Accumulated cable early fault enhanced recognition device and enhanced recognition method thereof

Abstract

The invention discloses an enhanced recognition device and an enhanced recognition method for an early fault of a waterlogged cable. The enhanced identification device is located on a grounding wire between a copper shielding layer and a grounding point, the current acquisition device adopts a current transformer (CT) wound on the grounding wire, and each cable joint is additionally provided with a repetitive signal counter besides a fault current measuring device. A fault signal from a fault current measuring device is uploaded to an information node through a detection point, and structural support is provided for repetitive detection of the fault signal. The enhanced identification method comprises five steps of enhancement processing of a current signal APE low-rank algorithm, timing threshold detection, continuity detection, harmonic sampling, harmonic content comparison and repeatability detection, and not only realizes identification of early faults of the cable joint, but also prevents false alarms caused by fault signals of other equipment. According to the enhanced identification logic provided by the invention, other interference signals can be effectively screened out, and the total accuracy is kept at a relatively high level.

Description

Technical Field

[0001] The present invention belongs to the field of early fault identification of cable joints, and in particular relates to an enhanced identification device for early faults of waterlogged cables and an enhanced identification method thereof. Background Art

[0002] With the rapid development of social economy and the accelerated advancement of urbanization, my country's power infrastructure, especially the distribution network cable system, has ushered in unprecedented development opportunities. However, due to the complex structure and changing operating environment of distribution network cables, as well as the various types of defects and their hidden nature, traditional detection methods are costly and inefficient, making it difficult to effectively capture early fault signals, resulting in frequent faults, which seriously restricts the stability and reliability of urban distribution networks.

[0003] Although domestic and foreign scholars have conducted research in the field of early fault and cable identification, such as using the LIME algorithm to capture the time domain characteristics of high-resistance grounding faults in distribution lines; using the Mahalanobis distance and support vector data to establish a cable early fault identification method; constructing a method based on harmonic correction and additional approximate correction to monitor cable early faults; and using the multiple correspondence analysis method to conduct correlation analysis and evaluation of many factors leading to cable faults; improving the existing identification method based on waveform wavelet domain features to achieve the identification of early cable faults. These studies have improved the accuracy and speed of fault identification. However, it cannot be ignored that there are still two core bottlenecks in current research: First, the deep correlation between early fault characteristics and fault development mechanisms has not been fully revealed, resulting in limited interpretability and accuracy of identification strategies; second, in the process of cable current signal acquisition, the widespread existence of multiple interference sources such as discrete spectrum interference from communication systems, pulse noise from electrical equipment, and electrical noise from the detection circuit itself seriously interferes with signal quality, increases the complexity and difficulty of signal processing, and greatly reduces the effectiveness of existing methods in practical applications.

[0004] Therefore, there is an urgent need for an innovative cable early fault identification and interference suppression technology, which can not only deeply analyze the intrinsic connection between fault characteristics and fault evolution process, improve the accuracy and explainability of identification, but also effectively respond to and suppress the influence of various interference sources on current signals, ensure signal quality, and provide solid technical support for the safe and efficient operation and maintenance of urban distribution network cables. Summary of the invention

[0005] In order to solve the above problems, the present invention proposes an enhanced identification device for early faults of waterlogged cables and an enhanced identification method thereof, which realizes efficient identification of early faults through refined fault feature extraction and enhanced processing.

[0006] The technical solution of the present invention is as follows:

[0007] On the one hand, the present invention provides a device for enhancing early fault identification of waterlogged cables, which is located on the grounding wire between the copper shielding layer and the grounding point, and is characterized in that it includes:

[0008] The current collection device uses a current transformer (CT) wound on the ground wire to collect the current signal in the cable;

[0009] A repeatability signal counter, installed at each cable connector, is used to record the repeatability of the fault signal from the current acquisition device;

[0010] An information node is used to receive a fault signal from a current acquisition device and support repeatability detection of the fault signal; wherein the fault signal includes an initial stage and a main stage, the acquisition period of the initial stage is set to 0.5 seconds from the appearance of the first arc discharge waveform, and the acquisition period of the main stage is 10 seconds after the end of the initial stage.

[0011] Preferably, it also includes a fault current measuring device, which is used to upload the collected fault signal to the information node through the detection point, so as to provide structural support for the repeatability detection of the fault signal.

[0012] On the other hand, the present invention also provides a method for enhancing early fault identification of waterlogged cables, which is characterized in that it includes the following steps:

[0013] S1. Enhanced processing of current signal APE low-rank algorithm:

[0014] S2. Timing threshold detection: when the signal amplitude of the current signal flowing through the sample in the initial stage is within the preset threshold range, the recording of the suspected fault signal is started;

[0015] S3. Continuity detection: During the signal acquisition process, if there is a continuous signal whose peak current amplitude is always lower than the preset threshold, the signal acquisition is stopped;

[0016] S4. Harmonic sampling: Segment the signal that meets the timing threshold detection and waveform continuity detection, and calculate the harmonic energy and harmonic contribution of the two signal segments;

[0017] S5. Comparison of harmonic content: By comparing the harmonic energy and harmonic contribution of the two signals, it is preliminarily determined whether the signal belongs to an early fault;

[0018] S6. Repeatability detection: Perform repeatability detection on the initially identified early fault signals. When the repeatability detection count exceeds the preset threshold within a certain period of time, a higher level early fault alarm signal is issued.

[0019] Furthermore, the enhanced processing of the current signal APE low-rank algorithm in the enhanced recognition logic specifically includes.

