A cable intermediate joint fault early warning method based on impedance monitoring

Abstract

The application discloses a cable intermediate joint fault early warning method based on impedance monitoring and relates to the technical field of online monitoring of power equipment. In the initial learning period after the cable intermediate joint is put into operation, the method collects a wideband input impedance spectrum, calculates the impedance amplitude ratio of different frequency points, obtains a temperature stability score and an insulation deterioration sensitivity score based on the variation coefficient of the ratio sequence and the disturbance injection or the look-up table method, and screens out the optimal characteristic frequency point pair through weighting. In the online monitoring stage, the current impedance ratio is calculated, the short-time change rate and the long-time deviation rate are calculated through short-time window and long-time window respectively, the abnormality is determined when the short-time change rate is continuously more than the first threshold value for multiple times and the long-time deviation rate is more than the second threshold value and is of the same sign, the comprehensive early warning index is calculated, and the early warning level is output. The application solves the problem that the existing impedance monitoring method is misjudged due to temperature drift, and realizes the online early warning of the insulation deterioration of the cable intermediate joint.

Description

Technical Field

[0001] This invention relates to the field of online monitoring technology for power equipment, and more specifically, to a method for early warning of cable joint faults based on impedance monitoring. Background Technology

[0002] Underground cables are widely used in urban power grids, and cable joints are the weakest links in the lines. Insulation dampness or interface degradation are the main causes of cable failures. Broadband impedance spectroscopy can reflect changes in the dielectric properties of cable insulation and has been used for cable condition assessment. However, fluctuations in ambient temperature can cause significant changes in the dielectric constant and loss factor of cable insulation materials, leading to impedance spectrum drift. This drift is often greater than the impedance change caused by early insulation degradation, making it difficult for online monitoring based on impedance spectra to distinguish between temperature interference and actual degradation.

[0003] In the prior art, CN119269961A discloses a fault early warning method for explosion-proof cable joints using multiple parameters of temperature and electric field. This method simultaneously collects temperature and electric field signals and performs early warning through multi-parameter correlation calculation. However, it still requires the additional installation of a temperature sensor, and temperature is only used as an independent parameter in the judgment. CN119355437A discloses a method and system for online monitoring and fault early warning of underground cables, which uses multi-modal data fusion and a cloud computing platform for fault identification, but does not solve the temperature drift problem in impedance monitoring. CN110794271B discloses a method for locating and diagnosing moisture-induced joints of power cables based on input impedance spectrum. This method requires offline acquisition of the reference impedance spectrum under healthy conditions and is not suitable for online monitoring scenarios.

[0004] Therefore, there is an urgent need for a method for monitoring the impedance of cable joints that does not require temperature sensors, does not rely on offline references, and can effectively eliminate temperature interference. Summary of the Invention

[0005] To overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a cable intermediate joint fault early warning method based on impedance monitoring. The technical problem to be solved is that in the online impedance monitoring of cable intermediate joints, it is difficult to distinguish between impedance drift caused by ambient temperature fluctuations and impedance changes caused by insulation degradation, which leads to a high misjudgment rate of existing methods or the need to rely on additional temperature sensors and offline reference spectra, thus failing to achieve reliable online early warning.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for early warning of cable joint faults based on impedance monitoring includes the following steps: S1: During the initial learning period after the cable joint is put into operation, a low-amplitude sweep frequency voltage signal is injected into the cable joint at fixed time intervals. The broadband input impedance response at the joint is collected, and multiple impedance amplitude sequences are obtained through Fourier transform. Impedance reference data of the cable joint under normal operating conditions is established, covering the temperature fluctuation range within a complete daily temperature difference cycle.

[0007] S2: For each pair of candidate frequency points, calculate the ratio of the impedance amplitudes at the two frequency points in each impedance amplitude sequence. This forms a ratio sequence of the frequency point pairs during the initial learning period. ; calculate coefficient of variation and will As a score for temperature stability The smaller the coefficient of variation, the lower the dispersion of the ratio sequence, indicating that the ratio is less sensitive to temperature changes.

