Self-adaptive dual-mode detection method and system for wire tree touch fault of power distribution network
By adopting an adaptive dual-mode detection method, combining adaptive window calculation baseline and trend analysis under high current and low current modes, the problem of balancing response speed and accuracy in the detection of tree contact faults in distribution network conductors is solved, and rapid and accurate fault determination is achieved.
Patent Information
- Application Number
- CN202510986816.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-17
- Publication Date
- 2025-11-18
AI Technical Summary
Existing methods for detecting tree-touching faults in power distribution networks struggle to balance response speed and accuracy. Traditional high-current detection is susceptible to interference and false alarms, while low-current detection has a slow response speed, and fixed threshold methods are difficult to adapt to different environments.
An adaptive dual-mode detection method is adopted, which collects zero-sequence current data in real time and combines adaptive window calculations, trend analysis and progressive confidence accumulation in high current and low current modes to achieve fast and accurate fault determination.
In high-current scenarios, faults can be quickly identified, while in low-current scenarios, false alarms can be suppressed through trend analysis and confidence accumulation, ensuring robust detection under different environments and improving detection accuracy and response speed.
Smart Images
Figure CN120971836A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of power distribution network fault detection, in particular to a power distribution network wire tree contact fault adaptive dual-mode detection method and system. BACKGROUND
[0002] With the continuous development of power distribution network automation and intelligent technology, the problem of tree contact fault caused by wire contact with trees in power distribution lines is increasingly prominent, especially in mountainous and suburban areas due to dense vegetation, resulting in a high incidence of faults. Traditional fault detection methods are mainly divided into two categories: (1) Fast detection method based on large current sudden change characteristics: a fixed threshold is preset for judgment, and when the zero sequence current instantaneously exceeds the threshold, it is directly judged as a fault. The advantage is fast response, but it is easily disturbed by load switching, weather disturbance and other factors, resulting in false positives.
[0003] (2) Monitoring method based on slow rising trend of small current: using the slow change trend of zero sequence current and long-term data statistics to identify the problem, although it can reduce false positives, but the response speed is slow when the fault has deteriorated, there is a risk of delayed confirmation.
[0004] The existing method uses only large current or small current detection strategy, which is not easy to balance response speed and accuracy, and the fixed threshold method is difficult to adapt in different operating environments. How to realize adaptive switching of detection mode during system operation, comprehensive utilization of multi-period trend characteristics and design of gradient type credibility accumulation mechanism has become a key problem to be solved.
[0005] In view of this, a power distribution network wire tree contact fault adaptive dual-mode detection method and system are needed. SUMMARY
[0006] In view of the problems in the prior art, the present application provides a power distribution network wire tree contact fault adaptive dual-mode detection method and system, which can quickly and accurately detect faults in large current scenarios, and at the same time, in small current fault scenarios, through dual-time scale trend analysis and gradual credibility accumulation to realize stable decision. The specific technical scheme is as follows: A power distribution network wire tree contact fault adaptive dual-mode detection method, characterized in that it comprises the following steps: Step S1, real-time acquisition of power distribution line zero sequence current data; Step S2, preprocessing of the collected power distribution line zero sequence current data; Step S3, calculating the zero sequence current baseline according to the preprocessed data through an adaptive window; Step S4, according to the preset large current threshold and the initial stable period, it is judged that the collected data is to enter the large current mode or the small current mode; if it is to enter the small current mode, it is transferred to step S5, if it is to enter the large current mode, it is transferred to step S7; Step S5, in the small current mode, using short-term and long-term data windows, respectively adopting a first-order polynomial fitting, according to the trend slope, the change rate, the fitting degree and the fluctuation index, the comprehensive trend score is calculated; Step S6, according to the comprehensive trend score, using the gradual cumulative confidence mechanism, the fault confidence is calculated; Step S7, in the large current mode, when the zero sequence current continuously exceeds the threshold value for a preset confirmation time, the fault is directly determined; in the small current mode, when the fault confidence reaches the high confidence threshold and maintains for a certain time, the fault is determined; Step S8, the fault state and the corresponding diagnostic information are output.
