero-mode current waveform similarity.

[0038] Step S620, Execution of the segment selection strategy (eliminating non-faulty line segments): Based on the upstream and downstream relationships of the suspected faulty line segments constructed in step S200, traverse the first-level branch nodes of the line, if the conditions are met... (in If the value is, for example, 0.3), then it is determined that the fault did not occur in that branch. Traverse the remaining feeder chain sections from downstream to upstream; if the downstream section... If the fault does not occur downstream of the faulty section, then it is determined that the fault did not occur downstream of that section. After completing the above steps, the non-faulty sections of the faulty line are eliminated and assessed.

[0039] Furthermore, if the branch line structure is complex and requires the assistance of this method for segment selection and positioning, the branch line can be regarded as a virtual line, and steps S300~S500 can be repeated.

[0040] Step S630, final location of the faulty section: Traverse each segment of the remaining chain structure from downstream to upstream, and check whether each segment meets the requirements. (in (For example, a value of 0.8). After verification, the last segment of the remaining suspected faulty circuit chain structure is determined to be the faulty segment (i.e., the fault selection is completed).

[0041] Example 2 This embodiment presents a method for locating single-phase grounding fault sections based on comprehensive analysis of multiple fault detection factors from a distribution cloud master station. Figure 1 As shown, the steps are as follows: Step S100: Obtain information such as single-phase grounding alarm of distribution terminal and small current grounding alarm of substation, use them as the start signal for fault analysis, and delineate the suspected grounding fault range of distribution network.

[0042] like Figure 2 As shown, in a specific implementation, the activation signals for single-phase ground fault assessment events in the distribution network include the following four categories: 1) The power distribution cloud master station updates the dispatch cloud master station with the newly added small current grounding line selection signal in the substation in real time; 2) Real-time monitoring of newly added FTU / DTU and traveling wave terminal grounding alarm signals and zero-sequence overcurrent alarm signals on the distribution network feeder of the distribution cloud master station; 3) The power distribution cloud master station has added a phase loss signal for the distribution transformer; 4) Manually trigger the grounding selection signal.

[0043] Based on any of the above signals, initiate the ground fault assessment procedure for the distribution cloud master station, and delineate the suspected ground fault range of the distribution network according to the signal indication.

[0044] Reference Figure 3 As shown, in one specific implementation, the step of delineating the suspected distribution network grounding fault range includes: 1) Control the newly added ground fault lines (or groups) of the low current ground fault selection device in the cloud synchronous substation. 2) FTU / DTU and traveling wave terminal grounding alarm signals installed in the distribution network lines, and newly added grounding fault lines (or collections). 3) After screening, the distribution network operators manually input the line set (or set).

[0045] Step S200: Obtain the topology of the suspected ground fault line, the upstream and downstream connection relationship of the terminal (adjacent matrix), the terminal ledger, the terminal operation and alarm information, complete the virtual measurement calculation of fault recording file recall and feature extraction, and identify the neutral point grounding mode of the distribution network.

[0046] In one specific embodiment, the step includes: 1) Connect to the power grid resource business platform to obtain the topology map of suspected faulty lines and the distribution automation terminal model (terminal type, connection relationship), and obtain the observation point map of suspected faulty lines; 2) Connect the network topology service, segment the feeder according to the FTU / DTU location, traverse all terminal telemetry, tele-signaling real-time data and fault recording data in each segment, and obtain detection factors (characteristic quantities) for single-phase grounding fault assessment, including but not limited to: three-phase load current, zero-sequence current / zero-sequence current mutation, zero-sequence voltage, and the angle between zero-sequence voltage and zero-sequence current. 3) Connect to the power distribution cloud master station to obtain the grounding signal from the substation's low-current grounding fault location device and the bus grounding signal; 4) Complete the virtual measurement calculation of the fault recording file and identify the neutral grounding method of the distribution network.

