Power distribution network fault type intelligent identification method based on multi-parameter analysis

Abstract

The application discloses a power distribution network fault type intelligent identification method based on multi-parameter analysis and relates to the technical field of power distribution network fault identification.The application aims at the problem that the prior art cannot accurately identify intermittent arc grounding faults, collects voltage and current signals of key nodes of a power distribution network through a fault recording device, marks a signal mutation period to form a fluctuation time sequence, analyzes dynamic changes of electric signal frequencies of each mutation period, calculates an overlapping degree and a frequency change speed in combination with inherent interference frequency bands, generates a discrimination marker value of arc mutation, separates mixed signals and extracts fault characteristic components based on multi-dimensional indexes such as time intervals and electric signal frequency change speeds, generates a fault characteristic marker value, and finally comprehensively judges whether the current mutation belongs to an intermittent arc fault and distinguishes other voltage disturbances or phase-to-phase short-circuit faults, so that the misjudgment of the intermittent arc fault can be effectively reduced, and the accuracy and reliability of power distribution network fault identification are improved.

Description

Technical Field

[0001] This invention relates to the field of distribution network fault identification technology, specifically a method for intelligent identification of distribution network fault types based on multi-parameter analysis. Background Technology

[0002] When a single-phase intermittent arcing-to-ground fault occurs in a distribution network, the arc at the fault point repeatedly extinguishes and reignites, generating a series of broadband transient pulse groups. The center frequency of these pulse groups drifts significantly during the arcing process. Existing technologies typically collect three-phase voltage, current, and zero-sequence electrical quantities, and use wavelet packet decomposition, Fourier transform, or empirical mode decomposition to extract features such as fixed-band energy ratio and transient amplitude, which are then input into classifiers such as neural networks or support vector machines. These methods have a high recognition rate for faults with stable characteristics, such as metallic grounding and two-phase short circuits. However, for intermittent arcing-to-ground faults, this fixed-band decomposition strategy is difficult to effectively capture the true fault characteristics.

[0003] The drift frequency band of intermittent arcing ground fault pulses overlaps with two types of signals: switching noise generated during normal operation of distributed power sources and the main frequency band of transient traveling waves generated by two-phase short-circuit faults. Simultaneously, the rapid rise in the recovery voltage of the faulted phase during the arc-extinguishing phase causes complex distortions and fluctuations in the waveforms of the zero-sequence and negative-sequence voltages, characteristics very similar to ferroresonance or inrush current during no-load closing. Existing methods based on fixed frequency bands or fixed time windows cannot dynamically track the real-time changes in the center frequency of the arc pulses, nor can they effectively separate the true fault characteristics from the aliased signals. This leads to frequent misjudgments of intermittent arcing ground faults as two-phase short circuits or ferroresonance in distribution networks with a high proportion of power electronic devices. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent identification method for distribution network fault types based on multi-parameter analysis, so as to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent identification method for distribution network fault types based on multi-parameter analysis, comprising the following steps: Step S100: Collect three-phase voltage, three-phase current, zero-sequence voltage and zero-sequence current signals at each feeder, transformer and distributed power source grid connection point through the fault recording device deployed in the distribution network; perform noise reduction and time alignment processing on the original signals, mark the start and end times of all voltage or current abrupt changes as signal abrupt change periods, and form an electrical signal fluctuation time series in chronological order. Step S200: For each marked signal abrupt change period, analyze the dynamic change of the electrical signal frequency over time and record the trajectory of the electrical signal frequency offset; at the same time, extract the characteristic frequency band of the inherent interference signal existing in the normal operation state of the fault recording device, calculate the degree of overlap between the frequency of the abrupt change period and the inherent interference frequency band and the rate of change of the electrical signal frequency, and generate the discrimination mark value of the corresponding abrupt change period belonging to the arc-type abrupt change. Step S300: Decompose the waveform recording signal of each signal change period into normal operation component, interference component and suspected fault component; calculate the difference value between suspected fault component and interference component, and generate the fault feature mark value corresponding to the change period. Step S400: Based on the discrimination marker value generated in step S200 and the fault feature marker value generated in step S300, identify the fault type of the current signal change and output the corresponding alarm prompt synchronously.

[0006] S100 specifically includes the following steps: Step S101: Deploy fault recording devices at each feeder, transformer and distributed power source grid connection point in the distribution network to collect three-phase voltage, three-phase current, zero-sequence voltage and zero-sequence current signals in real time at a set sampling frequency. Step S102: Denoise the acquired raw signal by using a median filter to remove impulse noise and a low-pass filter to eliminate high-frequency interference; normalize the denoised signal to transform all electrical quantities to the [0,1] interval. Step S103: Calculate the signal difference between adjacent sampling points, set a sudden change judgment threshold as an anomaly distinction standard, mark the signal sudden change start time when the absolute value of the difference between three consecutive sampling points exceeds the preset threshold; mark the signal sudden change end time when the absolute value of the difference between three consecutive sampling points is continuously lower than the threshold; extract the corresponding time period signal according to the sudden change start point and sudden change end point to form an electrical signal fluctuation time series.