[0008] A small perturbation harmonic signal is superimposed simultaneously with the frequency sweep voltage signal injection, and the impedance ratio of the frequency pair before and after superposition is measured. and ,Will As a score for sensitivity to insulation degradation Small perturbations simulate minute changes in the dielectric properties of the insulation, reflecting the sensitivity of that frequency point to degradation.

[0009] S3: Calculate the weighted sum Based on this weighted sum, multiple optimal feature frequency pairs are selected from all candidate frequency pairs, among which... and The weighting coefficients are used to weight and balance temperature stability and degradation sensitivity. The selected frequency pairs can remain stable under temperature fluctuations and effectively respond to insulation degradation.

[0010] S4: During the online monitoring phase after the initial learning period, after each acquisition of a new broadband input impedance spectrum, calculate the current impedance ratio for each optimal characteristic frequency pair. The current ratio is used to compare with historical data to capture trends in impedance changes.

[0011] S5: Calculate the short-time rate of change for each optimal feature frequency pair using a short-time sliding window. The long-term offset rate of each optimal feature frequency point pair is calculated using a long-term sliding window. Short-term rate of change reflects recent fluctuations, while long-term deviation reflects cumulative changes. The combination of the two is used to distinguish between temperature fluctuations and deterioration trends.

[0012] S6: When a certain optimal feature frequency point pair Multiple consecutive occurrences of the same number exceeding the first threshold and Exceeding the second threshold and and When the frequencies are the same, an insulation degradation anomaly is determined to have occurred at that frequency point. Repeated occurrences of the same frequency eliminate random noise, and directional consistency ensures that the short-term trend and the long-term cumulative direction are the same, which is consistent with the unidirectional gradual characteristic of insulation degradation.

[0013] S7: Calculate the comprehensive early warning index The number of optimal feature frequency point pairs that are judged as anomalous hour ,otherwise ,in For the first The long-term offset rate of anomalous frequency point pairs; according to The value outputs the fault warning level of the cable intermediate joint. The comprehensive warning index integrates information from multiple abnormal frequency points and outputs it in a graded manner to facilitate operation and maintenance decisions.

[0014] Furthermore, the initial learning period lasts for 24 hours, and the fixed time interval is 15 minutes or 30 minutes. The 24-hour period covers the complete cycle of daily temperature differences, and the 15-minute or 30-minute intervals balance data density and storage overhead.

[0015] Furthermore, the wideband input impedance spectrum has a frequency range of 10kHz to 30MHz, which is divided into 128 discrete frequency points. This frequency band covers the dielectric relaxation characteristic region of cable insulation, and the 128 points ensure frequency resolution. The candidate frequency point pairs are those where the low-frequency value is less than the high-frequency value, and the ratio of the high-frequency value to the low-frequency value is greater than or equal to 2. This multiple condition ensures that the dielectric relaxation characteristics of the two frequency points have a sufficient difference, reducing the temperature sensitivity of the ratio.

[0016] Furthermore, the coefficient of variation ,in Ratio sequence standard deviation Ratio sequence The mean of the values. The coefficient of variation is dimensionless and is used to compare the stability of different frequency pairs under temperature fluctuations.

[0017] Furthermore, the amplitude of the small disturbance harmonic signal does not exceed 5% of the amplitude of the swept frequency voltage signal. This limitation ensures that the disturbance does not affect the normal operation of the cable, while being sufficient to induce a measurable impedance change.

[0018] Furthermore, the weighting coefficients and satisfy and Temperature stability is given a higher weight than sensitivity, prioritizing resistance to temperature-related disturbances. The number of optimal feature frequency pairs selected is between 3 and 8. This number strikes a balance between computational efficiency and redundancy.

[0019] Furthermore, the short-term sliding window has a duration of 1 hour, and the long-term sliding window has a duration of 24 hours. The 1-hour window captures short-cycle temperature fluctuations, while the 24-hour window reflects daily cumulative changes. Short-term rate of change. ,in This is the median of the impedance ratios within a short-time sliding window immediately preceding the current measurement time. The median is insensitive to outliers, while the short-time rate of change reflects the percentage of recent relative change.