[0047] Specifically, the virtual measurement calculation for completing the feature extraction of the fault recording file and identifying the neutral grounding mode of the distribution network includes: 1) If the distribution network line within the fault range is equipped with an FTU / DTU / transient waveform type fault indicator, and the master station supports fault waveform file recall, then the file transfer protocol is started to perform the recall. 2) This embodiment does not specifically limit the identification method of the neutral grounding mode of the distribution network. In actual implementation, the following two schemes can be adopted: Option 1: Connect the substation model to the control cloud master station to directly obtain the substation neutral point grounding method (including ungrounded, arc suppression coil grounded, arc suppression coil parallel medium resistance grounded, and small resistance grounded), and at the same time obtain the operating status of the neutral point arc suppression coil to avoid misjudgment that the coil is under maintenance. Option 2: Calculate the zero-sequence impedance using fault recording files. The zero-sequence impedance is calculated by using the incremental method (division by variation) to measure the minute changes in zero-sequence voltage and current (ΔU0, ΔI0) before and after the fault. The formula for calculating the zero-sequence impedance is:

[0048] When zero-sequence reactive power Or the phase difference between zero-sequence voltage and zero-sequence current This indicates that the power distribution network is an ungrounded system.

[0049] Step S300: Construct an alarm signal observer according to the line and line section division. and characteristic quantity observer Based on the route selection and segment selection business scenarios, and combined with multiple fault detection factors, a fault detection factor set is dynamically constructed. And complete the calculation of fault detection factors.

[0050] Step S310: Configure the feeder start, segment, and branch nodes according to the feeder segment location. Observer (container); Normally, the two fault observers of the line are reused with the two observers of the first segment after the line section is divided; in a specific embodiment, if the first segment of the line is not configured with an FTU, the feeder outlet switch or the first segment switch can be configured as a line observer.

[0051] In a preferred embodiment, the step of dividing the feeder into sections according to a preset feeder segmentation principle specifically involves: preferably dividing the main line into sections based on the upstream and downstream positions of the distribution network feeder switches, and dividing the branch lines into sections based on the first-level branch switches. If a feeder section has multiple terminals (such as FTUs, fault indicators, and traveling wave terminals), then the grounding alarm information and grounding assessment telemetry data of the relevant terminals need to be configured into the corresponding observers.

[0052] Step S320: Model the telemetry and teleindication data of the distribution cloud master station, as well as the virtual measurements obtained by analyzing the fault waveform files, into fault detection factors, and dynamically configure the fault detection factors into the two observers according to the jurisdiction of the section.

[0053] When dynamically loading detection factors into observers, different types of fault indication factors are typically loaded into different observers based on signal type, detection principle, and source of trait values. In a preferred embodiment, an alarm signal observer is provided. and characteristic quantity observer .in: Feature Observer It includes steady-state telemetry and fault event telemetry of smart terminals, and can also include virtual measurements obtained from fault recording signal analysis (as a fault detection factor).

[0054] Set of detection factors in two observers The type and quantity of virtual telemetry are determined by the analysis of alarm signals, telemetry data, and transient signals within the jurisdiction.

[0055] This invention emphasizes the flexibility of multi-platform solutions and focuses on the dynamic aggregation of multiple detection factors from various terminals based on on-site telemetry measurements to form a multi-dimensional comprehensive observer. Furthermore, a transient signal analysis observer can be configured according to actual needs. A preferred embodiment of the detection factor set configuration is shown in Table 1: Table 1 Detection Factor Set Configuration Table

[0056] Note*: This invention does not require the use of a specific transient fault feature detection algorithm; relevant transient fault feature detection methods can refer to existing technologies.

[0057] Step S330: Calculate the output result of each detection factor (where “1” indicates that a grounding event was detected and “0” indicates that no grounding event was detected).