[0007] S200 specifically includes the following steps: Step S201: For each marked signal abrupt change period in the electrical signal fluctuation time series, starting from the start time of the abrupt change, the first signal segment is extracted as a sliding window with a preset fixed time width; then, the sliding window is moved backward with a preset fixed time step, and subsequent signal segments are extracted sequentially until the entire abrupt change period is covered; for each extracted sliding window, the average value of the signal within the window is first calculated, and the average value is subtracted from each sample value; then, the sample values ​​at both ends of the window are smoothed and attenuated; next, a fast Fourier transform is performed on the processed window signal to obtain the amplitude distribution of the signal at different frequencies within the sliding window; the frequency point with the highest amplitude is found from the amplitude distribution as the instantaneous frequency corresponding to the center time of the current sliding window; each window is processed sequentially to obtain an instantaneous frequency sequence arranged in chronological order, denoted as {f1, f2, ..., f n}, where n is the total number of sliding windows, f n The instantaneous frequency values ​​corresponding to n sliding windows are represented; the center time of each window is used as the abscissa and the corresponding instantaneous frequency is used as the ordinate. Points are plotted in a plane rectangular coordinate system and the points are connected in sequence to form the frequency offset trajectory curve F(t); Step S202: Based on the instantaneous frequency sequence, calculate the rate of change vf of the electrical signal frequency during the abrupt change period. The calculation formula is as follows: ; Where Δt is the time interval between adjacent sliding windows, f k and f k+1 These represent the instantaneous frequency values ​​corresponding to the k-th and (k+1)-th sliding windows, respectively; simultaneously, the frequency drift range Δf = max(f1, f2, ..., f n )-min(f1,f2,…,f n ), which represents the maximum fluctuation width of the frequency during the mutation period.

[0008] Step S203: Extract a segment of steady-state data collected by the fault recording device under normal operating conditions of the distribution network, and the data duration shall not be less than a preset duration threshold; for each phase voltage or current signal in the steady-state data, first calculate the average value of the entire signal segment, and subtract the average value from each sample value; then, perform a fast Fourier transform on the processed signal to obtain the amplitude results of the signal at each frequency point; plot the amplitude distribution curve of the amplitude changing with frequency with frequency as the abscissa and the corresponding amplitude as the ordinate; find the maximum amplitude in the entire frequency range on the amplitude distribution curve and denot it as Amax; Set the amplitude detection threshold Ath, Ath = Amax × η; where η is the adjustment coefficient, 0.1 < η < 0.3; starting from the lowest frequency, scan the amplitude curve along the frequency increase direction. When the amplitude of all frequency points in a continuous frequency range is higher than Ath, mark the frequency range as a candidate frequency band and record the start and end frequencies of the candidate frequency band. For each candidate frequency band, the steady-state data is divided into multiple sub-segments according to a preset number of segments. The aforementioned Fast Fourier Transform and candidate frequency band extraction process is performed on each sub-segment. If a candidate frequency band appears in all sub-segments, and the bandwidth variation of the candidate frequency band among the sub-segments does not exceed a preset bandwidth ratio threshold, and the ratio of the maximum amplitude to the minimum amplitude within the frequency band is less than a preset amplitude ratio threshold, then the frequency band is determined to be an inherent interference frequency band, denoted as the interval set [L]. j U j ] m j=1 Where m is the number of inherent interference frequency bands identified, and L j and U j These are the upper and lower limits of the j-th inherent interference frequency band, respectively.

[0009] Step S204: Calculate the degree of overlap O between the instantaneous frequency sequence and the inherent interference frequency band. The calculation formula is as follows: ; Where |•| represents the total length of the interval; This represents the union of all inherent interference frequency bands identified in step S203. If two frequency bands overlap, they are merged into a continuous interval; otherwise, they remain as multiple separate intervals. The value of 0 ranges from [0,1]. For the k-th instantaneous frequency value f in the instantaneous frequency sequence... k Define the neighborhood interval as [f k -δ,f k +δ]; If the neighborhood interval is completely contained within a certain inherent interference frequency band. If the neighborhood interval has no intersection with any inherent interference frequency band, then ; δ is the frequency sampling half-width, set as the standard deviation of the frequency difference between adjacent windows: ; Where, d k =f k+1 -f k ; .

[0010] Step S205: Combining the frequency change rate, frequency drift range, and overlap, generate a discriminant value λ indicating that the abrupt change period belongs to an arc-type abrupt change. The calculation formula is as follows: ; Wherein, vf0 is the baseline value of the frequency change rate of the historical arc sample, which is obtained from the statistical data of confirmed intermittent arc faults; Δf0 is the baseline value of the frequency drift range of the historical arc sample, which is also obtained from the statistical data of historical faults; w1 is the weight corresponding to the frequency change rate, with a value range of 0.5 to 0.7; w2 is the weight corresponding to the degree of overlap between the frequency and the interference frequency band, with a value range of 0.2 to 0.3; w3 is the weight corresponding to the frequency drift range, with a value range of 0.1 to 0.2; and w1+w2+w3=1.