[0007] Preferably, the preprocessing includes downsampling, calculating the effective value of the primary zero sequence current and filtering processing, the step of downsampling includes: The collected set distribution line zero sequence current signal is downsampled to obtain the downsampled zero sequence current data, wherein the expression of the downsampling is: ; Wherein, is the downsampled zero sequence current data, which is a subset of the original data; X is the original distribution line zero sequence current data, which is obtained by real-time collection of the line monitoring device; d is the downsampling factor, N is the length of the original distribution line zero sequence current data, indicating the total number of sample points collected; The frequency after downsampling is: : ; Wherein, is the original sampling frequency.
[0008] Preferably, the expression of calculating the effective value of the primary zero sequence current is: ; Wherein, t is the time index, indicating the tthsecond; is the effective value of the zero sequence current in the tthsecond; n is the sampling point number per second, which is equal to the sampling frequency after downsampling ; is the starting sampling point index of the tthsecond, is the ending sampling point index of the tthsecond; j is the sampling point index in the range of to ; is the downsampled zero sequence current data value at the jthpoint.
[0009] Preferably, the step S3 of calculating the zero-sequence current baseline according to the preprocessed data through the adaptive window specifically comprises the following steps: In the initial stage, a short-term average is used as the zero-sequence current baseline, and the calculation formula is as follows: ; Wherein, B(t) is the zero-sequence current baseline value at time t; is a time threshold set in the initial stage; min(t0, t) represents the smaller value of t0 and the current time t; max(1, t- t0): the larger value of 1 and t- t0; In the stable operation, the average value of the sliding window is used to update the zero-sequence current baseline, and the calculation formula is as follows: ; Wherein, is the zero-sequence current baseline calculation window width.
[0010] Preferably, it further comprises: In the fault state, the updated rate frozen baseline is used, and the specific calculation formula is as follows: ; ; Wherein, and are smoothing factors, and , .
[0011] Preferably, the determination condition of the large current mode is: ; Wherein, M(t) is the detection mode at time t, M(t)=1 represents the small current mode, and M(t)=2 represents the large current mode; is the fault judgment threshold in the large current mode, is the initial stable period time threshold.
[0012] Preferably, the calculation of the comprehensive trend score specifically comprises: The short-term trend score is calculated in the short-term data window, and the specific calculation formula is as follows: , ; Wherein, is the zero-sequence current short-term trend score at time t; is the zero-sequence current short-term relative change rate at time t; is the short-term goodness of fit; is the zero-sequence current short-term fluctuation index; Fluctuation weights are used to penalize unstable data. The slope of the short-term trend of the zero-sequence current; The minimum short-term slope threshold; The minimum short-term slope threshold; This is the width of the short-term trend analysis window; The short-term trend score is calculated using the following formula within a long-term data window: ; in, The score represents the long-term trend of the zero-sequence current at time t. Let be the long-term relative rate of change of the zero-sequence current at time t; For long-term goodness of fit; This is the long-term fluctuation index of zero-sequence current; The slope of the long-term trend of the zero-sequence current; The minimum long-term slope threshold; This represents the long-term absolute change in zero-sequence current, indicating the absolute value of the current difference between the beginning and end of the long-term data window. The threshold for the minimum absolute change; The weights of the short-term and long-term trend scores are dynamically adjusted based on the actual running time, and a final composite trend score is obtained. The specific calculation formula is as follows: ; ; ; in, The overall trend score at time t; Let t be the weight of the short-term trend. Let be the weight of the long-term trend at time t. This is the weight conversion period.
[0013] Preferably, the step of calculating the fault credibility based on the comprehensive trend score using a progressive cumulative credibility mechanism includes the following specific calculation formula: ; in, Let be the reliability of the fault at time t. This is the initial credibility threshold. This is the threshold for increasing credibility.
[0014] Preferably, the step of directly determining a fault when the zero-sequence current continuously exceeds a threshold for a preset confirmation time specifically includes: The specific formula for calculating the high-current counter value is as follows: ; wherein, is a large current counter value at time t, counted once per second, the counter value is incremented by one as long as the current second is in the large current mode, otherwise it is cleared; when a fault is determined, wherein is a large current preset confirmation time; when and the duration a fault is confirmed; wherein: is the fault credibility at time t, is a high credibility threshold; is a small current confirmation time; is the cumulative time required for the rising trend in the small current mode.