[0058] In a preferred embodiment, the based The observer records the output of the multivariate detection factors (single detection factor) respectively. The output is either "1" or "0", depending on the corresponding threshold setting. This invention does not recommend a specific activation threshold; the threshold setting can be determined based on the power grid parameters of the application location and historical data analysis. One set of empirical thresholds is shown in Table 2: Table 2 Detection Factor Thresholds and Judgment Criteria

[0059] Step S400: Use the enhanced multi-factor Bayesian algorithm to calculate the posterior probabilities of the two observer events for each line and each section.

[0060] Reference Figure 4 As shown, in step S410: Line observers are set up at L1#2 of feeder 1 and L2#2 of feeder 2; segment observers are set up at L1#2, L1#5, L1#7 of feeder 1 and the first segment L1-1#1, L1-2#1 of the first-level branch line, and similarly at L2#2, L2#3, L2#6 of feeder 2.

[0061] Taking L1#2, L1#5, and L2#2 as examples, segment ( / line) observers and In the middle, the detection accuracy parameters of alarm signals and telemetry detection factors ( , , and ), detection factor failure and resolution weight Example data is shown in Table 3.

[0062] Step S420: Calculate the weighted likelihood of the event occurring and the weighted likelihood of the event not occurring, and calculate the prior probability of the event according to the policy. and ; Step S430: Calculate the weighted evidence factor ; Step S440: Calculate the posterior probability of the event occurrence for the two observers using the enhanced multi-factor Bayesian algorithm. and .

[0063] Table 3 L1#2 Node Observer Detection Factor Parameter Table

[0064] Based on the posterior probability of the event and Based on the calculation results, it can be determined that the grounding event downstream of node L1#2 is highly likely.

[0065] Table 4 shows the multi-factor Bayesian posterior probability calculation data for the L1#5 node observer: Table 4 L1#5 Node Observer Detection Factor Parameter Table

[0066] The multifactor Bayesian posterior probability calculation data for the L1# node observer are shown in Table 5: Table 5 L1#7 Node Observer Detection Factor Parameter Table

[0067] Step S500: When the alarm signal observer... With characteristic quantity observer When there is conflict of evidence, the alarm signal observer is modified using the evidence consistency assumption. The signal with the highest confidence weight is used to eliminate false alarms and missed alarms from the terminal.

[0068] Step S510: Verify the posterior probability of the event occurrence for observer L1#5, and calculate... According to the ground fault characteristic consistency rule, it was determined that the L1#5 FTU had missed the ground fault alarm signal; the "ground fault alarm signal" of the L1#5 node was corrected to "1", and the parameters in Table 4 were recalculated, with the results shown in Table 6. After correction, the L1#5... =0.9993 and =0.9850, which indicates that the probability of a grounding event occurring in this section is relatively high.

[0069] Table 6. Detection factor parameters after correction for node observer L1#5

[0070] Step S520: Verify the posterior probability of the L1#7 observer event, and calculate... And the zero-mode current centroid frequency of node L1#7 The detection factor is indicated as "0"; based on the ground fault characteristic consistency rule, the FTU of node L1#7 is determined to be a false ground fault alarm signal; the "ground fault alarm signal" of node L1#7 is corrected to "0", and Table 5 is recalculated, with the results shown in Table 7. After correction, L1#7's... =0.001 and =0.013, which indicates that the probability of a grounding event occurring downstream of L1#7 is low.

[0071] Table 7. Corrected detection factor parameters for L1#7 node observer

[0072] Step S600: Select the line according to the event posterior probability comparison strategy of the feature quantity observer; select the segment according to the arithmetic mean of the event posterior probabilities of the two observers, and eliminate the non-faulty line segment; finally, check the arithmetic mean of the event posterior probabilities of the upstream and downstream observers according to the remaining chain topology to complete the fault segment location.