[0020] a) Decompose the collected cable current signal into 6 sub-signals through EMD, and each sub-signal is expressed as:

[0021] s={s1,s2,...,s N}

[0022] b) For each sub-signal, select embedding dimension d=3 and time delay l=1, then embed each sub-signal IMF into a d-dimensional space to obtain a reconstructed vector:

[0023]

[0024] c) Reconstruct the vector Sort the elements in to get a new vector:

[0025]

[0026] Among them, j1, j2, ..., jd are The index of the element in .

[0027] d) For a d-dimensional reconstruction vector, there are d! sorting modes, and the probability of the i-th sorting mode appearing is recorded as The probability of each sorting mode occurring is calculated according to the following formula.

[0028]

[0029] e) Calculate the APE entropy value of each sub-signal according to the following formula.

[0030]

[0031] The enhanced processing of the current signal APE low-rank algorithm in the enhanced recognition logic specifically includes:

[0032] a) Cluster the six sub-signals into two clusters based on the clustering algorithm and APE entropy value.

[0033] b) Each sub-signal in the sub-signal cluster with high APE entropy value is enhanced by a low-rank matrix decomposition algorithm, specifically including:

[0034] Each sub-signal in the sub-signal cluster with high APE entropy value is represented as: x = {x1, x2, ..., x N}, and convert it into the observation matrix form X:

[0035]

[0036] Where L is the embedding dimension, L = 0.5*N, K = N-L+1.

[0037] The SVD decomposition result of X is calculated by the following formula.

[0038]

[0039] Where H represents the conjugate transpose, are the descending singular values ​​of X, d is the number of singular values, They represent the left and right singular vectors respectively.

[0040] Set the number of clusters k=3 and use the clustering algorithm to cluster the singular values ​​of X. Then calculate the average singular value of each cluster, regard the singular values ​​in the two clusters with larger average values ​​as being associated with the sub-signal and obtain the matrix rank r, which is the number of singular values ​​in the two larger clusters.

[0041] Calculate the i-th reconstruction coefficient according to the following formula

[0042]

[0043] in, I is the identity matrix, for The differential form of , Tr is the sum operation of the diagonal elements of a square matrix.

[0044] Compute the subsignal matrix using the reconstruction coefficients and singular value vectors As shown below.

[0045]

[0046] Finally, the enhanced sub-signal is obtained

[0047]

[0048] Each enhanced sub-signal in the sub-signal cluster with a high APE entropy value is added together to obtain an enhanced current signal.

[0049] Furthermore, the timing threshold detection specifically includes:

[0050] The detection device set on the grounding wire of the cable joint is in the normally open state. When the current signal I flowing through the sample f When the signal amplitude at the initial stage is within the threshold range as shown in the following formula, the recording of the suspected fault signal is started.

[0051]

[0052] I set1 <|I start |<I set2

[0053] In the formula, [I1,I2,…,I n ] The array is the amplitude of the current signal at each peak point. From the experimental phenomenon, we know that n is determined by the duration of the initial stage of the early fault and can be 5 to 20; I start is the current signal at the initial stage of the fault signal; I set1 To determine the lower limit of the current threshold for fault recording, I set2 The upper limit of the current threshold for determining whether fault recording can continue.

[0054] Furthermore, the waveform continuity detection specifically includes:

[0055] During the signal acquisition process, if the signal I f There is a continuous signal I f1 , if the peak current amplitude of the signal is always lower than the threshold, the signal acquisition is stopped. The specific detection method is shown in the following formula:

[0056]

[0057] I f1 =[I m ,I m+1 ,..,I m+c ]

[0058] |I f1 |<I set

[0059] In the formula, [I m ,I m+c ,…,I m+c ] is I f1 The amplitude of each peak point in the segment signal. m For I f1 The current amplitude at the initial point of the signal; I set is the lower limit of the current for determining whether the signal is continuous; c is the threshold for signal continuity detection, I f1 The signal duration is 0.01c.

[0060] Furthermore, the harmonic sampling specifically includes:

[0061] For signals that meet the timing threshold detection and waveform continuity detection conditions, the timing starts from the initial point I1 of the signal that meets the timing threshold detection, and the recorded suspected fault signal is further divided in the middle of the signal. The division is at the time point T1 after the initial point I1. The value of T1 should be slightly higher than the duration of the initial stage of the early fault in the actual environment. After the dividing point, the current signal with a time length of T2 is recorded again as the waveform signal of the main stage.

[0062] The two new signals are named I b , I s . Calculate the harmonic energy of the two signals. The specific calculation method is as follows.

[0063]

[0064]

[0065] Among them, i b (k) and i s (k) are the time domain signals I b with I s The frequency domain signal after discrete Fourier transform, m is the harmonic order, m = 1 represents the fundamental wave. N is the number of points of the time domain and frequency domain discrete signals, k is the number of the time domain discrete signal. bm With E sm i b (k) and i s (k) Energy at the mth harmonic. When m=1, it indicates the fundamental wave energy. PE bm With PE sm are signal i b (k) and i s (k) is the ratio of the mth harmonic energy to the fundamental energy.

[0066] Furthermore, the harmonic content comparison specifically includes:

[0067] The harmonic energy method can be used to evaluate the relative strength of different harmonics and the fundamental wave. In order to prevent the misjudgment problem that may exist in a single method, the harmonic contribution H is introduced as a supplementary criterion in comparing the harmonic content of the two stages. b (k) and i s (k) The harmonic contribution of the initial stage and the main stage of the early fault is further calculated. The calculation method is as follows:

[0068]

[0069] In the formula, H bx With H sx They are the time domain signals I b with I s The harmonic contribution of the xth harmonic, I bx is the root mean square of the xth harmonic current, I b1 is the root mean square current of the collected waveform in one stage. sx is the root mean square of the corresponding xth harmonic current, I s1is the root mean square current of the second-stage waveform collected. When evaluating the harmonic contribution, only the third harmonic is calculated. For the signal I after steps (1) to (4), f , if the PE of the signal bm >PE sm , and H bx >H sx . It can be preliminarily determined that the signal belongs to an early fault, and the enhanced recognition device can issue a suspected early fault determination signal.