[0011] Step S300 specifically includes the following steps: Step S301: For each marked signal abrupt change period, calculate the following five basic characteristic quantities: the duration from the start time to the end time, the rate of change of the electrical signal frequency obtained from step S202, the difference between the voltage waveform and the standard sine waveform, the maximum difference ratio between each phase voltage, and the voltage recovery rate; wherein, the difference between the voltage waveform and the standard sine waveform is obtained by dividing the average of the point-by-point deviations of the actual voltage sample value and the standard sine wave of the same frequency and amplitude by the rated voltage amplitude; the maximum difference ratio between each phase voltage is obtained by dividing the maximum absolute value of the difference between any two phases of the three-phase voltage effective value by the rated phase voltage; the voltage recovery rate is the average rate of change of the voltage within the time required for the voltage to recover to more than 95% of the rated value from the end of the abrupt change; Step S302: Establish the normal waveform range and interference value range for each characteristic quantity in advance by using historical normal operation waveform data and interference event statistics of the distribution network; the interference event is an electrical quantity disturbance event caused by load switching, switching operation, operation of power electronic equipment and distributed power source fluctuations in the distribution network under fault-free conditions; If all basic characteristic quantities fall within the normal recording range, the electrical signal segment during the abrupt change period is classified as a normal operation component; if all basic characteristic quantities fall within the interference range, it is classified as an interference component; if any basic characteristic quantity exceeds the upper limit of the interference range, the electrical signal segment during the abrupt change period is identified as a suspected fault component. Record the basic characteristic quantities and their actual measured values ​​corresponding to all electrical signal segments identified as suspected fault components. Step S303: For each basic characteristic quantity that is judged as a suspected fault, take its average value in historical interference events as the interference benchmark value; if a basic characteristic quantity is not judged as a suspected fault, its difference is recorded as zero; otherwise, calculate the absolute difference between the actual measured value of the basic characteristic quantity and the interference benchmark value, and then divide it by the maximum difference that the corresponding basic characteristic quantity has in historical fault events to obtain the normalized difference value; add up the normalized differences of all basic characteristic quantities to obtain the total difference score, denoted as D, then the fault characteristic label value μ=min(D / 5,1), D∈[0,5], μ∈[0,1].

[0012] Step S400 selects several confirmed intermittent arc faults, phase-to-phase short-circuit faults, and conventional voltage disturbance events from the historical waveform data of the distribution network. It calculates the arc-induced abrupt change discrimination flag value and fault characteristic flag value corresponding to each historical event, forming template point sets for the three types of events. For the current abrupt change period to be identified, a feature point (λ,μ) is constructed based on the calculated λ and μ. The Euclidean distance between the feature point (λ,μ) and each template point in the three template point sets is calculated. The category to which the template point with the smallest Euclidean distance belongs is selected as the fault type identification result for the current signal abrupt change, and a corresponding alarm prompt is output synchronously.

[0013] Compared with the prior art, the beneficial effects of the present invention are: This invention uses instantaneous frequency analysis via a sliding window to dynamically track the frequency drift trajectory of intermittent electric arcs, calculate the frequency change rate and drift range, and extract the inherent interference frequency bands of the distribution network to calculate the degree of overlap. This overcomes the shortcomings of traditional fixed frequency band decomposition, which cannot capture real-time changes in electric arc frequency, and significantly reduces misjudgments caused by inherent interference such as distributed power supply switching noise and power electronic harmonics. This invention calculates the total difference score based on five basic characteristic quantities, including duration, three-phase imbalance, and voltage recovery rate, and generates fault characteristic marker values, thereby achieving reliable separation between real faults and conventional voltage disturbances such as load switching and no-load closing. This invention constructs two-dimensional feature points by combining arc fault discrimination marker values ​​and fault feature marker values. By matching the nearest neighbor with the historical template point set, it simultaneously identifies two dimensions: whether it is an intermittent arc fault and whether it is a real fault. This accurately distinguishes between intermittent arc faults, phase-to-phase short-circuit faults, and conventional voltage disturbances, thereby improving the accuracy of distribution network fault identification. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the method flow for the intelligent identification method of distribution network fault types based on multi-parameter analysis of the present invention. Detailed Implementation

[0015] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Example: Figure 1 As shown, this invention provides a technical solution: an intelligent identification method for distribution network fault types based on multi-parameter analysis, comprising the following steps: Step S100: Collect three-phase voltage, three-phase current, zero-sequence voltage and zero-sequence current signals at each feeder, transformer and distributed power source grid connection point through the fault recording device deployed in the distribution network; perform noise reduction and time alignment processing on the original signals, mark the start and end times of all voltage or current abrupt changes as signal abrupt change periods, and form an electrical signal fluctuation time series in chronological order. Step S200: For each marked signal abrupt change period, analyze the dynamic change of the electrical signal frequency over time and record the trajectory of the electrical signal frequency offset; at the same time, extract the characteristic frequency band of the inherent interference signal existing in the normal operation state of the fault recording device, calculate the degree of overlap between the frequency of the abrupt change period and the inherent interference frequency band and the rate of change of the electrical signal frequency, and generate the discrimination mark value of the corresponding abrupt change period belonging to the arc-type abrupt change. Step S300: Decompose the waveform recording signal of each signal change period into normal operation component, interference component and suspected fault component; calculate the difference value between suspected fault component and interference component, and generate the fault feature mark value corresponding to the change period. Step S400: Based on the discrimination marker value generated in step S200 and the fault feature marker value generated in step S300, identify the fault type of the current signal change and output the corresponding alarm prompt synchronously.

