Power distribution cable partial discharge online positioning method and system based on improved cross-correlation algorithm

Abstract

The present application relates to the technical field of power cable partial discharge, and more particularly to a power distribution cable partial discharge online positioning method based on an improved cross-correlation algorithm, comprising: on the basis of the GCC-PHAT method, an optimized pulse time window technology is proposed, and the time window size of the correlation function is adaptively adjusted according to the relative position of the partial discharge source in the cable. Considering the frequency characteristics of electromagnetic waves during cable propagation, the partial discharge wave velocity is accurately estimated by drawing a frequency-wave velocity characteristic curve. Finally, a pruning mean data filtering algorithm is proposed, and the sampling data is filtered by constructing a three-layer data filter. Compared with the traditional time delay estimation method, the results show that the proposed method can effectively improve the time delay estimation accuracy and reduce the positioning error under the conditions of serious environmental interference and limited sampling rate of the detection equipment. The cable partial discharge online positioning experiment is carried out, and the experimental results show that the relative error of the proposed method is 0.47%, which basically meets the requirements of engineering practice.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power cable partial discharge, and particularly relates to a power distribution cable partial discharge online positioning method based on an improved cross-correlation algorithm. BACKGROUND

[0002] With the continuous expansion of power distribution network cable lines and the in-depth optimization of power distribution network topology, the safe and stable operation of urban medium-voltage power distribution cables should be paid attention to, and once an accident occurs, the power supply reliability of the city will be affected. The insulation state of the medium-voltage power distribution cable is closely related to the cable partial discharge. Due to factors such as equipment aging or external damage, voids, impurities, bubbles and other defects are generated in the cross-linked polyethylene (XLPE) insulation layer of the cable. In the actual operation process, when the field strength on these local defects reaches the insulation breakdown field strength, partial discharge will occur.

[0003] Nowadays, the partial discharge online positioning technology of the power distribution cable still relies on the traditional time delay estimation positioning method, and the positioning accuracy is low in the case of low signal-to-noise ratio. The frequency characteristics of the partial discharge pulse wave speed are not considered when the partial discharge source position is estimated according to the time delay estimation, which increases the positioning error. In addition, the complexity of the cable operation environment and the abnormality of the partial discharge detection equipment increase the uncertainty of the online positioning result. Therefore, it is necessary to propose a cable partial discharge online detection and positioning method for engineering practical application. SUMMARY

[0004] The present application provides a power distribution cable partial discharge online positioning method based on an improved cross-correlation algorithm, which can effectively solve the problems in the background art.

[0005] In order to achieve the above purpose, the technical scheme adopted by the present application is:

[0006] A power distribution cable partial discharge online positioning method based on an improved cross-correlation algorithm, comprising the following steps:

[0007] Through empirical wavelet transform, the partial discharge signals x1(t) and x2(t) collected at both ends of the cable are subjected to data denoising pretreatment;

[0008] The time window technology is optimized to accurately obtain a time window covering complete partial discharge pulse information, the size of the time window of the correlation function is adaptively adjusted according to the relative position of the partial discharge source in the cable, and the GCC-PHAT algorithm is used to estimate the signal time delay;

[0009] The frequency-wave speed curve of electromagnetic wave propagation in the cable is established, and the signal wave speed is obtained according to the center frequency of the partial discharge signal, and then the position of the partial discharge source is located;

[0010] The pruning mean data filtering technology is used to statistically control the generated multiple partial discharge position samples through a three-layer data filter, so as to minimize the estimated partial discharge positioning error.

[0011] Further, the optimization pulse time window technology comprises the following specific technical process:

[0012] The separation threshold U of the noise signal and the pulse signal is determined by the following formula thr , so as to accurately identify the partial discharge pulse in the sampling signal:

[0013] ;

[0014] Wherein, P noise is the residual noise power, and k is a proportional coefficient, generally taken as 2;

[0015] The start and end times of the two partial discharge signal amplitudes just exceeding the threshold are t as , t ae and t bs , t be , and the start and end time points t ws , t we of the optimization time window are determined by the following formula:

[0016]

[0017] ;

[0018] Wherein, t r , t f are the time margins of the rising edge and the falling edge of the signal respectively, and the parameter t r is set to be smaller than t f ;

[0019] The optimization time window width T w is calculated by the following formula:

[0020] .