[0020] Long-term offset ,in This represents the median of the impedance ratios within the current long-time sliding window, including the current measurement time. This represents the median impedance ratio at that frequency point during the initial learning period. The long-term offset reflects the cumulative drift since the learning period began.

[0021] Furthermore, the "continuous multiple times" refers to 3 or 4 consecutive times, and the first threshold... Second threshold Instantaneous fluctuations are eliminated multiple times, and the threshold is set based on field experience to balance sensitivity and noise immunity.

[0022] Furthermore, according to The specific method for outputting the fault warning level is as follows: when Output normal level when Attention level when outputting, Output warning level in real time, when The system outputs the hazard level in real time. The level is related to the degree of deviation from the long-term offset rate; the higher the level, the more severe the degradation.

[0023] Statistically analyze the low-frequency values ​​of all abnormal frequency pairs. If more than 80% of the abnormal frequency pairs have low-frequency values ​​less than 500kHz, output a moisture risk warning; otherwise, output an interface degradation risk warning. Moisture primarily affects the low-frequency band, while interface degradation primarily affects the high-frequency band. Frequency band distribution is used to distinguish fault types.

[0024] Furthermore, the method also includes: during the online monitoring phase, recalculating every 7 days using all impedance spectrum data collected in the past 7 days. and the recalculated Replace the initial learning period Used for subsequent long-term offset Calculation. Regularly updating the baseline can track the normal aging drift of cable insulation and avoid misjudgments caused by baseline deviation.

[0025] The technical effects and advantages of this invention are as follows: (1) In the initial learning period after the cable joint is put into operation, the present invention collects broadband input impedance spectrum and calculates the ratio of impedance amplitude at different frequency points. The temperature stability score is constructed using the coefficient of variation of the ratio sequence. At the same time, the relative change of impedance ratio is measured by superimposing small disturbance harmonic signals or the insulation degradation sensitivity score is obtained by looking up a table. The optimal characteristic frequency point pair is adaptively selected from all candidate frequency point pairs by weighted comprehensive scoring. This selection process only relies on the impedance data collected during the initial learning period. It does not require the prior acquisition of the reference impedance spectrum under healthy conditions, nor does it require the installation of temperature sensors at the cable joint.

[0026] (2) In the online monitoring phase, this invention calculates the short-time change rate and long-time offset rate for each optimal characteristic frequency pair. The short-time change rate is based on the median impedance ratio within the short-time sliding window immediately preceding the current measurement time, reflecting the fluctuation trend over a recent period. The long-time offset rate is based on the median impedance ratio within the initial learning period, reflecting the cumulative offset since commissioning. By setting a first threshold and a second threshold, when the short-time change rate exceeds the first threshold multiple times consecutively and the long-time offset rate exceeds the second threshold with the same sign, an abnormal degradation is determined. This dual-window differential judgment logic utilizes the waveform characteristic difference between the unidirectional gradual change caused by insulation degradation and the bidirectional alternating change caused by temperature fluctuations to separate temperature interference from the impedance measurement signal.

[0027] (3) This invention calculates a comprehensive early warning index, which is the average of the ratio of the long-term offset rate of all optimal feature frequency point pairs judged as abnormal to the second threshold. Based on the numerical range of this index, the early warning level is divided into four levels: normal, attention, warning, and danger. Simultaneously, the distribution of low-frequency points in all abnormal frequency point pairs is statistically analyzed. When the proportion of low-frequency points less than 500kHz exceeds 80%, a moisture risk warning is output; otherwise, an interface degradation risk warning is output. This graded early warning mechanism and fault type differentiation function provide maintenance personnel with different levels of handling suggestions and preliminary judgment criteria for fault types.