[0017] S100 specifically includes the following steps: Step S101: Deploy fault recording devices at each feeder, transformer and distributed power source grid connection point in the distribution network to collect three-phase voltage, three-phase current, zero-sequence voltage and zero-sequence current signals in real time at a set sampling frequency. Step S102: Denoise the acquired raw signal by using a median filter to remove impulse noise and a low-pass filter to eliminate high-frequency interference; normalize the denoised signal to transform all electrical quantities to the [0,1] interval. Step S103: Calculate the signal difference between adjacent sampling points, set a sudden change judgment threshold as an anomaly distinction standard, mark the start time of signal sudden change when the absolute value of the difference between three consecutive sampling points exceeds the preset threshold; mark the end time of signal sudden change when the absolute value of the difference between three consecutive sampling points is continuously lower than the threshold; extract the corresponding time period signal according to the start point and end point of sudden change to form an electrical signal fluctuation time series, and set the preset sudden change detection threshold to 10%~20% of the rated voltage or rated current.

[0018] Example 1: A fault recording device is deployed at the feeder of the distribution network, with a sampling frequency of 10 kHz to collect three-phase voltage, current and zero-sequence signals in real time; after median filtering and low-pass filtering to denoise the original signal, it is normalized to the [0,1] interval; the difference between adjacent sampling points is calculated, and the mutation threshold is set to 10% of the rated value; when the difference between three consecutive sampling points exceeds the threshold, the mutation is marked as starting, and when it is continuously below the threshold, the mutation is marked as ending; in a certain event, the mutation starts at 0.200 seconds, ends at 0.230 seconds, and lasts for 0.030 seconds. The signal during this period is extracted to form an electrical signal fluctuation time series.

[0019] S200 specifically includes the following steps: Step S201: For each marked signal abrupt change period in the electrical signal fluctuation time series, starting from the start time of the abrupt change, the first signal segment is extracted as a sliding window with a preset fixed time width; then, the sliding window is moved backward with a preset fixed time step, and subsequent signal segments are extracted sequentially until the entire abrupt change period is covered; for each extracted sliding window, the average value of the signal within the window is first calculated, and the average value is subtracted from each sample value; then, the sample values ​​at both ends of the window are smoothed and attenuated; next, a fast Fourier transform is performed on the processed window signal to obtain the amplitude distribution of the signal at different frequencies within the sliding window; the frequency point with the highest amplitude is found from the amplitude distribution as the instantaneous frequency corresponding to the center time of the current sliding window; each window is processed sequentially to obtain an instantaneous frequency sequence arranged in chronological order, denoted as {f1, f2, ..., f n}, where n is the total number of sliding windows, f n The instantaneous frequency values ​​corresponding to n sliding windows are represented; the center time of each window is used as the abscissa and the corresponding instantaneous frequency is used as the ordinate. Points are plotted in a plane rectangular coordinate system and the points are connected in sequence to form the frequency offset trajectory curve F(t); Step S202: Based on the instantaneous frequency sequence, calculate the rate of change vf of the electrical signal frequency during the abrupt change period. The calculation formula is as follows: ; Where Δt is the time interval between adjacent sliding windows, fk and f k+1 These represent the instantaneous frequency values ​​corresponding to the k-th and (k+1)-th sliding windows, respectively; simultaneously, the frequency drift range Δf = max(f1, f2, ..., f n )-min(f1,f2,…,f n ), which represents the maximum fluctuation width of the frequency during the mutation period.

[0020] Step S203: Extract a segment of steady-state data collected by the fault recording device under normal operating conditions of the distribution network, and the data duration shall not be less than a preset duration threshold; for each phase voltage or current signal in the steady-state data, first calculate the average value of the entire signal segment, and subtract the average value from each sample value; then, perform a fast Fourier transform on the processed signal to obtain the amplitude results of the signal at each frequency point; plot the amplitude distribution curve of the amplitude changing with frequency with frequency as the abscissa and the corresponding amplitude as the ordinate; find the maximum amplitude in the entire frequency range on the amplitude distribution curve and denot it as Amax; Set the amplitude detection threshold Ath, Ath = Amax × η; where η is the adjustment coefficient, 0.1 < η < 0.3; starting from the lowest frequency, scan the amplitude curve along the frequency increase direction. When the amplitude of all frequency points in a continuous frequency range is higher than Ath, mark the frequency range as a candidate frequency band and record the start and end frequencies of the candidate frequency band. For each candidate frequency band, the steady-state data is divided into multiple sub-segments according to a preset number of segments. The aforementioned Fast Fourier Transform and candidate frequency band extraction process is performed on each sub-segment. If a candidate frequency band appears in all sub-segments, and the bandwidth variation of the candidate frequency band among the sub-segments does not exceed a preset bandwidth ratio threshold, and the ratio of the maximum amplitude to the minimum amplitude within the frequency band is less than a preset amplitude ratio threshold, then the frequency band is determined to be an inherent interference frequency band, denoted as the interval set [L]. j U j ] m j=1 Where m is the number of identified inherent interference frequency bands, and L j and U j These are the upper and lower limits of the j-th inherent interference frequency band, respectively.