[0021] Further, the received sampling signal is preprocessed based on the empirical wavelet transform, and the residual noise is power spectrum estimated to obtain the residual noise power.

[0022] Further, the partial discharge pulses x1(t), x2(t) in the two sampling signals are extracted respectively, and the GCC-PHAT time delay estimation of the two signals is performed in the optimized time window.

[0023] Further, the frequency-wave velocity characteristic curve of the electromagnetic wave in the cable is drawn according to the basic parameters of the cable, which comprises the following steps:

[0024] The propagation constant of the coaxial cable is:

[0025] ;

[0026] Where the unit equivalent conductance G is negligible when the cable is in normal operation, the formula can be approximately expressed as:

[0027] ;

[0028] The ideal unit equivalent inductance L0' is:

[0029] ;

[0030] Where μ0 is the magnetic permeability of the insulation layer between the inner diameter of the cable and the outer diameter of the cable; D A is the outer diameter of the cable, D C is the inner diameter of the cable;

[0031] The actual equivalent inductance corresponding to the outer semi-conductive layer and the inner semi-conductive layer of the cable is:

[0032]

[0033] ;

[0034] Where δ A is the skin depth of the outer semi-conductive layer of the cable, δ C is the skin depth of the inner semi-conductive layer of the cable, and its formula is:

[0035] ;

[0036] Where σ is the conductivity;

[0037] The actual unit equivalent inductance of the cable is:

[0038]

[0039] ;

[0040] The unit equivalent resistance R0 is expressed as

[0041] ;

[0042] We get:

[0043] ;

[0044] Where,

[0045] ;

[0046] Expanding the above formula by Taylor and keeping the first two terms, we get:

[0047] ;

[0048] Again

[0049] ;

[0050] From the relationship between wave velocity and beta:

[0051] ;

[0052] According to the above formula, the frequency-wave velocity characteristic curve of the electromagnetic wave in the cable is drawn, and the corresponding wave velocity is obtained by measuring the center frequency of the electromagnetic wave.

[0053] Further, by constructing a three-layer data filter to eliminate the detected bad data, accurate estimation of the multi-sample partial discharge positioning result is realized.

[0054] Among them, the three-layer data filter includes a pruning filter, a clustering filter and an average filter, the pruning filter acts as a first layer filter of positioning samples by pruning and eliminating outliers, a maximum density clustering (MDCA) algorithm is used as a second layer clustering filter, and finally an average filter is used to calculate the average value of the positioning samples in the cluster as the final cable partial discharge online positioning result.

[0055] Further, the pruning filter sorts the partial discharge positioning result samples from the minimum value to the maximum value of the partial discharge position, and filters 25%-75% of the samples in the positioning samples as the input of the clustering filter.

[0056] Further, the maximum density clustering algorithm introduces division clustering based on the idea of density, which can automatically determine the number of clusters and discover clusters of any shape.

[0057] The specific process of the maximum density clustering algorithm is as follows:

[0058] Calculate the local density of the i-th data point in the sample set, with a cutoff distance d c : i

[0059] ;

[0060] Among them:

[0061] ;

[0062] Calculate the distance δi of each data point to the high local density point:

[0063] ;

[0064] ​In the rho-delta graph, data points with large local density rho and large high local density point distance delta are selected as cluster centers, and data points with small local density rho and large high local density point distance delta are selected as data abnormal points.

[0065] A power distribution cable partial discharge online positioning system based on an improved cross-correlation algorithm is applied to the above method and comprises:

[0066] At least two groups of discharge signal sensors are arranged at both ends of the measured cable and are used to collect the partial discharge signals of the measured cable.

[0067] At least two monitoring terminals are connected to the at least two groups of discharge signal sensors, are used to receive the partial discharge signals and obtain pulse waveform data.

[0068] A network switch is connected to the at least two monitoring terminals and is used to upload the pulse waveform data and the partial discharge signals.

[0069] A diagnosis platform is connected to the network switch and is used to perform correlation analysis on the received pulse waveform data and partial discharge signals and determine the partial discharge position.