[0070] Furthermore, the repeatability detection specifically includes: since early faults are repeatable and cumulative, based on this feature, repeatability detection can be performed on the initially identified early fault signals. The specific method is as follows:

[0071] A suspected early fault signal at any of the cable joints can increase the count of the repeatability detection device by one. When the repeatability detection count exceeds the threshold within a certain period of time, a higher level early fault alarm signal can be issued. The use of repeatability detection can further eliminate occasional interference signals similar to early fault characteristics.

[0072] Furthermore, the harmonic sampling also includes: the harmonic energy value with a multi-order harmonic contribution greater than 0.0131 in the initial stage of the early fault is taken as the total harmonic energy, and the harmonic energy value with a multi-order harmonic contribution greater than 0.0060 in the main stage is taken as the total harmonic energy.

[0073] Furthermore, the threshold and duration are defined as:

[0074] Iset1 (lower current threshold for starting recording) was set to 0.01 (A).

[0075] Iset2 (the upper current threshold for starting recording) is set to 0.2 (A).

[0076] 0.01c (continuity detection timing threshold) is set to 0.05(s).

[0077] Iset (lower current threshold for continuity detection) is set to 0.01 (A).

[0078] T1 (timing duration of the initial stage) was set to 0.5 (s).

[0079] T2 (timing duration of the main phase) was set to 10 (s).

[0080] Beneficial effects of the present invention:

[0081] The present invention proposes an enhanced recognition logic for early faults, which includes six steps: enhanced processing of the APE low-rank algorithm of the current signal, timing threshold detection, continuity detection, harmonic sampling, harmonic content comparison and repeatability detection, which can effectively screen out interference signals generated by other equipment on the distribution network line. By analyzing the time domain and frequency domain characteristics of early faults, the current amplitude, duration and spectrum difference between stages of early faults are determined, which helps to identify early faults more accurately. Preliminary experimental verification shows that the enhanced recognition logic can effectively distinguish early faults from other types of faults, especially for the fault current waveforms of equipment such as circuit breakers and overhead lines. It has good anti-interference ability. The overall accuracy of the enhanced recognition logic is maintained at a high level, and the recognition rate of early faults of cable joints reaches 97.5%. At the same time, other types of interference signals can also be well identified, and the overall accuracy reaches 99.3%. The reliability and accuracy of early fault recognition of cable joints are improved, which helps to prevent the development of early faults into permanent faults, thereby reducing the occurrence of unplanned power outages and improving the safety and stability of power grid operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 A flowchart of the enhanced recognition method of the present invention;

[0083] Figure 2 The current peak signal collected by the enhanced identification device of the present invention;

[0084] Figure 3 This is the current waveform at the beginning of the early fault;

[0085] Figure 4 Comparison of current waveforms of different samples when early faults occur for the first time;

[0086] Figure 5 It is the local current waveform in the main stage of the early fault;

[0087] Figure 6 It is the harmonic component of early fault;

[0088] Figure 7 It is a structural schematic diagram of an early fault detection unit;

[0089] Figure 8 is the position of the repeatability detection point in the grid;

[0090] Fig. 9 Enhance the before and after contrast for the current signal;

[0091] Specific implementation cases

[0092] The logic and device of the present invention are further explained below in conjunction with the accompanying drawings and specific cases. Considering that the actual operating environment of the cable, such as cable trenches and cable wells, is relatively complex, and that early faults usually occur inside the cable joints, it is difficult to directly observe the specific circumstances and characteristics of the faults. To this end, before formulating an implementation strategy, we conducted simulation experiments of early faults to better understand this problem. Given that the internal temperature of the cable is usually much higher than the external environment during normal operation, condensation generally does not occur inside the joint. The liquid water that causes early faults mainly comes from the outside, such as ground water, rainwater or dew. These liquid waters can penetrate into the tiny gaps caused by the aging of the cable joints through capillary phenomena and electrophoresis. Under the action of the electric field inside the joint, the dispersed water droplets gradually gather to form a continuous water film, which is connected to the ground position and acts as a liquid electrode on the low potential side during early faults.

[0093] For the experiments, we used a 50mm 2 The experimental sample is a specially designed 10kV medium voltage cable joint, whose outer surface and main insulation layer are made of transparent materials, which is convenient for observing the internal fault. Since the early faults mainly occur on the inner surface of the cable joint, the influence of the transparent joint design on the fault characteristics can be ignored. Some parameters of the experimental sample are shown in Table 1.

[0094] Table 1 Some parameters of experimental samples

[0095]

[0096] When an early fault occurs, the damage is usually limited to the inside of a single cable joint, so this type of fault is a typical single-phase grounding fault. In order to simulate the common mixed connection of overhead lines and cable lines in urban distribution networks, we configured a combination of RLC elements in the external circuit to be equivalent to the series connection of 5 kilometers of overhead lines and 5 kilometers of cable lines. The system neutral point adopts a small resistance grounding method, and the set grounding resistance value is 10 ohms.