[0073] Step S610, Execution of line selection strategy: In the distribution network suspected grounding fault range defined in step S100, when there are ≥2 suspected grounding faults on feeders, the event posterior probability comparison of the pairwise line characteristic quantity observers is adopted. If the value of the event posterior probability of the first line minus the event posterior probability of the second line is greater than the second preset threshold, then the fault probability of the first line is determined to be high. Step S620, execution of the segment selection strategy: Based on the upstream and downstream relationships of the suspected faulty line segment constructed in step S200, branch line elimination: traverse each first-level branch node of the line. If the arithmetic mean of the event posterior probability of the alarm signal observer and the event posterior probability of the feature quantity observer is less than the third preset threshold, it is determined that the fault did not occur in that branch; end feeder segment elimination: traverse the remaining feeder chain segments from downstream to upstream. If the arithmetic mean of the event posterior probability of the downstream segment is less than the third preset threshold, it is determined that the fault did not occur downstream of that segment. Step S630: Verify the remaining suspected line segments of feeder 1 [L1#2~L1#7), L1#2's =0.9893, L1#5 =0.9921, the arithmetic mean of the posterior probabilities of observer events in each segment. All are greater than 0.8, while L1#7's =0.0068, therefore, based on comprehensive analysis, the feeder section from L1#5 to L1#7 is the section where the ground fault occurred.

[0074] The above description of the embodiments is provided to enable those skilled in the art to understand and apply the present invention. Those skilled in the art can readily make various modifications to the above embodiments and apply the general principles described herein to other embodiments without creative effort. Therefore, the present invention is not limited to the above embodiments, and any improvements and modifications made to the present invention by those skilled in the art based on the disclosure thereof should be within the scope of protection of the present invention.

Claims

1. A method for locating single-phase grounding fault sections based on centralized multi-factor comprehensive analysis at a main station, characterized in that, Including the following steps: S100 acquires single-phase grounding alarm information from distribution terminals and low-current grounding alarm information from substations, uses them as the start signal for fault analysis, and delineates the suspected grounding fault range of the distribution network. S200 acquires the topology of suspected ground fault lines, upstream and downstream terminal connection relationships, terminal ledgers, terminal operation and alarm information, completes virtual measurement calculations for fault recording file recall and feature extraction, and identifies the neutral point grounding mode of the distribution network. S300 constructs alarm signal observers and feature quantity observers according to the division of lines and line sections. Based on the line selection and section selection service scenarios, it dynamically constructs a fault detection factor set and completes the fault detection factor calculation by combining multiple fault detection factors. S400 uses an enhanced multi-factor Bayesian algorithm to calculate the posterior probability of events for two observers in each line and each section. S500: When there is a conflict between the evidence from the alarm signal observer and the feature quantity observer, the evidence consistency assumption is used to correct the signal indication with the highest confidence weight in the alarm signal observer, and to eliminate false alarms and missed alarms at the terminal. S600 selects lines based on the posterior probability comparison strategy of the feature quantity observers; it selects segments based on the arithmetic mean of the posterior probabilities of the two observers, eliminating non-faulty line segments; and finally, based on the remaining chain topology, it verifies the arithmetic mean of the posterior probabilities of the upstream and downstream observers to complete the fault segment location.

2. The method for locating a single-phase ground fault section according to claim 1, characterized in that, Step S300 includes: Step S310: Configure alarm signal observers and characteristic quantity observers for the feeder head, segment and branch nodes according to the feeder segment position.

3. The method for locating a single-phase ground fault section according to claim 2, characterized in that, Step S300 also includes: Step S320: The telemetry data, teleindication data, and virtual measurements obtained from the analysis of fault recording files of the power distribution cloud master station are modeled as fault detection factors, and the fault detection factors are dynamically configured into two observers according to the jurisdiction of the section.

4. The method for locating a single-phase ground fault section according to claim 3, characterized in that, Step S300 also includes: Step S330: Calculate the output result for each fault detection factor.

5. The method for locating a single-phase ground fault section according to claim 1, characterized in that, Step S400 includes: Step S410: In the alarm signal observer and the feature quantity observer, it is assumed that a fault detection factor set with N fault detection factors is configured. ,in ; For each fault detection factor Preset performance parameters, including hit rate underreporting rate False alarm rate and correct rejection rate ; Define the confidence weight of the detection factor ,in This indicates that the fault detection factor is effective.