[0015] A power distribution network conductor tree fault adaptive dual-mode detection system applies the method, comprising: a data acquisition unit for real-time acquisition of power distribution line zero sequence current data; a preprocessing unit for preprocessing the acquired power distribution line zero sequence current data, the preprocessing including downsampling, calculating the effective value of the primary zero sequence current, and filtering processing; a baseline calculation unit for calculating the zero sequence current baseline through an adaptive window according to the preprocessed data; a mode selection unit for determining whether the collected data enters the large current mode or the small current mode according to a preset large current threshold and an initial stable period; if it enters the small current mode, it is transferred to the trend analysis unit, and if it enters the large current mode, it is transferred to the fault judgment unit; a trend analysis unit for using short-term and long-term data windows in the small current mode, respectively adopting a first-order polynomial fitting, calculating the comprehensive trend score according to the trend slope, the change rate, the fitting goodness and the fluctuation index; a credibility calculation unit for calculating the fault credibility by using a gradual accumulation credibility mechanism according to the comprehensive trend score; a fault judgment unit for directly determining a fault when the zero sequence current continuously exceeds the threshold value for a preset confirmation time in the large current mode, and determining a fault when the fault credibility reaches a high credibility threshold and maintains for a certain time in the small current mode; a fault output unit for outputting the fault state and the corresponding diagnostic information.
[0016] Compared with the prior art, the beneficial effects of the present application are: (1) The present application takes into account the response speed and accuracy: the large current mode can confirm the fault in a short time, and the small current mode effectively suppresses false positives through trend analysis and credibility accumulation, ensuring that the system can quickly and accurately respond to various faults.
[0017] (2) The application has strong adaptability: the adaptive update of the baseline and the dynamic switching of the mode adapt to different operating environments and system disturbances, and realize robust detection in multiple scenarios.
[0018] (3) The application has high intelligence: the comprehensive evaluation of multi-dimensional features (trend slope, goodness of fit, volatility index, etc.) makes the fault determination closer to expert experience judgment, and has good engineering practicability and popularization value. BRIEF DESCRIPTION OF DRAWINGS
[0019] In order to more clearly illustrate the technical solutions in the specific embodiments or prior art of the present application, the drawings needed in the specific embodiments or prior art description will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, each element or part is not necessarily drawn according to the actual proportion.
[0020] Figure 1 The method flowchart of the present application.
[0021] Figure 2 The simulation result figure for fast confirmation of the large current threshold.
[0022] Figure 3 The simulation result figure for slow confirmation of the small current trend.
[0023] Figure 4 The simulation result figure for switching from the initial small current to the large current mode.
[0024] Figure 5 The system principle diagram of the present application. DETAILED DESCRIPTION
[0025] The technical solutions in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.
[0026] It should be understood that when used in the present specification and the appended claims, the terms "comprise" and "include" indicate the presence of the described features, whole, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, whole, steps, operations, elements, components and / or sets thereof.
[0027] It should also be understood that the terms used in the specification and the following claims are intended to describe particular embodiments by way of example only and are not intended to limit the application. As used in the specification and the appended claims, the singular forms "a," "an" and "the" are intended to include plural referents unless the context clearly dictates otherwise.
[0028] It should further be understood that the term "and / or" as used in the specification and the following claims indicates any combination of one or more of the associated listed items and all possible combinations of those items.
[0029] Embodiment 1: As shown in the figure, the embodiment provides a power distribution network conductor tree fault adaptive dual-mode detection method, including the following steps: Figure 1 Step S1, real-time collection of power distribution line zero sequence current data. The sampling frequency is usually 2 kHz or higher.
[0030] Step S2, preprocessing of the collected power distribution line zero sequence current data, including downsampling, calculation of the effective value of the primary zero sequence current and filtering processing. The collected power distribution line zero sequence current signal is downsampled to obtain the downsampled zero sequence current data, wherein the expression of the downsampling is: ; Wherein, is the downsampled zero sequence current data, which is a subset of the original data; X is the original power distribution line zero sequence current data, which is collected in real time by the line monitoring device; d is the downsampling factor, N is the length of the original power distribution line zero sequence current data, indicating the total number of collected sample points; The frequency after downsampling is: ; Wherein, is the original sampling frequency, typically 2 kHz or higher, indicating the number of samples collected per second.
