CN-121978476-A - High-voltage cable partial discharge on-line monitoring method and device

CN121978476ACN 121978476 ACN121978476 ACN 121978476ACN-121978476-A

Abstract

The invention discloses a high-voltage cable partial discharge on-line monitoring method, and relates to the technical field of on-line monitoring. The method comprises the steps of embedding a distributed optical fiber sensing unit in a high-voltage cable sheath layer, converting physical field change data generated by partial discharge into an initial partial discharge optical signal sequence, using a wavelet transformation algorithm to reduce noise of the initial partial discharge optical signal sequence, extracting features to obtain partial discharge feature vectors, constructing a partial discharge classification model, inputting the feature vectors, outputting classification results and time-space information, using a DBSCAN algorithm to perform time-space clustering according to the classification results and the time-space information, generating a time-space evolution path, constructing a discharge development trend prediction model based on an ARIMA model, inputting time-space evolution path feature parameters, outputting the prediction results, calculating partial discharge risk values, and performing hierarchical early warning. The invention realizes the on-line monitoring of the partial discharge of the high-voltage cable based on the partial discharge risk value.

Inventors

MEI HAIJUN
MEI SHENGWEI
YANG DUNGAO
LI CHUNYAN
LIU YI
ZHANG FENG
Wu sai

Assignees

武汉比邻科技发展有限公司

Dates

Publication Date: 20260505
Application Date: 20260122

Claims (10)