6. The method for locating a single-phase ground fault section according to claim 5, characterized in that, Step S400 also includes: Step S420: Based on the independence of fault detection factors, calculate the weighted likelihood of event occurrence and the weighted likelihood of event non-occurrence. The likelihood is the weighted product of the output probabilities of individual fault detection factors within the observer. Logarithmic transformation is used to optimize the likelihood value calculation. Step S430: Calculate the weighted evidence factor based on the weighted likelihood of the event occurring and the weighted likelihood of the event not occurring, and adjust the prior probability of the dynamically updated event according to the alarm signal indication. Step S440: Based on the weighted likelihood, weighted evidence factor, and dynamically updated prior probability, the enhanced multi-factor Bayesian algorithm is used to calculate the posterior probability of events for the two observers of each line and each section.

7. The method for locating a single-phase ground fault section according to claim 6, characterized in that, Step S500 includes: S510, Missed Report Correction: Traverse the observer set of each segment along the feeder, if... If the value is greater than the first preset threshold, it is determined that the alarm signal observer has missed an alarm. The alarm signal with the highest confidence weight in the alarm signal observer is corrected to "1", and steps S430-S440 are executed again. in, For characteristic quantity observer The posterior probability of the event, Alarm signal observer The posterior probability of the event.

8. The method for locating a single-phase ground fault section according to claim 7, characterized in that, Step S500 also includes: Step S520, False Alarm Correction: Traverse the observer set of each segment along the feeder, if If the current centroid frequency is "0" and the zero mode is traced from downstream to upstream according to the upstream and downstream relationship of the feeder section, it is determined that the alarm signal observer has a false alarm. The alarm signal with the highest confidence weight in the alarm signal observer is corrected to "0", and steps S430-S440 are re-executed.

9. The method for locating a single-phase ground fault section according to claim 8, characterized in that, Step S500 also includes: Step S530, Fault Channel Correction: According to the upstream and downstream relationship of the feeder section, trace from downstream to upstream. If the value of the posterior probability of the downstream alarm signal observer event minus the posterior probability of the upstream alarm signal observer event is greater than the first preset threshold, then based on the ground fault characteristic continuity assumption, it is determined that the upstream alarm signal observer has missed alarm reporting. The alarm signal with the highest confidence weight in the upstream alarm signal observer is corrected to "1", and steps S430-S440 are re-executed.

10. The method for locating a single-phase ground fault section according to claim 1, characterized in that, Step S600 includes: Step S610, Execution of line selection strategy: In the distribution network suspected grounding fault range defined in step S100, when there are ≥2 suspected grounding faults on feeders, the event posterior probability comparison of the pairwise line characteristic quantity observers is adopted. If the value of the event posterior probability of the first line minus the event posterior probability of the second line is greater than the second preset threshold, then the fault probability of the first line is determined to be high. Step S620, execution of the segment selection strategy: Based on the upstream and downstream relationships of the suspected grounding fault line segment constructed in step S200, traverse the first-level branch nodes of the line. If the arithmetic mean of the event posterior probability of the alarm signal observer and the event posterior probability of the feature quantity observer is less than the third preset threshold, it is determined that the fault did not occur in that branch. Traverse the remaining feeder chain segments from downstream to upstream. If the arithmetic mean of the event posterior probability of the downstream segment is less than the third preset threshold, it is determined that the fault did not occur downstream of that segment. Step S630, final location of the faulty section: Traverse each segment of the remaining chain structure from downstream to upstream, and check whether each segment meets the following condition: the arithmetic mean of the posterior probability of the alarm signal observer event and the posterior probability of the feature quantity observer event is greater than the fourth preset threshold. After the check is passed, the last segment of the remaining suspected faulty line chain structure is determined to be the faulty segment.

Images