[0097] The experimental samples were placed in a constant temperature and humidity environment. The experimental power supply provided a constant voltage of 5.8 kV, which is equivalent to the single-phase to ground voltage in a 10 kV distribution network. The environmental conditions were set at a temperature of 20 degrees Celsius and a relative humidity of 90%. In order to capture the discharge phenomenon during the fault process, a high-speed camera with a lens horizontally aimed at the fault point was installed in the experiment. The resolution of the camera was 1074×768 pixels and the shutter speed reached 5000 frames per second. In addition, the recorder was used to record the voltage at both ends of the fault point and the ground current, with a sampling rate of 10,000 sample points per second (10kSPs). The specific experimental steps include:

[0098] 1) Inject local tap water quantitatively into the cable joint in the horizontal direction from the bottom of the dry cable joint. The conductivity of the tap water is 0.05S / m and the single injection volume is 0.1mL.

[0099] 2) Turn on the power. If an arc discharge of at least milliampere level appears in the air gap of the cable connector at this time, record the non-electrical and electrical quantities of the fault. Continue to observe. If the discharge stops automatically without external interference, the discharge is considered an early fault. If the discharge persists and eventually causes violent combustion, it is considered a permanent fault.

[0100] 3) If no discharge occurs in step 2), turn off the power supply and continue to increase the water injection at the same position in small steps of 0.1 mL to further increase the length of the water accumulation. Repeat step 2) until discharge occurs.

[0101] 4) Repeat steps 1) to 3) until permanent failure occurs. Replace new samples and repeat steps 1) to 4) until the early failure test of all samples is completed.

[0102] The fault scenario of the early fault directly determines the electrical quantity characteristics of the fault. If the fault electrical quantity characteristics are to be extracted, it is necessary to analyze the stable fault scenario and analyze the electrical quantity in the scenario on this basis.

[0103] The internal environment of the cable connector in a waterlogged environment can be regarded as the discharge between the air and the cross-linked polyethylene-silicone rubber composite interface. Referring to existing research, it can be seen that the leakage current amplitude in a waterless environment is much lower than the early fault current, so the influence of the leakage current can be ignored when evaluating the early fault characteristics. Figure 2 This is a complete early fault current waveform. Combined with experimental phenomena, it can be obtained that early faults can be divided into an initial stage with a sharp increase in current amplitude and a subsequent main stage. The initial stage starts at the time when the arc discharge occurs and ends when the accumulated water and the semi-conductive layer come into direct contact for the first time. It lasts for about 0.1 to 1 second. The main stage starts from the end of the initial stage until the fault stops completely, and lasts for several seconds to tens of seconds. Figure 3 This is the current waveform at the initial stage of an early fault. As can be seen from the figure, the current at this stage has a "flat shoulder" that gradually decreases with time, indicating that the influence of the arc on the fault current characteristics is gradually decreasing. Figure 4 The current waveforms of four intact experimental samples when the early fault first occurred are compared. The initial time point of current acquisition is the first current peak point when the discharge occurs, and the acquisition time is 0.15s. Figure 3 and Figure 4It can be seen that the current at this time is increasing as a whole, and there may be fluctuations at some time points. From the experimental phenomenon, it can be seen that during this period of time, as the liquid electrode approaches the semi-conductive layer of the joint further, the arc length decreases, the total fault resistance decreases, and the current amplitude generally shows an increasing trend. The arc discharge between the solid electrode and the liquid electrode hinders the further displacement of the liquid electrode, which affects the change trend of the current amplitude and may cause short-term fluctuations. The subtle differences in the distribution of dust on the insulating surface and the microstructure make the volume of accumulated water and the resistance of accumulated water required for early faults in different samples different, thereby affecting the amplitude of the fault current. In the experiment, the current peak value of different devices in the initial stage of early faults is in the range of 0.05 to 0.2A. The accumulated water of sample 4 has four single discharges of half a cycle during the stretching process, with a duration of about 0.04s. Combined with the experimental phenomenon, it can be seen that during this period, the field strength between the tip of the accumulated water and the solid electrode cannot maintain a stable arc, and discharge can only occur when the voltage exceeds the negative half cycle. The main stage of the early fault occurs after the initial stage, corresponding to the evaporation process of the accumulated water. The current waveform of this stage is as follows: Figure 5 As shown. Figure 5 (a) with Figure 5 (b) Figure 2 Each 2s segment in the main stage of the early to mid-term fault, Figure 5 (c) Figure 5 (a) with Figure 5 (b) Fault resistance change curve within 4s. It can be seen from the experimental phenomenon that it takes tens of seconds for the accumulated water to completely evaporate. In a short time scale, the external circuit and the total amount of accumulated water can be considered to have no obvious changes. Figure 5 As shown, the amplitude of the fault current fluctuates violently, and the current peak value can reach 1.5 times the valley value in a short time. Combined with the experimental phenomena, the reason for this situation is that the resistance of the accumulated water fluctuates continuously due to deformation and displacement. Since there is always a stretching process of accumulated water in the initial stage of early faults, the fault scenario in the initial stage of early faults is relatively stable. Similarly, for any early fault, there is an evaporation process of accumulated water, which is the main stage of the early fault. This makes the initial stage and the main stage of the early fault stable. The difference in fault scenarios and corresponding fault characteristics in the two stages can be used as an identification criterion for early faults. On this basis, the fault currents in the two stages are discrete Fourier transformed respectively to obtain their frequency domain characteristics, such as Figure 6As shown. The experimental phenomenon can be used to obtain the source of the two-stage harmonics and the reason for the change of this frequency domain feature. Since the early fault is a typical high-resistance grounding fault, its fault resistance is much larger than the line inductive reactance and capacitive reactance. Therefore, the main source of the harmonics of the fault current is the nonlinear change of the fault resistance. The size of the harmonic component in a certain period of time is directly determined by the magnitude of the nonlinear change of the resistance during this period of time. For the early fault in the initial stage, since the discharge in this stage has not yet started, the total amount of accumulated water and the resistance have not changed significantly, and there is no obvious change in the flashover trace. Therefore, the harmonic component mainly comes from the arc, that is, as the accumulated water continues to stretch, the total length of the arc decreases, the arc resistance decreases, and at the same time, the reduction in the distance between the liquid-solid electrodes also reduces the harmonic components caused by the arc "flat shoulder" process.