[0021] Step S204: Calculate the degree of overlap O between the instantaneous frequency sequence and the inherent interference frequency band. The calculation formula is as follows: ; Where |•| represents the total length of the interval; This represents the union of all inherent interference frequency bands identified in step S203. If two frequency bands overlap, they are merged into a continuous interval; otherwise, they remain as multiple separate intervals. The value of 0 ranges from [0,1]. For the k-th instantaneous frequency value f in the instantaneous frequency sequence... k Define the neighborhood interval as [f k -δ,f k +δ]; If the neighborhood interval is completely contained within a certain inherent interference frequency band. If the neighborhood interval has no intersection with any inherent interference frequency band, then ; δ is the frequency sampling half-width, set as the standard deviation of the frequency difference between adjacent windows: ; Where, d k =f k+1 -f k ; .

[0022] Step S205: Combining the frequency change rate, frequency drift range, and overlap, generate a discriminant value λ indicating that the abrupt change period belongs to an arc-type abrupt change. The calculation formula is as follows: ; Wherein, vf0 is the baseline value of the frequency change rate of the historical arc sample, which is obtained from the statistical data of confirmed intermittent arc faults; Δf0 is the baseline value of the frequency drift range of the historical arc sample, which is also obtained from the statistical data of historical faults; w1 is the weight corresponding to the frequency change rate, with a value range of 0.5 to 0.7; w2 is the weight corresponding to the degree of overlap between the frequency and the interference frequency band, with a value range of 0.2 to 0.3; w3 is the weight corresponding to the frequency drift range, with a value range of 0.1 to 0.2; and w1+w2+w3=1.

[0023] Example 2: A sliding window was used to extract instantaneous frequencies during abrupt change periods. The window width was 10 ms and the step size was 2 ms, resulting in a total of 10 instantaneous frequency values. 49.5, 49.0, 48.2, 47.5, 47.0, 47.2, 47.8, 48.5, 49.0, 49.3; Calculate the frequency change rate, where Δt = 0.002 seconds, and obtain vf = 250 Hz / s according to the formula; the frequency drift range Δf = max(f) - min(f) = 49.5 - 47.0 = 2.5 Hz; Extracting the inherent interference frequency band: Taking 0.2 seconds of steady-state data under normal operating conditions, and performing FFT and stability verification, the inherent interference frequency bands are obtained as [140,160] and [240,260] Hz; calculating the frequency overlap degree O, in this embodiment, all instantaneous frequencies are far from the interference frequency band, so O=0; based on the statistics of historical arc samples, taking the baseline values ​​vf0=200Hz / s and Δf0=3.0Hz, and setting the weights to w1=0.6, w2=0.2, w3=0.2, calculate the discrimination label value: λ=0.6×(250 / 200)+0.2×(2.5 / 3.0)+0.2×1=1.117; Step S300 specifically includes the following steps: Step S301: For each marked signal abrupt change period, calculate the following five basic characteristic quantities: the duration from the start time to the end time, the rate of change of the electrical signal frequency obtained from step S202, the difference between the voltage waveform and the standard sine waveform, the maximum difference ratio between each phase voltage, and the voltage recovery rate; wherein, the difference between the voltage waveform and the standard sine waveform is obtained by dividing the average of the point-by-point deviations of the actual voltage sample value and the standard sine wave of the same frequency and amplitude by the rated voltage amplitude; the maximum difference ratio between each phase voltage is obtained by dividing the maximum absolute value of the difference between any two phases of the three-phase voltage effective value by the rated phase voltage; the voltage recovery rate is the average rate of change of the voltage within the time required for the voltage to recover to more than 95% of the rated value from the end of the abrupt change; Step S302: Establish the normal waveform range and interference value range for each characteristic quantity in advance by using historical normal operation waveform data and interference event statistics of the distribution network; the interference event is an electrical quantity disturbance event caused by load switching, switching operation, operation of power electronic equipment and distributed power source fluctuations in the distribution network under fault-free conditions; If all basic characteristic quantities fall within the normal recording range, the electrical signal segment during the abrupt change period is classified as a normal operation component; if all basic characteristic quantities fall within the interference range, it is classified as an interference component; if any basic characteristic quantity exceeds the upper limit of the interference range, the electrical signal segment during the abrupt change period is identified as a suspected fault component. Record the basic characteristic quantities and their actual measured values ​​corresponding to all electrical signal segments identified as suspected fault components. Step S303: For each basic characteristic quantity that is judged as a suspected fault, take its average value in historical interference events as the interference benchmark value; if a basic characteristic quantity is not judged as a suspected fault, its difference is recorded as zero; otherwise, calculate the absolute difference between the actual measured value of the basic characteristic quantity and the interference benchmark value, and then divide it by the maximum difference that the corresponding basic characteristic quantity has in historical fault events to obtain the normalized difference value; add up the normalized differences of all basic characteristic quantities to obtain the total difference score, denoted as D, then the fault characteristic label value μ=min(D / 5,1), D∈[0,5], μ∈[0,1].