[0028] (4) In the online monitoring phase, this invention sets up a dynamic baseline update mechanism. Every fixed period ranging from 7 to 30 days, all impedance spectrum data collected in the previous period are taken with the current time as the endpoint. The long-term sliding window baseline median of each optimal characteristic frequency point pair is recalculated, and the original baseline median is replaced by the newly calculated baseline median for subsequent long-term offset calculation. This update mechanism enables the baseline to be adjusted to follow the normal aging drift of the cable insulation material, reducing the cumulative error of offset caused by the long-term invariance of the baseline. Attached Figure Description

[0029] Figure 1 This is a flowchart illustrating the method execution of the present invention. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0031] Example 1 As attached Figure 1 This paper presents a fault early warning method for cable joints based on impedance monitoring. This method does not rely on an external temperature sensor. It utilizes the characteristic that the impedance ratio at different frequency points in the dielectric relaxation properties of cable insulation is insensitive to temperature changes but sensitive to insulation degradation. Through adaptive selection of characteristic frequency point pairs and dual-window differential early warning logic, online early warning of the insulation status of cable joints is achieved. The following detailed explanation of this method is provided in conjunction with specific implementation steps.

[0032] I. Data Collection During the Initial Learning Period After a new cable joint is put into operation and its insulation condition is confirmed to be good through offline withstand voltage testing and partial discharge detection, the initial learning period is initiated.

[0033] The initial learning period begins when the cable joint is first energized, and its duration is set to cover at least one complete diurnal temperature range cycle, such as 24 hours. During the initial learning period, measurements are repeated at fixed time intervals. In one implementation, this fixed time interval is 15 minutes; in another implementation, it is 30 minutes.

[0034] During each measurement, a low-amplitude sweep voltage signal is injected into the cable joint via a non-invasive coupling method (e.g., a high-frequency current transformer).

[0035] The swept-frequency voltage signal is connected in series to the grounding circuit of the cable line through the primary side of a high-frequency current transformer, and the secondary side of the transformer is connected to a signal generator. The amplitude of the swept-frequency voltage signal is set to a value that does not interfere with the normal operation of the line, such as 5V to 10V; the swept-frequency range is 10kHz to 30MHz, using logarithmic intervals. To ensure impedance measurement accuracy, the amplitude output accuracy of the signal generator should be better than ±1%, the sampling rate of the sampling device should be no less than 100MHz, and the analog-to-digital conversion resolution should be no less than 12 bits.

[0036] The frequency range is divided into multiple discrete frequency points. In one implementation, the number of discrete frequency points is 128, denoted as . At each frequency point, the voltage response time-domain signal at both ends of the cable connector is acquired, and a fast Fourier transform is performed on the time-domain signal to obtain the frequency-domain voltage spectrum.

[0037] Simultaneously, the current in the injection circuit is measured to obtain the injection current spectrum. The impedance amplitude at that frequency point is obtained by dividing the frequency domain voltage spectrum by the injection current spectrum. To eliminate random noise, the measurement at each frequency point can be repeated four times and the average value is taken.

[0038] After a complete frequency sweep, a sequence of impedance amplitudes covering all discrete frequency points is obtained. Multiple impedance amplitude sequences are obtained during the initial learning period, each represented as follows: ,in The measurement ordinal number ranges from 1 to the total number of measurements during the initial learning period. .

[0039] If the number of measurements is insufficient during the initial learning period due to power outages or communication interruptions. Then extend the learning period until complete data is collected. Groups of data. Each group of data is accompanied by a timestamp for subsequent window alignment.

[0040] II. Adaptive Feature Frequency Pair Filtering From all discrete frequency points, select all that satisfy and Frequency point pairs As candidate frequency point pairs, among For a preset multiple greater than or equal to 2, for example The generation of candidate frequency pairs iterates through all combinations that meet the above conditions, without missing any pair. This multiple condition ensures that the dielectric relaxation characteristics of the two frequency points are sufficiently different, so that the ratio is less sensitive to temperature.

[0041] For each candidate frequency pair, the ratio of the impedance magnitudes of the two frequency points in each pair is calculated using the entire impedance magnitude sequence during the initial learning period: ; This yields the ratio sequence of the frequency point pair. Before calculating the mean and standard deviation, you can... Outlier removal: using 3 The criterion is to remove points that deviate from the mean by more than 3 standard deviations, and the remaining number of data points is denoted as n'. Subsequent calculations are based on n' valid data points.

[0042] Calculate the mean of the sequence. and standard deviation Calculate the coefficient of variation. Define temperature stability score .like If it is zero (which theoretically would not happen), then set it to zero. For example, a constant that is much larger than the scores of other candidate pairs. .