[0104] The nonlinear change of resistance caused by water splashing and arc in the main stage is mainly manifested as short-term disturbances of varying amplitudes. The main factor that dominates its long-term waveform characteristics is the increase in water resistance caused by the evaporation of water. Therefore, the current shows local fluctuations and overall stability. According to experimental phenomena, compared with the overall stable main stage, the total duration of the water stretching process in the initial stage is only 0.1 to 0.5s. The rate of change of current amplitude and the intensity of the "flat shoulder" process are much higher than those in the main stage, that is, in the same time, the nonlinear change degree of fault resistance is much higher than that in the subsequent main stage. This also leads to a greater harmonic contribution in the initial stage than in the main stage. The damage caused by early faults is cumulative, and there are several to dozens of similar early faults with similar characteristics from the first early fault to the permanent fault. This is conducive to its identification.

[0105] For this purpose, a device for enhancing early fault identification of waterlogged cables is described, which specifically includes:

[0106] The cable joint is wrapped with a copper shielding layer to connect the semi-conductive layers of the cables on both sides that are disconnected when installing the cable joint. The grounding wire is wrapped around the copper shielding. The external current path of the early fault occurring in the cable joint is busbar-front line-cable joint-copper shielding-grounding wire-grounding point.

[0107] Early fault detection unit such as Figure 7As shown. The device is installed on the grounding wire between the copper shielding layer and the grounding point, and the current acquisition equipment uses a current transformer (CT) wound on the grounding wire. Compared with general grounding faults, the current value generated by early faults is much smaller, usually at the milliampere level. Therefore, the winding design of the CT is intended to enhance the fault current signal. For the mutual inductance current signal that is amplified at the same time due to the winding of the coil, the current flowing through the internal cable line is less likely to change rapidly during the several seconds to tens of seconds of the early fault, and the possibility of causing large fluctuations in the induced current on the copper shielding layer is also low. Therefore, in the initial setting, the induced current on the copper shielding layer can be regarded as a background noise of a fixed amplitude and filtered out, so as to obtain an early fault current signal with the induced current interference removed when the fault occurs. In addition to the equipment for measuring the fault current, a repetitive signal counter must be installed on each cable joint. Its specific structure is as follows Figure 8 As shown. The fault signal collected from the fault current measurement device will be transmitted to the information node through the detection point to monitor the repeatability of the fault signal. When a fault occurs, the current flow path is: cable conductor - joint semi-conductive layer - arc - water accumulation - joint stress cone - grounding point.

[0108] For this purpose, a waterlogged cable early fault enhanced identification logic is described, specifically including:

[0109] a) Decompose the collected cable current signal into 6 sub-signals through EMD, and each sub-signal is expressed as:

[0110] s={s1,s2,...,s N}

[0111] b) For each sub-signal, select embedding dimension d=3 and time delay l=1, then embed each sub-signal IMF into a d-dimensional space to obtain a reconstructed vector:

[0112]

[0113] c) Reconstruct the vector Sort the elements in to get a new vector:

[0114]

[0115] Among them, j1, j2, ..., jd are The index of the element in .

[0116] d) For a d-dimensional reconstruction vector, there are d! sorting modes, and the probability of the i-th sorting mode appearing is recorded as The probability of each sorting mode occurring is calculated according to the following formula.

[0117]

[0118] e) Calculate the APE entropy value of each sub-signal according to the following formula.

[0119]

[0120] The enhanced processing of the current signal APE low-rank algorithm in the enhanced recognition logic specifically includes:

[0121] a) Cluster the six sub-signals into two clusters based on the clustering algorithm and APE entropy value.

[0122] b) Each sub-signal in the sub-signal cluster with high APE entropy value is enhanced by a low-rank matrix decomposition algorithm, specifically including:

[0123] Each sub-signal in the sub-signal cluster with high APE entropy value is represented as: x = {x1, x2, ..., x N}, and convert it into the observation matrix form X:

[0124]

[0125] Where L is the embedding dimension, L = 0.5*N, K = N-L+1.

[0126] The SVD decomposition result of X is calculated by the following formula.

[0127]

[0128] Where H represents the conjugate transpose, are the descending singular values ​​of X, d is the number of singular values, They represent the left and right singular vectors respectively.

[0129] Set the number of clusters k=3 and use the clustering algorithm to cluster the singular values ​​of X. Then calculate the average singular value of each cluster, regard the singular values ​​in the two clusters with larger average values ​​as being associated with the sub-signal and obtain the matrix rank r, which is the number of singular values ​​in the two larger clusters.

[0130] Calculate the i-th reconstruction coefficient according to the following formula

[0131]

[0132] in, I is the identity matrix, for The differential form of , Tr is the sum operation of the diagonal elements of a square matrix.

[0133] Compute the subsignal matrix using the reconstruction coefficients and singular value vectors As shown below.

[0134]

[0135] Finally, the enhanced sub-signal is obtained

[0136]

[0137] Each enhanced sub-signal in the sub-signal cluster with a high APE entropy value is added together to obtain an enhanced current signal.

[0138] On the other hand, when estimating the resistance of the accumulated water, it can be assumed that the accumulated water is homogeneous, that is, the resistivity ρ of each part of the accumulated water is the same, and this physical quantity is affected by the environment in which the cable joint is located. In this case, the resistance of the accumulated water can be calculated by the following formula.

[0139]

[0140] Where l is the length of the pond, ds is the differential of the area, and Tr is the sum of the diagonal elements of a square matrix.