[0031] The expression for calculating the effective value of the primary zero sequence current is: ; Wherein, t is the time index, indicating the tthsecond; is the effective value of the zero sequence current at the tthsecond; n is the number of samples per second, equal to the sampling frequency after downsampling ; is the starting sample point index at the tthsecond, is the index of the end sampling point of the t-th second; j is a summation variable to the index of the sampling point within the range; is the down-sampled zero sequence current data value at point j.
[0032] Step S3, calculate the zero sequence current baseline through adaptive windowing according to the preprocessed data. The baseline calculation combines the short-term average in the initial stage and the sliding window average in the normal operation stage, and adopts an extremely low update rate to freeze the baseline when a fault occurs. The present application adopts a dynamic baseline calculation method to adapt to system fluctuations and fault states. Specifically, the following steps are included: In the initial stage, the short-term average is used as the zero sequence current baseline, and the calculation formula is as follows: ; where t is the current time point, seconds; B(t) is the zero sequence current baseline value at time t; is the time threshold set in the initial stage, generally set to 30 seconds; min(t0, t) indicates the smaller value of t0 and the current time t, which is used to dynamically adjust the calculation window size in the initial stage; max(1, t- t0): take the larger value of 1 and t- t0, to ensure that the index is not less than 1.
[0033] In stable operation, the average value of the sliding window (recommended 180-second window) is used to update the zero sequence current baseline, and the calculation method is as follows: ; where is the zero sequence current baseline calculation window width, and the recommended value is 180 seconds.
[0034] In the fault state, an extremely low update rate is used to freeze the baseline to ensure the relative stability of the detection index. Specifically as follows: ; ; where and are smoothing factors, and , .
[0035] In the present embodiment , where = 0.995, = 0.005 are smoothing factors, which control the baseline update speed, and the smaller 0.005 makes the baseline hardly change in the fault state.
[0036] Add a baseline minimum value limit, take the larger one of the baseline value and 0.001, to avoid division by zero error, specifically as follows: .
[0037] Step S4, according to the preset large current threshold and the initial stable period, it is judged whether the collected data is in the large current mode or the small current mode; if it is in the small current mode, it is transferred to step S5, and if it is in the large current mode, it is transferred to step S7. In the dual-mode detection, the detection mode is determined by comparing the real-time zero sequence current with the high current threshold, and in the large current mode, it is directly confirmed, and in the small current mode, it is confirmed by relying on long-term trend and credibility accumulation.
[0038] The determination condition of the large current mode is: ; Among them, M(t) is the detection mode at t moment, M(t)=1 indicates the small current mode, and M(t)=2 indicates the large current mode; is the fault determination threshold in the large current mode, and the recommended value is 1.0A; is the initial stable period time threshold, and the recommended value is 30 seconds, avoiding false alarm in the initial system startup. Based on the determination condition, the application can automatically identify the large current and small current fault scenes, and then adaptively select the best detection mode.
[0039] Step S5, in the small current mode, short-term and long-term data windows are used, first-order polynomial fitting is used respectively, and comprehensive trend score is calculated according to the trend slope, change rate, fitting goodness and fluctuation index. In the small current mode, the short-term and long-term trend scores are combined by weighting, wherein the weight gradually transitions from short-term to long-term with the system running time, and the final comprehensive trend score is adjusted combined with the rising trend duration. The calculation method of the comprehensive trend score is as follows: (1) the short-term trend score is calculated in the short-term data window, and the specific method is as follows: ; ; Among them, is the zero sequence current short-term trend score at t moment; is the zero sequence current short-term relative change rate at t moment, with the unit of % / minute; is the short-term fitting goodness, which measures the quality of fitting; is the zero sequence current short-term fluctuation index, which represents the instability degree of the zero sequence current; is the fluctuation weight, and the recommended value is 2, which is used to punish unstable data; is the zero sequence current short-term trend slope, which represents the change rate of the zero sequence current; is the minimum short-term slope threshold; is the minimum short-term slope threshold, and the recommended value is 2.0%, which filters the small changes; For short-term trend analysis window width. Only when the upward trend is obvious, the short-term trend score is calculated.
[0040] 1) First-order polynomial fitting is as follows: ; Where, is the zero sequence current short-term trend intercept, the formula is fitted in window, is the short-term trend analysis window width, the recommended value is 30 seconds.