[0070] The present application has the following advantages:

[0071] The present application discloses a power distribution cable partial discharge online positioning technology based on an improved cross-correlation algorithm, which can effectively improve the time delay estimation accuracy and reduce the positioning error under the condition that the environmental interference is serious and the sampling rate of the detection equipment is limited, and basically meets the actual engineering requirements. BRIEF DESCRIPTION OF DRAWINGS

[0072] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0073] Figure 1 The present application discloses a power distribution cable partial discharge online positioning technology flow chart based on an improved cross-correlation algorithm.

[0074] Figure 2 The present application discloses a power distribution cable partial discharge online positioning experimental principle diagram based on an improved cross-correlation algorithm.

[0075] Figure 3 The present application discloses a power distribution cable partial discharge online positioning experimental wiring diagram based on an improved cross-correlation algorithm.

[0076] Figure 4 An improved cross-correlation algorithm-based power distribution cable partial discharge online positioning experimental measurement system;

[0077] Figure 5(1) is a waveform diagram of the A-end partial discharge signal measured in the experiment;

[0078] Figure 5(2) is a waveform diagram of the B-end partial discharge signal measured in the experiment;

[0079] Figure 6(1) is a frequency spectrum diagram of the A-end partial discharge signal measured in the experiment;

[0080] Figure 6(2) is a frequency spectrum diagram of the B-end partial discharge signal measured in the experiment;

[0081] Figure 7 Figure 7 is a cable frequency-wave velocity characteristic curve diagram in the experiment;

[0082] Figure 8 GCC-PHAT time delay estimation result;

[0083] Figure 9 Partial discharge positioning result based on MDCA. DETAILED DESCRIPTION

[0084] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments of the present application.

[0085] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0086] As Figures 1 to 9 shown in a power distribution cable partial discharge online positioning method based on an improved cross-correlation algorithm, comprising the following steps:

[0087] Through empirical wavelet transform, the collected partial discharge signals x1(t), x2(t) are preprocessed for data denoising;

[0088] Through optimized pulse time window technology, a time window covering complete partial discharge pulse information is accurately obtained, the size of the time window of the correlation function is adaptively adjusted according to the relative position of the partial discharge source in the cable, and the GCC-PHAT (generalized cross-correlation phase transform) algorithm is used to estimate the signal time delay;

[0089] The frequency-wave velocity curve of electromagnetic wave propagation in the cable is established, and the signal wave velocity is obtained according to the center frequency of the partial discharge signal, and then the partial discharge source position is located;

[0090] The pruning mean data filtering technology is utilized, the generated multi-local discharge position sample is statistically controlled through the construction of three-layer data filters, and the estimated local discharge positioning error is minimized.

[0091] Specifically, considering that the fixed time window introduces too much redundant noise, which easily causes the cross-correlation function to appear "false peak", affecting the time delay estimation accuracy, an optimized pulse time window technology is proposed, including the following specific technical processes:

[0092] The received sampling signal is preprocessed based on empirical wavelet transform, and the residual noise is power spectrum estimated to obtain the residual noise power, and the separation threshold U of the noise signal and the pulse signal is determined by the following formula thr To accurately identify the partial discharge pulse in the sampling signal:

[0093] ;

[0094] Where, P noise is the residual noise power, k is the proportional coefficient, generally 2;

[0095] The initial and final times of the two partial discharge signal amplitudes just exceeding the threshold are t as , t ae and t bs , t be , and the initial and final time points t ws , t we of the optimized time window are determined by the following formula:

[0096]

[0097] ;

[0098] Where, t r , t f are the time margins of the signal rising edge and falling edge respectively, considering that the GCC method does not need to accurately identify the initial time of the two signals, t r , t f can be empirically set, the rising edge of the partial discharge pulse signal is generally smaller than the falling edge, therefore, when setting the parameters, t r is smaller than t f ;

[0099] The optimized time window width T w is calculated by the following formula:

[0100] ;

[0101] Extract partial discharge pulses x1(t), x2(t) in the two end sampling signals respectively, and perform GCC-PHAT time delay estimation of the two signals in the optimized time window.

[0102] Further, considering that the actual electromagnetic wave velocity is a parameter affected by cable structure parameters, operation environment, electromagnetic wave frequency and other factors, the frequency characteristic function of the electromagnetic wave phase velocity is studied.