1. The method for monitoring the partial discharge of the high-voltage cable on line is characterized by comprising the following steps of: Step S1, embedding a distributed optical fiber sensing unit in a high-voltage cable sheath layer, and converting physical field change data generated when a partial discharge event occurs into an initial partial discharge optical signal sequence through the distributed optical fiber sensing unit; S2, carrying out noise reduction treatment on the initial partial discharge optical signal sequence through a wavelet transformation algorithm to obtain a noise reduction partial discharge optical signal sequence, and carrying out feature extraction on the noise reduction partial discharge optical signal sequence to obtain a partial discharge feature vector; S3, constructing a partial discharge classification model based on a one-dimensional convolutional neural network 1D-CNN model, inputting the partial discharge feature vector into the partial discharge classification model, and outputting to obtain a partial discharge classification result and space-time information of a partial discharge event; s4, based on the partial discharge classification result and the space-time information of the partial discharge event, performing space-time clustering on the partial discharge event by using a spatial clustering DBSCAN algorithm to generate a space-time evolution path of the partial discharge; and S5, constructing a discharge development trend prediction model based on an autoregressive integral moving average model, inputting characteristic parameters of the space-time evolution path into the discharge development trend prediction model, outputting a discharge development trend prediction result, calculating a local discharge risk value according to the discharge development trend prediction result, and carrying out hierarchical early warning on a local discharge event based on the local discharge risk value.
2. The method for online monitoring of partial discharge of a high-voltage cable according to claim 1, wherein the step of embedding the distributed optical fiber sensing unit in the high-voltage cable sheath layer comprises the steps of: In the manufacturing process of the sheath layer of the high-voltage cable, the distributed optical fiber sensing unit is directly embedded, and tight physical contact and signal coupling with the cable metal shielding layer are realized through the conductive adhesive layer; The distributed optical fiber sensing unit adopts a four-layer composite structure design from outside to inside, wherein the outermost layer is a nickel plating fiber woven shielding layer, the middle layer is a polyimide electric erosion resistant coating, a heat conducting silicon rubber layer is arranged between the outermost layer and the middle layer, and the innermost layer is a standard single-mode fiber core.
3. The method for on-line monitoring of partial discharge of a high voltage cable according to claim 2, wherein the step of converting the data of the change of the physical field generated when the partial discharge event occurs into the initial partial discharge optical signal sequence comprises the steps of: Nanosecond laser pulse is emitted to the optical fiber, the back scattering signal is detected, and the phase change quantity delta phi of Rayleigh scattering light and the change quantity delta of Rayleigh scattering light intensity are extracted Brillouin scattering frequency shift change amount ; Marking each emitted laser pulse with an absolute timestamp by a high precision clock When in the optical fiber position Recording the absolute time stamp of receipt of the scattered signal when a scattering event is detected Light is emitted from the position The round trip time of returning to the receiving end is ; Finally, the Rayleigh scattered light phase change quantity delta phi and the Rayleigh scattered light intensity change quantity delta Brillouin scattering frequency shift change amount Correlating with corresponding space-time coordinates (t, x) to construct an initial partial discharge optical signal sequence with the space-time coordinates 。
4. The method for online monitoring of partial discharge of a high voltage cable according to claim 3, wherein the step of performing noise reduction processing on the initial partial discharge optical signal sequence by a wavelet transform algorithm to obtain a noise reduction partial discharge optical signal sequence comprises the steps of: First to the initial partial discharge optical signal sequence In terms of space coordinates Grouping, each of Corresponds to a three-dimensional time sequence Performing wavelet decomposition and noise reduction on each time sequence to obtain wavelet coefficients ; Then, threshold processing is carried out to obtain detail coefficients of each layer through decomposition Adaptive thresholding using unbiased risk estimation thresholding, thresholding using the method Determined by the following formula: ; Wherein, the Is the standard deviation of noise, which is formed by the detail coefficient of the first layer Is used for the median estimation of (1), Is the signal length; Subsequently, soft threshold function processing coefficients are applied: ; Wherein, the Is after soft threshold processing Reconstruction detail coefficients of the layer; Is after wavelet decomposition, the first Original detail coefficients of the layer; Is a sign function; finally, wavelet reconstruction is carried out, and detail coefficients after threshold processing are processed Approximation coefficient with retention Performing wavelet inverse transformation and reconstructing to obtain a noise-reducing partial discharge optical signal sequence 。
5. The method for online monitoring of partial discharge of a high-voltage cable according to claim 4, wherein the feature extraction of the noise-reduced partial discharge optical signal sequence to obtain a partial discharge feature vector comprises: first, the Rayleigh scattering light intensity peak value is extracted This feature is directly derived from the noise-reduced Equal to the measured intensity at discharge Reference intensity to no discharge Is the difference of the reference strength Calibrated by initialization time, peak value Is the pulse duration In the inner part of the inner part, Is the maximum value of (2); second, phase change peak value is extracted Derived directly from The extraction method comprises detecting peak offset of phase pulse waveform ; Furthermore, the duration of the extraction pulse ; The Rayleigh scattering pulse energy is then extracted Also from noise reduction The extraction method is to calculate the integral energy of the pulse waveform in the duration: ; Wherein, the The variation of the intensity of the noise-reducing Rayleigh scattering light; Finally, extracting the Brillouin frequency shift offset Directly from the noise reduction Brillouin scattering light frequency shift variable quantity Capturing a frequency shift pulse peak value synchronous with a partial discharge event during extraction, wherein a calculation formula is as follows: ; Wherein, the Is the brillouin shift offset peak, Noise reduction Brillouin scattering light frequency shift variable quantity; finally, a corresponding partial discharge characteristic vector is generated for each partial discharge event 。