[0024] Example 3: For the marked signal abrupt change period, the duration of which is 0.03 seconds, five basic characteristic quantities are first calculated: duration is 0.03 seconds; the rate of change of electrical signal frequency is 250 Hz / s; the difference between the voltage waveform and the standard sine wave is 0.10; the maximum difference ratio between the phase voltages is 0.08; and the voltage recovery rate is 15. Based on historical normal operation waveform recording data and interference event statistics, the pre-established normal waveform recording range and interference value range for each characteristic quantity are as follows: duration is normally less than 0.005 seconds, and the interference range is 0.005 to 0.020 seconds; the rate of change of electrical signal frequency is normally less than 10 Hz / s, and the interference range is 10 to 80 Hz / s; the difference between the voltage waveform and the standard sine wave is normally less than 0.02, and the interference range is 0.02 to 0.06; the maximum difference ratio between the phase voltages is normally less than 0.03, and the interference range is 0.03 to 0.10; and the voltage recovery rate is normally greater than 100, and the interference range is 20 to 100. Comparing the measured values ​​with the above range, the duration of 0.030 seconds exceeds the upper limit of the interference range by 0.020 seconds; the frequency change rate of the electrical signal is 250Hz / s exceeding 80Hz / s; the difference between the voltage waveform and the standard sine wave is 0.10 exceeding 0.06; the voltage recovery rate is 15 below 20; and the maximum difference ratio between the phase voltages is 0.08 falling within 0.03 to 0.10. Therefore, the first four basic characteristic quantities exceed or fall below the upper or lower limit of the interference range, indicating that the waveform signal recorded during this abrupt change period contains a suspected fault component. The maximum difference ratio between the phase voltages falls within the interference range, corresponding to the interference component. The four basic characteristic quantities identified as suspected fault components and their actual measured values ​​are recorded. Subsequently, the interference baseline average values ​​of these four characteristic quantities are obtained from historical interference events, which are 0.012 seconds, 45Hz / s, 0.04, and 55, respectively. The maximum differences of these four characteristic quantities were obtained from historical fault events, which were 0.050 seconds, 300 Hz / s, 0.15, and 50, respectively. The absolute differences between the measured value and the interference reference value of each suspected fault characteristic quantity were calculated: the duration difference was 0.018 seconds, the electrical signal frequency change rate difference was 205 Hz / s, the voltage waveform difference from the standard sine wave was 0.06, and the voltage recovery rate difference was 40. These differences were divided by the corresponding maximum difference of historical faults to obtain normalized differences: 0.36, 0.683, 0.40, and 0.80. The maximum difference ratio between each phase voltage was not identified as a suspected fault, and its difference was recorded as zero. The normalized differences of all five basic characteristic quantities were added together to obtain a total difference score of 2.243. Finally, the fault characteristic label value was calculated by dividing the total difference score by 5 to obtain 2.243 / 5 = 0.4486. Step S400 selects several confirmed intermittent arc faults, phase-to-phase short-circuit faults, and conventional voltage disturbance events from the historical waveform data of the distribution network. It calculates the arc-induced abrupt change discrimination flag value and fault characteristic flag value corresponding to each historical event, forming template point sets for the three types of events. For the current abrupt change period to be identified, a feature point (λ,μ) is constructed based on the calculated λ and μ. The Euclidean distance between the feature point (λ,μ) and each template point in the three template point sets is calculated. The category to which the template point with the smallest Euclidean distance belongs is selected as the fault type identification result for the current signal abrupt change, and a corresponding alarm prompt is output synchronously.

[0025] Example 4: From the historical waveform data of the distribution network, 20 confirmed intermittent arc faults, 20 phase-to-phase short-circuit faults, and 20 conventional voltage disturbance events were selected respectively. The arc-induced abrupt change discrimination value and the fault feature discrimination value corresponding to each historical event were calculated to form template point sets for the three types of events. Through statistics, the center positions of the three types of templates are as follows: the discrimination value of intermittent arc faults is approximately 1.10, and the fault feature discrimination value is approximately 0.50; the discrimination value of phase-to-phase short-circuit faults is approximately 0.40, and the fault feature discrimination value is approximately 0.85; the discrimination value of conventional voltage disturbance events is approximately 0.65, and the fault feature discrimination value is approximately 0.25. For the current abrupt change period to be identified, based on the calculation results of Examples 2 and 3, the arcing change discrimination marker value is 1.117, and the fault characteristic marker value is 0.4486. These two values ​​are used as feature points, and the Euclidean distances between these points and the centers of the three types of templates are calculated respectively: the distance to the intermittent arcing fault template is approximately 0.056, the distance to the phase-to-phase short-circuit fault template is approximately 0.80, and the distance to the conventional voltage disturbance template is approximately 0.52. Among them, the distance to the intermittent arcing fault template is the smallest. Therefore, the current signal abrupt change is identified as an intermittent arcing fault, and an alarm prompt is output synchronously: an intermittent arcing fault has occurred on a certain feeder. Please arrange maintenance.