[0103] According to the basic parameters of the cable, a frequency-wave velocity characteristic curve of the electromagnetic wave in the cable is drawn, including the following steps:

[0104] The propagation constant of the coaxial cable is:

[0105] ;

[0106] Wherein, R, L and C represent the resistance, inductance and capacitance per unit length respectively, ω is the frequency, and the unit equivalent conductance G can be ignored during normal operation of the cable, so the formula can be approximately expressed as:

[0107] ;

[0108] The ideal unit equivalent inductance L0' is:

[0109] ;

[0110] Wherein, μ0 is the magnetic permeability of the insulation layer between the inner diameter of the cable and the outer diameter of the cable; D A is the outer diameter of the cable, D C is the inner diameter of the cable;

[0111] The actual equivalent inductance corresponding to the outer semi-conductive layer and the inner semi-conductive layer of the cable is:

[0112]

[0113] ;

[0114] Wherein, δ A is the skin depth of the outer semi-conductive layer of the cable, δ C is the skin depth of the inner semi-conductive layer of the cable, and the formula is:

[0115] ;

[0116] Wherein, σ is the conductivity;

[0117] The actual unit equivalent inductance of the cable is:

[0118]

[0119] ;

[0120] The unit equivalent resistance R0 is expressed as

[0121] ;

[0122] We get:

[0123] ;

[0124] Where,

[0125] ;

[0126] The Taylor expansion of the above formula, the first two terms can be obtained:

[0127] ;

[0128] Also

[0129] ;

[0130] From the relationship between wave velocity and β, we get:

[0131] ;

[0132] According to the above formula, the frequency-wave velocity characteristic curve of electromagnetic wave in the cable is drawn, and the corresponding wave velocity is obtained by measuring the center frequency of electromagnetic wave. The problem of inaccurate reference wave velocity set by human is solved.

[0133] In view of the problems of more abnormal data and large positioning error in engineering practice, a pruning mean data filtering algorithm is proposed. By constructing three layers of data filter to eliminate the detected bad data, the accurate estimation of multi-sample partial discharge positioning result is realized;

[0134] Among them, the three-layer data filter includes pruning filter, clustering filter and average filter. The pruning filter eliminates outliers by pruning and serves as the first layer filter of positioning samples. The partial discharge positioning result samples are sorted from the minimum value to the maximum value of the partial discharge position by the pruning filter, and 25%-75% of the samples in the positioning sample are selected as the input of the clustering filter. The maximum density clustering (MDCA) algorithm is used as the second layer clustering filter, and finally the average filter is used to calculate the average value of the positioning samples in the cluster as the final cable partial discharge online positioning result.

[0135] The density maximum clustering algorithm introduces the division clustering based on the density idea, uses the density instead of the initial point as the basis for studying the cluster attribution, and can automatically determine the cluster number and find clusters of arbitrary shape;

[0136] The specific process of the density maximum clustering algorithm is as follows:

[0137] The i-th data point in the sample set is calculated as d c The local density p i of the i-th data point is calculated as p

[0138] ;

[0139] Wherein:

[0140] ;

[0141] The high local density point distance d of each data point is calculated as d

[0142] ;

[0143] In the p-d map, the data points with larger local density p and larger high local density point distance d are selected as cluster centers, and the data points with smaller local density p and larger high local density point distance d are selected as data outliers.

[0144] In the specific implementation process, as shown in Figure 2 , in the experiment, a cable with a length of 90 meters and a model of ZR-YJV26 / 35kV is used, and a 30mm long, 2mm wide and 1mm deep surface scratch defect is set at a distance of 27 meters from the cable A end. The high-frequency current transformer (HFCT) clamps the ground lead of the cable shielding layer, and the partial discharge signal emitted by the cable defect is detected by the HFCT and sent to the partial discharge detector double port through the RG316 coaxial line. As shown in Figure 4 , the measurement system for the partial discharge positioning experiment is built.