[0043] Evaluate the sensitivity of this frequency point to insulation degradation (such as moisture or interface degradation). As one implementation method, a perturbation injection method is used: During the initial learning period, a small perturbation harmonic signal is superimposed on the normal frequency sweep signal. The frequency of this harmonic signal is set to a fixed value (e.g., 1MHz) within the frequency sweep range, and the amplitude is set to the amplitude of the frequency sweep voltage signal. ,in The value ranges from 1% to 5%.

[0044] Measure the impedance ratio of the pair of frequencies before and after the superimposed harmonic signal. The ratio of the summed impedances Calculate single-shot sensitivity .

[0045] Repeat the measurement multiple times (e.g., 3 times), and take the arithmetic mean as the sensitivity score. When disturbance signals are superimposed, keep the parameters of the sweep frequency signal unchanged and only add a single-frequency sine wave.

[0046] As another implementation method, a lookup table method is used: a frequency band sensitivity weight table is established in advance based on the offline test results of the dielectric spectrum characteristics of the cable insulation material, the frequency range is divided into several continuous frequency bands, and each frequency band is assigned a sensitivity weight.

[0047] The weighting method is as follows: Impedance tests are performed on cable joint samples with known different degrees of degradation. The correlation coefficient between the impedance change rate of each frequency band and the overall degree of degradation is calculated, and this correlation coefficient is used as the weight. For candidate frequency points... Query each Weight of the frequency band and Weight of the frequency band Calculate sensitivity score .like and If they are in the same frequency band, then It equals the weight of that frequency band.

[0048] For each candidate frequency pair, calculate the weighted sum. ,in and For the weighting coefficients, satisfying and Pair candidate frequency points with... Sort the values ​​from largest to smallest and select the first few. 1 is selected as the optimal feature frequency point pair, among which The value range is from 3 to 8.

[0049] If the total number of candidate frequency point pairs is insufficient If the optimal feature frequency pair is selected, all candidate pairs are taken. The selected optimal feature frequency pair is used continuously in subsequent online monitoring until the next learning period (such as after equipment replacement) when it is re-selected.

[0050] III. Feature Calculation in the Online Monitoring Phase After the initial learning period, the online monitoring phase begins. Each measurement performs the same frequency sweep operation as during the initial learning period. For each optimal feature frequency pair... Calculate the impedance ratio at the current moment: ; in and These are the frequency points in the current measurement. and The system measures the impedance amplitude. If the impedance amplitude at a certain frequency point in a measurement is invalid (e.g., exceeding the range or with a low signal-to-noise ratio), the measurement is discarded, and the system waits for the next measurement interval to re-acquire the data. If more than three consecutive measurements are discarded, the system should issue a measurement anomaly warning.

[0051] IV. Calculation of Dual-Window Difference Characteristics Set the duration of the short-time sliding window and the duration of the long sliding window As one implementation method, Hour, The number of measurements for each window is determined based on the measurement interval. The number of measurements for a window is obtained by dividing the time length by the measurement interval and rounding down. For example, when the measurement interval is 15 minutes... Corresponding to 4 measurements; with a measurement interval of 30 minutes, This corresponds to two measurements. The window boundary alignment is as follows: based on the timestamp of the current measurement moment, trace back... and The window boundary falls on the timestamp of the most recent measurement, and no interpolation is performed.

[0052] Define the current measurement time as the [number]th [time]. The first measurement during the online monitoring phase. Set short-time rate of change Starting from the second measurement ( The short-time sliding window is defined as the window immediately preceding the current measurement time, with a time length of [missing information]. The time period (excluding the current measurement).