[0026] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent identification of distribution network fault types based on multi-parameter analysis, characterized in that: The method includes: Step S100: Collect three-phase voltage, three-phase current, zero-sequence voltage and zero-sequence current signals at each feeder, transformer and distributed power source grid connection point through the fault recording device deployed in the distribution network; perform noise reduction and time alignment processing on the original signals, mark the start and end times of all voltage or current abrupt changes as signal abrupt change periods, and form an electrical signal fluctuation time series in chronological order. Step S200: For each marked signal abrupt change period, analyze the dynamic change of the electrical signal frequency over time and record the trajectory of the electrical signal frequency offset; simultaneously extract the characteristic frequency band of the inherent interference signal existing in the normal operation state of the fault recording device, calculate the degree of overlap between the frequency of the abrupt change period and the inherent interference frequency band, as well as the rate of change of the electrical signal frequency, and generate a discrimination mark value for the corresponding abrupt change period belonging to an arc-type abrupt change; the calculation of the degree of overlap between the frequency of the abrupt change period and the inherent interference frequency band includes extracting the instantaneous frequency sequence from the abrupt change period and calculating the degree of overlap between the instantaneous frequency sequence and the inherent interference frequency band. The calculation formula is: ； Where |•| represents the total length of the interval; It represents the union of all inherent interference frequency bands. If two frequency bands overlap, they are merged into a continuous interval; otherwise, they remain as multiple separate intervals. The value range is [0,1]; for the k-th instantaneous frequency value f in the instantaneous frequency sequence k Define the neighborhood interval as [f k -δ,f k +δ]; If the neighborhood interval is completely contained within a certain inherent interference frequency band. If the neighborhood interval has no intersection with any inherent interference frequency band, then ; δ is the frequency sampling half-width, set as the standard deviation of the frequency difference between adjacent windows: ； Where, d k =f k+1 -f k ; ; Step S300: Decompose the waveform recording signal of each signal change period into normal operation component, interference component and suspected fault component; calculate the difference value between suspected fault component and interference component, and generate the fault feature mark value corresponding to the change period. Step S400: Based on the discrimination marker value generated in step S200 and the fault feature marker value generated in step S300, identify the fault type of the current signal change and output the corresponding alarm prompt synchronously.

2. The intelligent identification method for distribution network fault types based on multi-parameter analysis according to claim 1, characterized in that: S100 specifically includes the following steps: Step S101: Deploy fault recording devices at each feeder, transformer and distributed power source grid connection point in the distribution network to collect three-phase voltage, three-phase current, zero-sequence voltage and zero-sequence current signals in real time at a set sampling frequency. Step S102: Denoise the acquired raw signal by using a median filter to remove impulse noise and a low-pass filter to eliminate high-frequency interference; normalize the denoised signal to transform all electrical quantities to the [0,1] interval. Step S103: Calculate the signal difference between adjacent sampling points, set a sudden change judgment threshold as an anomaly distinction standard, mark the signal sudden change start time when the absolute value of the difference between three consecutive sampling points exceeds the preset threshold; mark the signal sudden change end time when the absolute value of the difference between three consecutive sampling points is continuously lower than the threshold; extract the corresponding time period signal according to the sudden change start point and sudden change end point to form an electrical signal fluctuation time series.

3. The intelligent identification method for distribution network fault types based on multi-parameter analysis according to claim 2, characterized in that: S200 specifically includes the following steps: For each marked signal abrupt change period in the electrical signal fluctuation time series, starting from the initial time of the abrupt change, the first signal segment is extracted as a sliding window with a preset fixed time width. Then, the sliding window is moved backward with a preset fixed time step, and subsequent signal segments are extracted sequentially until the entire abrupt change period is covered. For each extracted sliding window, the average value of the signal within the window is first calculated, and this average value is subtracted from each sample value. Then, the sample values ​​at both ends of the window are smoothed and attenuated. Next, a Fast Fourier Transform is performed on the processed window signal to obtain the amplitude distribution of the signal at different frequencies within the sliding window. The frequency point with the highest amplitude is found from this amplitude distribution as the instantaneous frequency corresponding to the center time of the current sliding window. Each window is processed sequentially to obtain a sequence of instantaneous frequencies arranged in chronological order, denoted as {f1, f2, ..., f...}. n }, where n is the total number of sliding windows, f n The instantaneous frequency values ​​corresponding to n sliding windows are represented; the center time of each window is used as the abscissa and the corresponding instantaneous frequency is used as the ordinate. Points are plotted in a plane rectangular coordinate system and the points are connected in sequence to form the frequency offset trajectory curve F(t); Based on the instantaneous frequency sequence, the rate of change vf of the electrical signal frequency during the abrupt change period is calculated using the following formula: ； Where Δt is the time interval between adjacent sliding windows, f k and f k+1 These represent the instantaneous frequency values ​​corresponding to the k-th and (k+1)-th sliding windows, respectively; simultaneously, the frequency drift range Δf = max(f1, f2, ..., f n )-min(f1,f2,…,f n ), which represents the maximum fluctuation width of the frequency during the mutation period.