[0145] Further, the empirical wavelet transform is used for denoising preprocessing of the sampling data, and the time domain waveform graph and the frequency spectrum graph of the processed partial discharge signal are shown in Figures 5(1) and 5(2), Figures 6(1) and 6(2). As shown in Figures 6(1) and 6(2), the frequency coverage of the measured partial discharge signal is relatively dispersed, which makes it difficult to determine the center frequency. Therefore, wavelet decomposition is considered to decompose the partial discharge signal into each frequency band, read the center frequency of each frequency band, and determine the final signal center frequency through weighted average operation. The center frequency of the measured A-end partial discharge signal spectrum is 4.375MHz, and the center frequency of the B-end partial discharge signal spectrum is 4.398MHz. According to the basic parameters of the experimental cable, the corresponding frequency-wave velocity characteristic curve is drawn, as shown in Figures 5(1) and 5(2). The measured partial discharge center frequency estimates the frequency-varying wave velocity of the partial discharge signal as vA=1.7791×10 8 m / s, vB=1.7795×10 8m / s.

[0146] Based on the optimized pulse time window technique, the time delay of the double-ended partial discharge pulse pair is estimated using GCC-PHAT, such as... Figure 8 As shown in the table below. To reduce the impact of the sampling rate on the discrimination of the maximum time delay estimation of the correlation function, five sampling points were selected near the peak of the correlation function to fit the peak curve and obtain the most accurate time delay estimate. The final single-sample localization results of the partial discharge experiment are shown in the table below.

[0147] Single sample localization results

[0148]

[0149] Data analysis was performed using a trimmed mean data filtering algorithm. The 100 location result samples were normalized. The first layer of the trimmed mean data filtering technique removed location estimates significantly greater or less than the partial discharge amount, retaining the sampling results between the lower and upper quartiles. This yielded a location result-characteristic amplitude feature map of the partial discharge signal, as shown below. Figure 9 As shown, the density maximum clustering (MDCA) algorithm was used to divide the sampling results into three clusters and identify four data outliers. The cluster with the largest number of samples was identified as the true partial discharge source location sample set. After averaging, the final partial discharge location result was 26.58m, with a relative location error of 0.47%, which meets the requirements of engineering practice.

[0150] To compare and analyze with other time-delay estimation localization methods, the peak method, energy method, and CC method were used to analyze the localization of sampled partial discharge signals. All four methods employed the pruned mean data filtering algorithm proposed in this paper for data processing. The final localization results are shown in the table below.

[0151] Comparison of localization results from different methods

[0152]

[0153] As shown in the table above, the experimental positioning results of the four methods are consistent with the simulation analysis results. The peak method, energy method, and CC method all fail to consider the frequency characteristics of the partial discharge wave velocity, leading to errors in the positioning results. The proposed method, however, accurately estimates the partial discharge wave velocity by plotting the corresponding frequency-wave velocity characteristic curve of the experimental cable. Furthermore, the optimized pulse window technique adaptively adjusts the correlation function window size based on the signal-to-noise ratio and device sampling rate, minimizing the probability of "distorted peaks" and "false peaks" in the correlation function and improving the accuracy of time delay estimation.

[0154] Those skilled in the art should understand that the present application is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only to illustrate the principles of the present application, and various changes and improvements can be made without departing from the spirit and scope of the present application, and all such changes and improvements fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A power distribution cable partial discharge on-line locating method based on an improved cross-correlation algorithm, characterized in that, The method comprises the following steps: The partial discharge signals x1(t) and x2(t) collected at both ends of the cable are preprocessed by empirical wavelet transform to reduce data noise; A time window covering complete partial discharge pulse information is obtained, the size of the time window of the correlation function is adaptively adjusted according to the relative position of the partial discharge source in the cable, and the GCC-PHAT algorithm is used to estimate the signal time delay; A frequency-wave velocity curve of electromagnetic wave propagation in the cable is established, and the signal wave velocity is obtained according to the center frequency of the partial discharge signal, and then the position of the partial discharge source is located; The generated multiple partial discharge position samples are statistically controlled by constructing a three-layer data filter to minimize the estimated partial discharge positioning error by using the pruning mean data filtering technology; An optimized pulse time window technology is used to obtain a time window covering complete partial discharge pulse information, and the optimized pulse time window technology comprises the following specific technical process: The separation threshold U of the noise signal and the pulse signal is determined by the following formula thr To accurately identify the partial discharge pulse in the sampling signal: ； where P noise is the residual noise power, and k is a proportionality constant with a value of 2. The start and end times of the two partial discharge signals whose amplitudes just exceed the threshold value are estimated as t as , t ae , t bs , t be , and the start and end time points t ws , t we of the optimized time window are determined by the following formula: ； Wherein, t r , t f are the time margins of the rising and falling edges of the respective signals, and t r is set to be less than t f . The optimization time window width T is calculated by the following equation w : ； The three-layer data filter comprises a pruning filter, a clustering filter and an average filter, the pruning filter serves as the first layer filter of the positioning sample by pruning and removing outliers, the MDCA algorithm is used as the second layer clustering filter, and finally the average filter is used to calculate the average value of the positioning samples in the cluster as the final online positioning result of the cable partial discharge. The received sampling signal is preprocessed based on empirical wavelet transform, and the residual noise is power spectrum estimated to obtain the residual noise power.