[0041] 2) The calculation method of zero sequence current short-term relative change rate at time t is as follows: ; Where, 60 is the coefficient for conversion to per minute, and 100 is the coefficient for conversion to percentage.
[0042] 3) The calculation method of short-term fitting goodness is as follows: ; is the zero sequence current short-term fitting residual sum of squares, indicating the square sum of the deviation of the data points of the zero sequence current from the fitting line, and the calculation formula is: .
[0043] is the total deviation square sum of zero sequence current short-term, indicating the square sum of the deviation of the data points of the zero sequence current from the mean, and the calculation formula is: ; Where is the mean of zero sequence current short-term data, and the calculation formula is: ; Where the recommended value is 30 seconds.
[0044] 4) The calculation method of zero sequence current short-term fluctuation index is as follows: ; Where, indicates taking the larger value between the mean of short-term data and 0.001 to avoid division by zero.
[0045] is the residual standard deviation of zero sequence current short-term fitting, indicating the dispersion degree of the fitting residual, and the calculation formula is: (2) Calculate the long-term trend score under the long-term data window, also use the first-order polynomial fitting, in addition to the calculation of the trend slope, also pay attention to the absolute change of data to determine the reliability of the long-term trend. Specifically as follows: ; Where, is the long-term trend score of zero sequence current at time t; t is the long-term relative change rate of zero sequence current at time t, unit: % / min; is the long-term goodness of fit; is the long-term fluctuation index of zero sequence current; is the long-term trend slope of zero sequence current; is the minimum long-term slope threshold, recommended value is 0.5%; is the long-term absolute change of zero sequence current, indicating the absolute value of the current difference between the beginning and end of the long-term data window; is the minimum absolute change threshold, recommended value is 0.05A. Long-term goodness of fit , long-term fluctuation index of zero sequence current The calculation formula of is the same as that of short-term goodness of fit , short-term fluctuation index of zero sequence current , only the calculation window is changed to the long-term trend analysis window width.
[0046] 1) The first-order fitting of zero sequence current is as follows: ; is the long-term trend intercept of zero sequence current. The formula is fitted within window, is the long-term trend analysis window width, recommended value is 180 seconds.
[0047] 2) The calculation method of long-term relative change rate of zero sequence current at time t is as follows: .
[0048] 3) The calculation method of long-term absolute change of zero sequence current is as follows: .
[0049] 4) The calculation method of long-term goodness of fit is as follows: ; .
[0050] Total deviation square sum, representing the square sum of deviation of data points from the mean, is calculated as follows: ; Wherein, Zero sequence current short-term data mean, is calculated as follows: .
[0051] 5) Zero sequence current long-term fluctuation index The calculation method is as follows: ; Wherein, Zero sequence current long-term fitting residual standard deviation, representing the dispersion degree of fitting residual, is calculated as follows: .
[0052] (3) According to the actual running time, the short-term trend score and the long-term trend score weight are dynamically adjusted, and the final comprehensive score is synthesized, and the calculation formula is as follows: ; ; ; Wherein, The comprehensive trend score at time t; The weight of short-term trend at time t, which gradually decreases with the system running time; The weight of long-term trend at time t, which gradually increases with the system running time; The weight conversion period is recommended to be 600 seconds, and the max and min functions ensure that the weight value remains between 0.1 and 0.9.
[0053] Step S6, according to the comprehensive trend score, the gradual accumulation confidence mechanism is adopted to calculate the fault confidence. The gradual confidence accumulation mechanism sets three threshold values (initial, rising and alarm), and updates the confidence based on the current comprehensive trend score at different accumulation rates.
[0054] The present application carries out the smooth, step-by-step rising confidence calculation of the comprehensive trend score, realizes it by setting three confidence thresholds (initial 0.35, confirmation 0.6 and alarm 0.8): when the comprehensive trend score is higher than a certain threshold, it is accumulated with different weights, and when the long-term rising trend disappears, the confidence decreases with the corresponding decay rate.
[0055] The calculation method of fault confidence is as follows: ; Wherein, The fault confidence at time t; is the initial credibility threshold, and the recommended value is 0.35; is the credibility rising threshold, and the recommended value is 0.6. Here is the zero sequence current long-term fluctuation index, which is used to accelerate the credibility accumulation.