[0141] To simplify the calculation, the accumulated water can be regarded as a cylinder with the same cross-sectional area. The equivalent cross-sectional area of ​​the cylinder at time t is s t , then the above formula can be transformed into:

[0142]

[0143] In the formula, r t is the water resistance at time t, l t is the length of water accumulation at time t, V t is the volume of accumulated water at time t.

[0144] When an early fault occurs for the first time, there are no signs of ablation in the circuit, and the resistance in the circuit is mainly composed of water resistance. According to experiments, when the number of discharges is not much different, there is no obvious difference in the initial value of the fault current. The estimation of the water resistance is helpful to set the trigger threshold of the fault measurement device, and the setting of the early fault duration threshold is based on the estimation of the heat of the evaporation process. If a variety of equipment (insulators, transformer elbows, overhead lines, etc.) in the distribution network fails, their fault signals may affect the early fault enhancement identification device and interfere with the early fault identification of the cable. Therefore, the early fault enhancement identification device and logic described in this article must not only realize the identification of early faults of cable joints, but also prevent the fault signals of other equipment from causing false alarms. In order to distinguish the early fault signal from other possible interference signals. To this end, the early fault enhancement identification logic can be divided into the following steps:

[0145] (1) Timing threshold detection

[0146] The detection device set on the grounding wire of the cable joint is in the normally open state. When the current signal I flowing through the sample f When the signal amplitude at the initial stage is within the threshold range as shown in the following formula, the recording of the suspected fault signal is started.

[0147]

[0148] I set1 <|I start |<I set2

[0149] In the formula, [I1,I2,…,I n ] The array is the amplitude of the current signal at each peak point. From the experimental phenomenon, we know that n is determined by the duration of the initial stage of the early fault and can be 5 to 20; I start is the current signal at the initial stage of the fault signal; I set1 To determine the lower limit of the current threshold for fault recording, I set2 The upper limit of the current threshold for determining whether fault recording can continue.

[0150] (2) Waveform continuity detection

[0151] During the signal acquisition process, if the signal I f There is a continuous signal I f1 , if the peak current amplitude of the signal is always lower than the threshold, the signal acquisition is stopped. The specific detection method is shown in the following formula:

[0152]

[0153] I f1 =[I m ,I m+1 ,..,I m+c ]

[0154] |I f1 |<I set

[0155] In the formula, [I m ,I m+c ,…,I m+c ] is I f1 The amplitude of each peak point in the segment signal. m For I f1 The current amplitude at the initial point of the signal; I set is the lower limit of the current for determining whether the signal is continuous; c is the threshold for signal continuity detection, I f1 The signal duration is 0.01c.

[0156] (3) Harmonic sampling

[0157] For signals that meet conditions (1) and (2), the timing starts from the initial point I1 of the signal that meets the timing threshold detection, and the recorded suspected fault signal is further divided in the middle of the signal. The division is at the time point T1 after the initial point I1. The value of T1 should be slightly higher than the duration of the initial stage of the early fault in the actual environment. After the dividing point, the current signal with a time length of T2 is recorded again as the waveform signal of the main stage.

[0158] The two new signals are named I b , I s . Calculate the harmonic energy of the two signals. The specific calculation method is as follows.

[0159]

[0160] Among them, i b (k) and i s (k) are the time domain signals I b with I s The frequency domain signal after discrete Fourier transform, m is the harmonic order, m = 1 represents the fundamental wave. N is the number of points of the time domain and frequency domain discrete signals, k is the number of the time domain discrete signal. bm With E sm i b (k) and i s (k) Energy at the mth harmonic. When m=1, it indicates the fundamental wave energy. PE bm With PE sm are signal i b (k) and i s (k) is the ratio of the mth harmonic energy to the fundamental energy.

[0161] (4) Comparison of harmonic content

[0162] The harmonic energy method in step (3) can be used to evaluate the relative strength of different harmonics and the fundamental wave. In order to prevent the misjudgment problem that may exist in a single method, the harmonic contribution H is introduced as a supplementary criterion in comparing the harmonic content of the two stages. b (k) and i s (k) The harmonic contribution of the initial stage and the main stage of the early fault is further calculated. The calculation method is as follows:

[0163]

[0164] In the formula, H bx With H sx They are the time domain signals I b with I sThe harmonic contribution of the xth harmonic, I bx is the root mean square of the xth harmonic current, I b1 is the root mean square current of the collected waveform in one stage. sx is the root mean square of the corresponding xth harmonic current, I s1 is the root mean square current of the second-stage waveform collected. When evaluating the harmonic contribution, only the third harmonic is calculated. For the signal I after steps (1) to (4), f , if the PE of the signal bm >PE sm , and H bx >H sx . It can be preliminarily determined that the signal belongs to an early fault, and the enhanced recognition device can issue a suspected early fault determination signal.

[0165] (5) Repeatability test

[0166] Since early faults are repetitive and cumulative, based on this feature, the initially identified early fault signals can be tested for repetitiveness. The specific method is as follows:

[0167] A suspected early fault signal at any of the cable joints can increase the count of the repeatability detection device by one. When the repeatability detection count exceeds the threshold within a certain period of time, a higher level early fault alarm signal can be issued. The use of repeatability detection can further eliminate occasional interference signals similar to early fault characteristics.

[0168] The (3) harmonic sampling also includes: the harmonic energy value of the multi-order harmonic contribution greater than 0.0131 in the initial stage of the early fault is taken as the total harmonic energy, and the harmonic energy value of the multi-order harmonic contribution greater than 0.0060 in the main stage is taken as the total harmonic energy.

[0169] The thresholds and durations in (1)-(3) are defined as.