[0053] If the number of valid measurements within this period is 0, then the most recent available measurement will be used. The value (initially 0). The median of all measured impedance ratios within this window is denoted as . The median is calculated as follows: The ratios within the window are sorted from smallest to largest. If the number of measurements is odd, the middle value is taken; if the number of measurements is even, the arithmetic mean of the two middle values ​​is taken. The short-time rate of change at the current moment is calculated using the following formula: ; Define a long-term sliding window as one that includes the current measurement time and has a duration of [value missing]. The time period (i.e., from the current measurement time before) (Time length up to the current measurement time). If the number of valid measurements within this time period is 0, then the median of the initial learning period is used. As The median of all measured impedance ratios within this window is denoted as . The median is calculated using the same rules as above. The median of the impedance ratio at that frequency point during the initial learning period is defined as... (Calculation method is the same) The long-term offset rate is calculated using the following formula: ; V. Judgment of Abnormal Insulation Deterioration Set the first threshold Second threshold ,in The value ranges from 1% to 3%. The value ranges from 3% to 6%. For each optimal characteristic frequency pair, the short-time rate of change of each measurement is continuously monitored. .

[0054] When a continuous Second measurement The values ​​are all greater than or both are less than When short-term trends are deemed abnormal, among which... The preset number of consecutive counts, which can be 3 or 4. The number of consecutive counts starts from the first time the direction condition is met. If the opposite direction occurs or the absolute value does not exceed a certain threshold, the count will be stopped. If the count is zeroed out, the counter will be reset to zero.

[0055] Simultaneously check the long-term offset at the current moment. Does it meet the requirements? ,and The sign of the above consecutive The values ​​have the same sign.

[0056] When both of the above conditions are met, it is determined that the insulation medium of the corresponding cable joint at that frequency point has deteriorated abnormally.

[0057] For multiple frequency point pairs that are simultaneously identified as abnormal, their frequency band distribution consistency can be further analyzed: if the low-frequency points of all abnormal point pairs... If all degradation types are concentrated in the same frequency band, the reliability of the degradation type judgment is higher.

[0058] VI. Calculation and Level Output of Comprehensive Early Warning Index Let the number of optimal feature frequency point pairs that are currently judged as anomalies be . Comprehensive Early Warning Index Calculate using the following formula: like ,but ; like ,but ,in For the first Long-term offset (percentage value) of anomalous frequency point pairs. This is the second threshold (percentage value).

[0059] according to Output the fault warning level: like Output the "Normal" level; like Output the "attention" level; like Output the "warning" level; like Output the "danger" level.

[0060] At the same time, the frequency values ​​of all abnormal frequency points paired with low- and mid-frequency points (i.e., If more than 80% of the abnormal frequency pairs have low-frequency values ​​less than a preset frequency threshold. (For example, 500kHz) will output a "moisture risk warning"; Otherwise, output a "UI degradation risk warning". The denominator for this percentage calculation is the total number of abnormal frequency point pairs. ,like No risk warning will be output.

[0061] For mixed degradation (where both moisture and interface degradation characteristics are present), two types of prompts can be output simultaneously.

[0062] VII. Dynamic Baseline Update During the online monitoring phase, updates are performed every preset interval. Taking the current moment as the endpoint, take the past. For all impedance spectrum data collected within the day, the long-time sliding window baseline median for each optimal characteristic frequency pair was recalculated. (Calculation method is the same) ),Will Replace the original Used for subsequent long-term offset The calculation.

[0063] After the baseline is updated, historical data within the long-term sliding window remains unchanged. Update cycle Timing begins from the online monitoring phase, with the first update occurring on the [number]th day after commissioning. Execute at the end of the day. Update cycle. The value ranges from 7 days to 30 days.

[0064] To avoid abrupt changes in the warning level due to sudden baseline shifts, a smooth transition can be adopted: the new baseline... With the old baseline Weighted by a certain proportion, for example Gradually adjust to The smoothing cycle can be set to 3 updates.

[0065] The broadband impedance measurement technique, fast Fourier transform, median calculation, and standard deviation calculation used in this method are all existing and well-known techniques. Its inputs are the time-domain voltage response signal and the injected current spectrum, and its outputs are the impedance amplitudes at various frequency points. These fundamental techniques are used only as tools in this method, and their specific implementation does not constitute a limitation on the invention.