[0056] The credibility value is limited between 0 and 1 to ensure the numerical stability of the calculation, and the details are as follows: .
[0057] Step S7, in the large current mode, when the zero sequence current continuously exceeds the threshold for a preset confirmation time, the fault is directly determined; in the small current mode, when the fault credibility reaches the high credibility threshold and maintains for a certain time, the fault is determined.
[0058] In the large current mode, the zero sequence current continuously exceeding the threshold is specifically calculated by the following method: ; wherein, is the large current counter value at time t, which is counted once per second. If the current second is in the large current mode, the counter value is incremented by one, otherwise it is cleared; when , the fault is determined, wherein is the large current preset confirmation time, and the recommended value is 5 seconds.
[0059] In the small current mode, it is required that the comprehensive trend score reaches the alarm threshold within the observation window, and the high credibility maintains for a certain time (for example, 30 seconds) to confirm the fault, that is: when and the duration , the fault is confirmed; wherein: is the fault credibility at time t, is the high credibility threshold, and the recommended value is 0.8; is the small current confirmation time, and the recommended value is 30 seconds; is the cumulative time required for the rising trend in the small current mode, and the recommended value is 120 seconds.
[0060] Based on the above fault determination method, the fault is quickly confirmed in the large current mode, avoiding false positives caused by transient disturbances. In the small current mode, the trend score and reliability are comprehensively calculated through short-term and long-term trend analysis, reducing the influence of noise.
[0061] Step S8, output the fault state and the corresponding diagnostic information. When the fault is confirmed, the fault confirmation time, the fault credibility, the detection basis and the related trend statistical data are output, which is convenient for subsequent diagnosis and disposal. The fault detection result can be visually displayed, and the communication module is used to report the abnormal situation to the upper system.
[0062] The application designs an adaptive baseline and a credibility accumulation mechanism to ensure that the system does not trigger false positives during normal fluctuations, and can quickly capture abnormal changes during faults.
[0063] Figure 2 The simulation result graph of large current threshold fast confirmation is given, the initial fault current is greater than 1A, the large current mode is started, and the fault confirmation is completed at about 5 seconds after the fault occurs at the data time of 39 seconds, and the detection basis is the long-term trend confirmation of small current.
[0064] Figure 3 The simulation result graph of small current trend slow confirmation is given, the fault current is less than 1A for a long time, and the zero sequence current threshold method is difficult to detect. Using the method of the application, the fault confirmation is completed at the data time of 214 seconds, the fault credibility is 1.00, and the detection basis is the long-term trend confirmation of small current.
[0065] Figure 4 The simulation result graph of initial small current switching large current mode is given, the initial fault current is less than 1A, the detection method starts in small current mode, and after the fault lasts for a few seconds, the current reaches 1A, the detection mode is changed from small current mode to large current mode, and the fault fast confirmation is completed at the data time of 35 seconds, and the detection basis is the direct confirmation of large current.
[0066] The detection method of the application is suitable for application in multiple power grid scenes, mainly including: Medium and low voltage distribution line monitoring: in the area with dense vegetation, the line tree contact situation is monitored in real time and early warning is provided.
[0067] Smart grid SCADA system: as a basic fault indication unit, detailed trend analysis data and fault diagnosis information are provided to the upper dispatching system.
[0068] Rural and mountainous area power grid reconstruction: in the power grid upgrading and reconstruction project, the adaptive detection device is installed to assist fault positioning and line optimization.
[0069] Embodiment 2: As Figure 5 shown, based on the same inventive concept as embodiment 1, the embodiment provides a distribution line tree contact fault adaptive dual-mode detection system, and the method comprises: A data acquisition unit is used to acquire zero sequence current data of the distribution line in real time. A preprocessing unit is used to preprocess the acquired zero sequence current data of the distribution line, and the preprocessing includes downsampling, calculating the effective value of the primary zero sequence current, and filtering processing. A baseline calculation unit is used to calculate the zero sequence current baseline through an adaptive window according to the preprocessed data. A mode selection unit is configured to determine whether the collected data is entering a large current mode or a small current mode according to a preset large current threshold and an initial stable period; if the data is entering the small current mode, the mode selection unit is connected to a trend analysis unit; if the data is entering the large current mode, the mode selection unit is connected to a fault judgment unit; The trend analysis unit is configured to use short-term and long-term data windows to respectively adopt a first-order polynomial fitting in the small current mode, and to calculate a comprehensive trend score according to a trend slope, a change rate, a fitting goodness and a fluctuation index. The credibility calculation unit is configured to calculate a fault credibility according to the comprehensive trend score by using a gradual cumulative credibility mechanism. The fault judgment unit is configured to directly determine a fault when the zero sequence current continuously exceeds a threshold for a preset confirmation time in the large current mode, and to determine a fault when the fault credibility reaches a high credibility threshold and is maintained for a certain time in the small current mode. The fault output unit is configured to output a fault state and corresponding diagnostic information.