[0170] Iset1 (lower current threshold for starting recording) was set to 0.01 (A).

[0171] Iset2 (the upper current threshold for starting recording) is set to 0.2 (A).

[0172] 0.01c (continuity detection timing threshold) is set to 0.05(s).

[0173] Iset (lower current threshold for continuity detection) is set to 0.01 (A).

[0174] T1 (timing duration of the initial stage) was set to 0.5 (s).

[0175] T2 (timing duration of the main phase) was set to 10 (s).

[0176] First, we conduct enhanced logic verification on the actual collected current signal. Fig. 9 As shown in (a) and (b), the de-enhancement effect is significant and the low-frequency interference components are suppressed.

[0177] Secondly, we identified the fault signals of the 10kV distribution network lines with low-resistance grounding at the neutral point in a northern city from 2021 to 2023. The total number of faulty equipment samples was 770, of which 40 were early fault samples of cable joints. The remaining 730 samples included fault waveforms of various 10kV distribution network equipment such as cable bodies, overhead lines, 35 / 10kV transformers and insulators, which were used as interference data to test the reliability of the enhanced identification logic. The specific information of the interference data is shown in Table 2.

[0178] Table 2 Sources of interference signals in test samples

[0179]

[0180] For the fault data monitored by the existing current protection device without repeatability detection, the enhanced identification logic (1) to (4) can be used for identification to preliminarily verify the reliability of the identification method. The test results are shown in Table 3.

[0181] Table 3 Enhanced recognition logic test results

[0182]

[0183] In the false alarm and missed alarm results, in the missed cable joint early fault cases, the total fault duration is shorter than the main stage duration in the setting. This causes the false alarm of the fault signal. The four false alarm results are all flashovers along the outer surface of the device, and their fault scenarios and fault characteristics are similar to those of the cable joint early fault. The threshold setting method of the cable early fault identification model will be further refined in subsequent research.

[0184] Preliminary test results show that the enhanced recognition logic proposed in this paper can effectively distinguish early faults from the fault current waveforms of other equipment such as circuit breakers and overhead lines. It has good resistance to interference caused by such fault waveforms. (1) and (2) in the enhanced recognition logic have a direct and effective screening effect on large current grounding faults such as cable metallic grounding and overhead line tree line grounding. The surface discharge of insulators in humid and dirty environments and the internal discharge of transformer elbows have similar fault circuit structures to early faults. In addition to amplitude judgment, (3) and (4) in the enhanced recognition logic can also play a good role in identifying such faults. This keeps the overall accuracy of the enhanced recognition logic at a high level.

Claims

1. A device for enhancing the early fault identification of waterlogged cables, located on the grounding wire between the copper shielding layer and the grounding point, characterized in that: include: The current collection device uses a current transformer (CT) wound on the ground wire to collect the current signal in the cable; A repeatability signal counter, installed at each cable connector, is used to record the repeatability of the fault signal from the current acquisition device; An information node is used to receive a fault signal from a current acquisition device and support repeatability detection of the fault signal; wherein the fault signal includes an initial stage and a main stage, the acquisition period of the initial stage is set to 0.5 seconds from the appearance of the first arc discharge waveform, and the acquisition period of the main stage is 10 seconds after the end of the initial stage.

2. The device for enhancing early fault identification of waterlogged cables according to claim 1 is characterized in that: It also includes a fault current measuring device, which is used to upload the collected fault signal to the information node through the detection point, providing structural support for the repeatability detection of the fault signal.

3. A method for enhancing early fault identification of waterlogged cables, characterized in that: The steps include: S1. Enhanced processing of current signal APE low-rank algorithm: S2. Timing threshold detection: when the signal amplitude of the current signal flowing through the sample in the initial stage is within the preset threshold range, the recording of the suspected fault signal is started; S3. Continuity detection: During the signal acquisition process, if there is a continuous signal whose peak current amplitude is always lower than the preset threshold, the signal acquisition is stopped; S4. Harmonic sampling: Segment the signal that meets the timing threshold detection and waveform continuity detection, and calculate the harmonic energy and harmonic contribution of the two signal segments; S5. Comparison of harmonic content: By comparing the harmonic energy and harmonic contribution of the two signals, it is preliminarily determined whether the signal belongs to an early fault; S6. Repeatability detection: Perform repeatability detection on the initially identified early fault signals. When the repeatability detection count exceeds the preset threshold within a certain period of time, a higher level early fault alarm signal is issued.

4. The method for enhancing early fault identification of waterlogged cables according to claim 3 is characterized in that: The step S1 is to perform enhanced processing of the current signal APE low-rank algorithm, and the specific steps are as follows: S1.1 decomposes the collected cable current signal into 6 sub-signals through EMD, and each sub-signal is expressed as: s={s1,s2,...,s N } Step S1.2 selects the embedding dimension d and time delay l for each sub-signal, and then embeds each sub-signal IMF into a d-dimensional space to obtain a reconstructed vector: Step S1.3 reconstructs the vector Sort the elements in to get a new vector: Among them, j1,j2,...,j d yes The index of the element in; Step S1.4 For a d-dimensional reconstruction vector, there are d! sorting modes, and the probability of the i-th sorting mode appearing is recorded as The probability of each sorting pattern occurring is calculated according to the following formula: Step S1.5 calculates the APE entropy value of each sub-signal, and the formula is as follows:

5. The method for enhancing early fault identification of waterlogged cables according to claim 4 is characterized in that: The step S1 of enhancing the current signal APE low-rank algorithm further includes: Step S1.6) clustering the six sub-signals into two clusters according to the clustering algorithm and the APE entropy value; Step S1.7) Each sub-signal in the sub-signal cluster with high APE entropy value is enhanced by a low-rank matrix decomposition algorithm, including: Each sub-signal in the sub-signal cluster with high APE entropy value is represented as: x = {x1, x2, ..., x N }, and convert it into the observation matrix form X: Where L is the embedding dimension, L = 0.5*N, K = N-L+1; The SVD decomposition result of X is calculated by the following formula: Where H represents the conjugate transpose, are the descending singular values ​​of X, d is the number of singular values, They represent the left singular vector and the right singular vector respectively; Cluster the singular values ​​of X using a clustering algorithm, then calculate the average singular value of each cluster, consider the singular values ​​in the two clusters with the larger average value as being associated with the sub-signal and obtain the matrix rank r, which is the number of singular values ​​in the two larger clusters; The i-th reconstruction coefficient is calculated according to the following formula in, I is the identity matrix, for The differential form of , Tr is the sum operation of the diagonal elements of a square matrix; Compute the subsignal matrix using the reconstruction coefficients and singular value vectors As shown below: Calculate the enhanced sub-signal As shown below: Each enhanced sub-signal in the sub-signal cluster with a high APE entropy value is added together to obtain an enhanced current signal.

6. The method for enhancing early fault identification of waterlogged cables according to claim 1 is characterized in that: S2. Timing threshold detection: When the signal amplitude of the current signal flowing through the sample in the initial stage is within the preset threshold range, the recording of the suspected fault signal is started, specifically including: When the current signal I flowing through the sample f When the signal amplitude at the initial stage is within the threshold range shown in the following formula, the recording of the suspected fault signal is started: I set1 <|I start |<I set2 In the formula, [I1,I2,…,I n ] array is the amplitude of the current signal at each peak point, n is determined by the duration of the initial stage of the early fault and can be 5 to 20; I start is the current signal at the initial stage of the fault signal; I set1 To determine the lower limit of the current threshold for fault recording, I set2 The upper limit of the current threshold for determining whether fault recording can continue.

7. The method for enhancing early fault identification of waterlogged cables according to claim 1 is characterized in that: S3. Continuity detection: During the signal acquisition process, if there is a continuous signal whose peak current amplitude is always lower than the preset threshold, the signal acquisition is stopped. The specific detection method is shown in the following formula: I f1 =[I m ,I m+1 ,..,I m+c ] |I f1 |<I set In the formula, the continuous signal I f1 Existence in signal I f In, [I m ,I m+c ,…,I m+c ] is I f1 The amplitude of each peak point in the segment signal, I m For I f1 The current amplitude at the initial point of the signal; I set is the lower limit of the current for determining whether the signal is continuous; c is the threshold for signal continuity detection, I f1 The signal duration is 0.01c.

8. The method for enhancing early fault identification of waterlogged cables according to claim 1 is characterized in that: S4. Harmonic sampling: Segmenting the signal that meets the timing threshold detection and waveform continuity detection, and calculating the harmonic energy and harmonic contribution of the two signal segments, specifically including: The timing starts from the initial point I1 of the signal that meets the timing threshold detection, and the recorded suspected fault signal is further divided in the middle of the signal. The division is at the time point T1 after the initial point I1. The value of T1 should be slightly higher than the duration of the initial stage of the early fault in the actual environment. After the dividing point, the current signal with a time length of T2 is recorded again as the waveform signal of the main stage; The two new signals are named I b , I s ; Calculate the harmonic energy of the two new signals using the following formula: Among them, i b (k) and i s (k) are the time domain signals I b with I s The frequency domain signal after discrete Fourier transform, m is the harmonic order, m = 1 represents the fundamental wave; N is the number of points of the time domain and frequency domain discrete signals, k is the number of the time domain discrete signal; E bm With E sm i b (k) and i s (k) Energy at the mth harmonic, when m=1, it indicates the fundamental energy; PE bm With PE sm are signal i b (k) and i s (k) The ratio of the mth harmonic energy to the fundamental energy.

9. The method for enhancing early fault identification of waterlogged cables according to claim 1, characterized in that: S5. Harmonic content comparison: By comparing the harmonic energy and harmonic contribution of the two signals, it is preliminarily determined whether the signal belongs to an early fault; specifically including: The harmonic contribution of the initial stage and the main stage of the early fault is calculated as follows: In the formula, H bx With H sx They are the time domain signals I b with I s The harmonic contribution of the xth harmonic, I bx is the root mean square of the xth harmonic current, I b1 is the root mean square current of the collected waveform in one stage; I sx is the root mean square of the corresponding xth harmonic current, I s1 is the collected two-stage waveform fundamental current RMS; For the signal I after steps (1) to (4) f , if the PE of the signal bm >PE sm , and H bx >H sx . It can be preliminarily determined that the signal belongs to an early fault, and the enhanced recognition device can issue a suspected early fault determination signal.

10. The method for enhancing early fault identification of waterlogged cables according to claim 9, characterized in that: The S5. harmonic content comparison also includes: The harmonic energy values ​​with multi-order harmonic contributions greater than 0.0131 in the initial stage of early faults are taken as the total harmonic energy; The harmonic energy value with multi-order harmonic contribution greater than 0.0060 in the main stage is taken as the total harmonic energy.

11. A method for enhancing early fault identification of waterlogged cables according to any one of claims 3 to 10, characterized in that: Iset1 (lower current threshold for starting recording) was set to 0.01 (A); Iset2 (the upper limit of the current threshold for starting recording) was set to 0.2 (A); 0.01c (continuity detection timing threshold) was set to 0.05(s); Iset (lower current threshold for continuity detection) is set to 0.01 (A); T1 (timing duration of the initial stage) was set to 0.5 (s); T2 (timing duration of the main phase) was set to 10 (s).

Images