[0066] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any equivalent substitutions or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for early warning of cable joint faults based on impedance monitoring, characterized in that, Includes the following steps: S1: During the initial learning period after the cable intermediate joint is put into operation, a low amplitude sweep frequency voltage signal is injected into the cable intermediate joint at fixed time intervals, and the wideband input impedance response at the joint is collected. Multiple impedance amplitude sequences are obtained through Fourier transform. S2: For each pair of candidate frequency points, calculate the ratio of the impedance amplitudes at the two frequency points in each impedance amplitude sequence. This forms a ratio sequence of the frequency point pairs during the initial learning period. ; calculate coefficient of variation and will As a score for temperature stability ; A small perturbation harmonic signal is superimposed simultaneously with the frequency sweep voltage signal injection, and the impedance ratio of the frequency pair before and after superposition is measured. and ,Will As a score for sensitivity to insulation degradation ; S3: Calculate the weighted sum Based on this weighted sum, multiple optimal feature frequency pairs are selected from all candidate frequency pairs, among which... and These are the weighting coefficients; S4: During the online monitoring phase after the initial learning period, after each acquisition of a new broadband input impedance spectrum, calculate the current impedance ratio for each optimal characteristic frequency pair. ; S5: Calculate the short-time rate of change for each optimal feature frequency pair using a short-time sliding window. The long-term offset rate of each optimal feature frequency point pair is calculated using a long-term sliding window. ; S6: When a certain optimal feature frequency pair Multiple consecutive occurrences of the same number exceeding the first threshold and Exceeding the second threshold and and When the same signal is present, it is determined that the frequency point pair has an insulation degradation abnormality; S7: Calculate the comprehensive early warning index The number of optimal feature frequency point pairs that are judged as anomalous hour ,otherwise ,in For the first The long-term offset rate of a pair of anomalous frequency points; according to The value outputs the fault warning level of the cable intermediate joint.

2. The cable joint fault early warning method based on impedance monitoring according to claim 1, characterized in that, The initial learning period lasts for 24 hours, and the fixed time interval is 15 minutes or 30 minutes.

3. The cable joint fault early warning method based on impedance monitoring according to claim 1, characterized in that, The frequency range of the wideband input impedance spectrum is from 10kHz to 30MHz, and this frequency range is divided into 128 discrete frequency points; the candidate frequency point pair is a frequency point pair that satisfies the condition that the low frequency point value is less than the high frequency point value and the ratio of the high frequency point value to the low frequency point value is greater than or equal to 2.

4. The cable joint fault early warning method based on impedance monitoring according to claim 1, characterized in that, The coefficient of variation ,in Ratio sequence standard deviation Ratio sequence The mean.

5. The cable joint fault early warning method based on impedance monitoring according to claim 1, characterized in that, The amplitude of the small perturbation harmonic signal does not exceed 5% of the amplitude of the swept frequency voltage signal.

6. The cable joint fault early warning method based on impedance monitoring according to claim 1, characterized in that, The weighting coefficient and satisfy and The number of optimal feature frequency point pairs selected is 3 to 8.

7. The cable joint fault early warning method based on impedance monitoring according to claim 1, characterized in that, The duration of the short-term sliding window is 1 hour, and the duration of the long-term sliding window is 24 hours. ,in This is the median of the impedance ratio within a short sliding window immediately preceding the current measurement time. ,in This represents the median of the impedance ratios within the current long-time sliding window, including the current measurement time. This represents the median impedance ratio at that frequency point during the initial learning period.

8. The cable joint fault early warning method based on impedance monitoring according to claim 1, characterized in that, The "consecutive times" refers to 3 or 4 consecutive times. , .

9. The cable joint fault early warning method based on impedance monitoring according to claim 1, characterized in that, According to The specific method for outputting the fault warning level is as follows: when Output normal level when Attention level when outputting, Output warning level in real time, when Output the danger level in real time; count the low frequency values ​​of all abnormal frequency point pairs. If the low frequency value of more than 80% of the abnormal frequency point pairs is less than 500kHz, output a moisture risk warning; otherwise, output an interface degradation risk warning.

10. The cable joint fault early warning method based on impedance monitoring according to claim 7, characterized in that, Also includes: During the online monitoring phase, the impedance spectrum data collected in the past 7 days is recalculated every 7 days. and the recalculated Replace the initial learning period For subsequent calculate.

Images