[0070] Those skilled in the art can understand that the units of the examples described in combination with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components of the examples have been described in general terms in the above description. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.
[0071] In the embodiments provided in the present application, it should be understood that the division of units is only a logical functional division, and actual implementation can have another division manner, for example, multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored, etc.
[0072] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.
[0073] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or say the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a plurality of instructions to make a computer device (which can be a personal computer, a server or a network device, etc.) execute all or part of the steps of the method described in various embodiments of the present application. The foregoing storage medium includes: a U disk, a read-only memory (ROM, Read-0nly Memory), a random access memory (RAM, Random Access Memory), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.
[0074] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and they should be covered in the scope of the claims and the specification of the present application.
Claims
1. A power distribution grid conductor arcing fault adaptive dual mode detection method, characterized in that, The method comprises the following steps: Step S1, collecting zero sequence current data of the power distribution line in real time; Step S2, preprocessing the collected zero sequence current data of the power distribution line; Step S3, calculating the zero sequence current baseline through an adaptive window according to the preprocessed data; Step S4, determining whether the collected data enters a large current mode or a small current mode according to a preset large current threshold and an initial stable period; if the data enters the small current mode, the method proceeds to step S5; if the data enters the large current mode, the method proceeds to step S7; Step S5, in the small current mode, using short-term and long-term data windows, respectively adopting a first-order polynomial fitting, and calculating a comprehensive trend score according to a trend slope, a change rate, a fitting goodness and a fluctuation index; Step S6, calculating a fault confidence according to the comprehensive trend score by using a gradual accumulation confidence mechanism; Step S7, in the large current mode, directly determining a fault when the zero sequence current continuously exceeds a threshold value for a preset confirmation time; in the small current mode, determining a fault when the fault confidence reaches a high confidence threshold and is maintained for a certain time; Step S8, outputting a fault state and corresponding diagnostic information.
2. The adaptive dual-mode detection method for line-to-tree faults of a power distribution network according to claim 1, characterized in that, The preprocessing comprises downsampling, calculating an effective value of the primary zero sequence current and filtering processing, and the downsampling step comprises: downsampling the collected zero sequence current signal of the power distribution line to obtain the downsampled zero sequence current data, wherein the downsampling expression is: ； wherein, is the down-sampled zero sequence current data, which is a subset of the original data; X is the original distribution line zero sequence current data, which is collected by the line monitoring device in real time; d is the down-sampling factor, and N is the length of the original distribution line zero sequence current data, representing the total number of sample points collected. down-sampled frequency is: ； wherein is the original sampling frequency.
3. The adaptive dual-mode detection method for line-to-tree faults of a distribution network according to claim 2, characterized in that, the expression for calculating the effective value of the primary zero sequence current is: ； wherein t is a time index, representing the tth second; is the zero sequence current effective value of the tth second; n is the number of sampling points per second, equal to the sampling frequency after down-sampling ; is the starting sampling point index of the tth second, is the ending sampling point index of the tth second; j is a summation variable in the range of to ; is the down-sampled zero sequence current data value at the jth point.
4. The adaptive dual-mode detection method for distribution network conductor fault according to claim 1, characterized in that, the step S3 of calculating the zero sequence current baseline through the adaptive window according to the preprocessed data specifically comprises the following steps: in the initial stage, using a short-term average as the zero sequence current baseline, wherein the calculation formula is as follows: ； Wherein, B(t) is the zero sequence current baseline value at t moment; a time threshold set for the initial stage; min(t0, t) represents taking the smaller value of t0 and the current time t; max(1, t-t0): taking the larger value of 1 and t-t0; in the stable operation, updating the zero sequence current baseline by using the average value of the sliding window, wherein the calculation formula is as follows: ； wherein, The window width for the zero sequence current baseline is calculated.