2. The power distribution cable partial discharge online locating method based on the improved cross-correlation algorithm according to claim 1, characterized in that, The partial discharge signals x1(t)' and x2(t)' at both ends of the cable are extracted, and the GCC-PHAT time delay estimation of the two signals is performed in the optimized time window.

3. The power distribution cable partial discharge online locating method based on the improved cross-correlation algorithm according to claim 1, characterized in that, According to the basic parameters of the cable, a frequency-wave velocity characteristic curve of electromagnetic wave in the cable is drawn, which comprises the following steps:

4. The power distribution cable partial discharge online locating method based on the improved cross-correlation algorithm according to claim 1, characterized in that, The propagation constant of the coaxial cable is: Wherein, the unit equivalent conductance G can be ignored when the cable is in normal operation, so the formula can be approximately expressed as: ； The ideal unit equivalent inductance L0' is: ； The actual equivalent inductance of the outer semi-conductive layer and the inner semi-conductive layer of the cable is: ； wherein μ0 is the permeability of the insulation layer between the inner diameter of the cable and the outer diameter of the cable; D A is the outer diameter of the cable, and D C is the inner diameter of the cable. Wherein, σ is the conductivity; ； where δ A is the skin depth of the outer semiconductive layer of the cable, δ C is the skin depth of the inner semiconductive layer of the cable, and is given by ； The actual unit equivalent inductance of the cable is: The unit equivalent resistance R0 is expressed as ； It is obtained that: ； Wherein, ； The Taylor expansion of the above formula is obtained by retaining the first two terms: ； And ； The relationship between the wave velocity and β is: ； According to the above formula, a frequency-wave velocity characteristic curve of electromagnetic wave in the cable is drawn, and the corresponding wave velocity is obtained by measuring the center frequency of the electromagnetic wave. ； The partial discharge positioning result samples are sorted from the minimum value to the maximum value of the partial discharge position by the pruning filter through data normalization processing of a plurality of partial discharge position samples, and 25%-75% of the positioning samples are selected as the input of the clustering filter; 5. The power distribution cable partial discharge online locating method based on the improved cross-correlation algorithm according to claim 1, characterized in that, In the screening process, the positioning estimate value obviously greater or less than the partial discharge amount is removed, and the sampling results between the lower quartile and the upper quartile are retained to obtain the positioning result-characteristic graph of the partial discharge signal. ​ 6. The power distribution cable partial discharge online locating method based on improved cross-correlation algorithm according to claim 1, characterized in that, The density maximum value clustering algorithm introduces the idea of density into division clustering, divides the input sampling results into multiple clusters, and identifies data abnormal points; The specific process of the density maximum value clustering algorithm is as follows: The i-th data point in the sample set is computed as d c the local density p i : ； Wherein: ； Calculate the distance δi of each data point to the high local density point: ； In the ρ-δ graph, select the data points with large local density ρ and large high local density point distance δ as cluster centers, and select the data points with small local density ρ and large high local density point distance δ as data abnormal points.

7. An improved cross-correlation algorithm based on-line partial discharge positioning system for power cable, applied to the improved cross-correlation algorithm based on-line partial discharge positioning method of any one of claims 1-6, characterized in that, Including: At least two groups of discharge signal sensors are arranged at both ends of the measured cable for collecting partial discharge signals of the measured cable; At least two monitoring terminals are connected with the at least two groups of discharge signal sensors, for receiving the partial discharge signals and acquiring pulse waveform data; A network switch is connected with the at least two monitoring terminals, for uploading the pulse waveform data and the partial discharge signals; A diagnosis platform is connected with the network switch, for performing correlation analysis on the received pulse waveform data and the partial discharge signals, and determining the partial discharge position.

Images