4. The intelligent identification method for distribution network fault types based on multi-parameter analysis according to claim 3, characterized in that: Specifically, S200 also includes: Extract a segment of steady-state data collected by the fault recording device under normal operating conditions of the distribution network, with a data duration of not less than a preset duration threshold; for each phase voltage or current signal in the steady-state data, first calculate the average value of the entire signal segment, and subtract the average value from each sample value; then, perform a fast Fourier transform on the processed signal to obtain the amplitude results of the signal at each frequency point; plot the amplitude distribution curve of the amplitude changing with frequency using frequency as the abscissa and the corresponding amplitude as the ordinate; find the maximum amplitude in the entire frequency range on the amplitude distribution curve, and denot it as Amax; Set the amplitude detection threshold Ath, Ath = Amax × η; where η is the adjustment coefficient, 0.1 < η < 0.3; starting from the lowest frequency, scan the amplitude curve along the frequency increase direction. When the amplitude of all frequency points in a continuous frequency range is higher than Ath, mark the frequency range as a candidate frequency band and record the start and end frequencies of the candidate frequency band. For each candidate frequency band, the steady-state data is divided into multiple sub-segments according to a preset number of segments. The aforementioned Fast Fourier Transform and candidate frequency band extraction process is performed on each sub-segment. If a candidate frequency band appears in all sub-segments, and the bandwidth variation of the candidate frequency band among the sub-segments does not exceed a preset bandwidth ratio threshold, and the ratio of the maximum amplitude to the minimum amplitude within the frequency band is less than a preset amplitude ratio threshold, then the frequency band is determined to be an inherent interference frequency band, denoted as the interval set [L]. j U j ] m j=1 Where m is the number of inherent interference frequency bands identified, and L j and U j These are the upper and lower limits of the j-th inherent interference frequency band, respectively.

5. The intelligent identification method for distribution network fault types based on multi-parameter analysis according to claim 4, characterized in that: Specifically, S200 also includes: Based on the frequency change rate, frequency drift range, and overlap, a discriminant value λ is generated to indicate that this abrupt change period belongs to an arc-type abrupt change. The calculation formula is as follows: ； Wherein, vf0 is the baseline value of the frequency change rate of the historical arc sample, which is obtained from the statistics of confirmed intermittent arc fault data; Δf0 is the baseline value of the frequency drift range of the historical arc sample, which is also obtained from the statistics of historical fault data; w1 is the weight corresponding to the frequency change rate, w2 is the weight corresponding to the degree of overlap between the frequency and the interference frequency band, w3 is the weight corresponding to the frequency drift range, and w1+w2+w3=1.

6. The intelligent identification method for distribution network fault types based on multi-parameter analysis according to claim 5, characterized in that: Step S300 specifically includes the following steps: Step S301: For each marked signal abrupt change period, calculate the following five basic characteristic quantities: the duration from the start time to the end time, the rate of change of the electrical signal frequency obtained from step S202, the difference between the voltage waveform and the standard sine waveform, the maximum difference ratio between each phase voltage, and the voltage recovery rate; wherein, the difference between the voltage waveform and the standard sine waveform is obtained by dividing the average of the point-by-point deviations of the actual voltage sample value and the standard sine wave of the same frequency and amplitude by the rated voltage amplitude; the maximum difference ratio between each phase voltage is obtained by dividing the maximum absolute value of the difference between any two phases of the three-phase voltage effective value by the rated phase voltage; the voltage recovery rate is the average rate of change of the voltage within the time required for the voltage to recover to more than 95% of the rated value from the end of the abrupt change; Step S302: Establish the normal waveform range and interference value range for each characteristic quantity in advance by using historical normal operation waveform data and interference event statistics of the distribution network; the interference event is an electrical quantity disturbance event caused by load switching, switching operation, operation of power electronic equipment and distributed power source fluctuations in the distribution network under fault-free conditions; If all basic characteristic quantities fall within the normal recording range, the electrical signal segment during the abrupt change period is classified as a normal operation component; if all basic characteristic quantities fall within the interference range, it is classified as an interference component; if any basic characteristic quantity exceeds the upper limit of the interference range, the electrical signal segment during the abrupt change period is identified as a suspected fault component; record the basic characteristic quantities and their actual measured values ​​corresponding to all electrical signal segments identified as suspected fault components. Step S303: For each basic characteristic quantity that is judged as a suspected fault, take its average value in historical interference events as the interference benchmark value; if a basic characteristic quantity is not judged as a suspected fault, its difference is recorded as zero; otherwise, calculate the absolute difference between the actual measured value of the basic characteristic quantity and the interference benchmark value, and then divide it by the maximum difference that the corresponding basic characteristic quantity has in historical fault events to obtain the normalized difference value; add up the normalized differences of all basic characteristic quantities to obtain the total difference score, denoted as D, then the fault characteristic label value μ=min(D / 5,1), D∈[0,5], μ∈[0,1].

7. The intelligent identification method for distribution network fault types based on multi-parameter analysis according to claim 6, characterized in that: In step S400, several groups of confirmed intermittent arc faults, phase-to-phase short-circuit faults, and conventional voltage disturbance events are selected from the historical waveform data of the distribution network. The arc-induced sudden change discrimination mark value and fault feature mark value corresponding to each group of historical events are calculated to form template point sets corresponding to the three types of events. For the sudden change period to be identified, a feature point (λ,μ) is constructed based on the calculated λ and μ. The Euclidean distance between the feature point (λ,μ) and each template point in the three types of template point sets is calculated. The category to which the template point with the smallest Euclidean distance belongs is selected as the fault type identification result of the current signal sudden change, and the corresponding alarm prompt is output synchronously.

Images