5. The adaptive dual-mode detection method for distribution network conductor fault according to claim 4, characterized in that, further comprising: in the fault state, using an update rate to freeze the baseline, and the specific calculation formula is as follows: ； ； wherein and are smoothing factors, respectively, and , .
6. The adaptive dual-mode detection method for distribution network conductor fault according to claim 1, characterized in that, the determination condition of the large current mode is: ； Wherein, M(t) is the detection mode at t moment, M(t)=1 represents small current mode, M(t)=2 represents large current mode; is the fault determination threshold value in large current mode, is the initial stable period time threshold value.
7. The adaptive dual-mode detection method for distribution network conductor fault according to claim 1, characterized in that, the calculation of the comprehensive trend score specifically comprises: calculating a short-term trend score in the short-term data window, and the specific calculation formula is as follows: ， ； wherein, is the zero sequence current short-term trend score at time t; is the zero sequence current short-term relative change rate at time t; is the short-term goodness of fit; is the zero sequence current short-term fluctuation index; is the fluctuation weight used to penalize unstable data; is the zero sequence current short-term trend slope; is the minimum short-term slope threshold; is the minimum short-term slope threshold; is the short-term trend analysis window width; calculating a long-term trend score in the long-term data window, and the specific calculation formula is as follows: ； wherein, is a zero sequence current long term trend score at time t; is a zero sequence current long term relative change rate at time t; is a long term goodness of fit; is a zero sequence current long term fluctuation index; is a zero sequence current long term trend slope; is a minimum long term slope threshold; is a zero sequence current long term absolute change, representing the absolute value of the difference between the beginning and end of the long term data window; is a minimum absolute change threshold; adjusting the short-term trend score and the long-term trend score weight according to the actual operation time to finally synthesize the comprehensive trend score, and the specific calculation formula is as follows: ； ； ； wherein, is the combined trend score at time t; is the weight of the short-term trend at time t, is the weight of the long-term trend at time t, is the weight conversion period.
8. The adaptive dual-mode detection method for distribution network conductor fault according to claim 7, characterized in that, the calculation of the fault confidence according to the comprehensive trend score by using the gradual accumulation confidence mechanism specifically comprises: ； wherein, is the failure credibility at time t, is the initial credibility threshold, is the credibility rise threshold.
9. The adaptive dual-mode detection method for distribution network conductor fault according to claim 1, characterized in that, the direct determination of the fault when the zero sequence current continuously exceeds the threshold value for the preset confirmation time specifically comprises: calculating a large current counter value, and the specific calculation formula is as follows: ； wherein, is the large current counter value at time t, counted once per second, and the counter value is incremented by one if the current second is in the large current mode, otherwise it is cleared; and a fault is determined when is the large current preset confirmation time; When and the duration a fault is confirmed; wherein: is the fault credibility at time t, is a high credibility threshold; is a small current confirmation time; is a cumulative time of rising trend requirement in the small current mode.
10. An adaptive dual mode detection system for detecting faults in distribution network conductors, comprising: The method comprises the following steps: a data collection unit, which collects zero sequence current data of the power distribution line in real time; a preprocessing unit, which preprocesses the collected zero sequence current data of the power distribution line; a baseline calculation unit, which calculates a zero sequence current baseline through an adaptive window according to the preprocessed data; A mode selection unit determines whether the collected data is entering a large current mode or a small current mode according to a preset large current threshold and an initial stable period; if it is entering the small current mode, it is transferred to a trend analysis unit, and if it is entering the large current mode, it is transferred to a fault judgment unit; The trend analysis unit uses short-term and long-term data windows in the small current mode, respectively adopts a first-order polynomial fitting, and calculates a comprehensive trend score according to a trend slope, a change rate, a fitting goodness and a fluctuation index; A credibility calculation unit calculates a fault credibility according to the comprehensive trend score and using a gradual cumulative credibility mechanism; The fault judgment unit directly determines a fault when the zero sequence current continuously exceeds a threshold value for a preset confirmation time in the large current mode, and determines the fault when the fault credibility reaches a high credibility threshold value and is maintained for a certain time in the small current mode; A fault output unit outputs a fault state and corresponding diagnostic information.