6. The method for online monitoring of partial discharge of a high voltage cable according to claim 5, wherein step S3 comprises: firstly, preprocessing input data, reconstructing five-dimensional feature vectors output in the step S2 into tensor format suitable for network processing, and reconstructing the feature vectors Converting into a preset data structure, and performing standardization processing to obtain preprocessed input data ; Then, entering a core feature extraction stage, and sequentially carrying out deep feature learning through two residual error modules; the first residual error module processing flow inputs the characteristics Firstly, extracting a local characteristic mode through convolution operation: ; Wherein, the Is a feature map of the output of the first convolutional layer; The convolution kernel weight is responsible for capturing local correlation among features; Is a bias term for adjusting the output distribution; is a convolution operation symbol; then, carrying out batch normalization processing on the convolution output, and stabilizing the training process; Finally, feature fusion is realized through residual connection: ; Wherein, the Is the final output of the first residual block; A composite function representing volume and batch normalization; ( ) Is a linear rectifying activation function; An original input of the residual error module; and the second residual error module processing flow takes the output of the first module as the input of the second module, so as to realize deep abstraction of the features: ; Wherein, the Is the input to the second module; Finally, the classification decision is completed through the full connection layer and the Softmax function, and the output of the second residual error module is outputted Flattening and then sending the flattened material into a full-connection layer to obtain the original score of each category; then converting the original score into probability distribution through a Softmax function; Using the category corresponding to the probability maximum value as a prediction result The probability value is used as confidence When confidence is When the recognition result is determined to be a high confidence recognition result, the method directly enters And (5) a flow.
7. The method for online monitoring of partial discharge of a high voltage cable according to claim 6, wherein the spatial clustering of partial discharge events is performed by using a spatial clustering DBSCAN algorithm to generate a spatial-temporal evolution path of partial discharge, comprising: using three-dimensional coordinates of discharge events Space-time clustering is carried out by adopting an improved spatial clustering DBSCAN algorithm: ; Wherein, the Represent the first The number of the discharge clusters, the space neighborhood radius 2m ensures the geographic relevance, and the time window 10s ensures the event continuity; Is the kth point in the cluster The time of arrival of the signal at points i and j, respectively; A spatial coordinate point representing two partial discharge events; And A type tag representing a discharge event; Finally, based on the clustering result, three key characteristic parameters of each discharge cluster are calculated, and the spatial distribution density is calculated Average discharge frequency Amplitude of disturbance of optical signal ; Then, by analyzing the time-space distribution rule of the events in the clusters, the propagation direction vector is calculated And evolution acceleration ; Finally, these information of all discharge clusters together constitute a complete space-time evolution path G of partial discharges.
8. The method for online monitoring of partial discharge of high-voltage cable according to claim 7, wherein the three-dimensional coordinates of the discharge event The determination of (2) comprises: For each discharge event, the three-dimensional coordinates of which have been determined by step S1, the spatial position is determined by the fiber coordinates Give elevation information Determining plane coordinates of the cable path diagram by combining the cable laying digital data to finally obtain three-dimensional coordinates of the discharge event 。
9. The method for online monitoring of partial discharge of a high-voltage cable according to claim 8, wherein the method for online monitoring of partial discharge of the high-voltage cable is characterized by constructing a discharge development trend prediction model based on an autoregressive integral moving average model, inputting characteristic parameters of the space-time evolution path into the discharge development trend prediction model, and outputting and obtaining a discharge development trend prediction result, and comprises the following steps: First according to the discharge type Grouping the discharge cluster set G to obtain a cluster subset of the same type Dividing the time axis into equidistant windows For each window Aggregate all cluster features it covers: ; Wherein, the Is a window The clusters of the coverage are clustered and, ; Is the average discharge density of the ith time window; is the average discharge frequency of the ith time window, the average discharge amplitude of the ith time window ; Arranging all windows in time sequence Forming a multivariate time series ; The model is improved by adopting Support Vector Regression (SVR), and the linear fitting capacity of ARIMA is combined with the nonlinear mapping capacity of support vector regression: ; Wherein, the Is a discharge development trend prediction model at a time point Predictive value of ARIMA ) Is ARIMA component and SVR # , ) Is an SVR component; At the moment for ARIMA component Is a prediction residual of (2); For the moment of time Is a feature gradient vector of (1); The discharge development trend prediction model training adopts a two-stage optimization strategy, firstly, an optimal parameter combination (p, d, q) of ARIMA is determined through AIC criteria, and then grid search is used for optimizing penalty parameters of SVR And core parameters ; After training is completed, the future is synchronously predicted Multidimensional feature values at individual time points : ; Wherein, the Is a sequence of spatial distribution densities of discharge events over time over a predicted future period of time; is a predicted sequence of average discharge frequencies over time for discharge events over a future period of time; Is a sequence of predicted changes in the amplitude of the optical signal disturbance of the discharge event over time within a future period of time; is the prediction step size.
10. A high voltage cable partial discharge on-line monitoring device comprising a memory, a processor and a computer program stored on the memory, characterized in that the processor executes the computer program to carry out the steps of the method according to any one of claims 1-9.

Description

High-voltage cable partial discharge on-line monitoring method and device Technical Field The invention relates to the field of high-voltage cable online monitoring engineering, in particular to a method and a device for online monitoring of partial discharge of a high-voltage cable. Background The reliability of the insulating state of the high-voltage cable serving as key equipment for urban power grid and remote power transmission is directly related to the safe and stable operation of a power supply system. Partial discharge is an important phenomenon for characterizing cable insulation degradation, and the activity characteristics of the partial discharge are closely related to the type and severity of insulation defects. Therefore, the cable partial discharge is effectively monitored, and the method is an important means for realizing state maintenance and preventing faults. The traditional high-voltage cable partial discharge monitoring method adopts a high-frequency current transformer coupling method. According to the method, the high-frequency current transformer is arranged at the cable grounding wire or the joint, the high-frequency current signal generated by partial discharge is coupled, and the characteristics of amplitude, frequency and the like are analyzed by a pulse current method, so that the discharge severity is judged. However, the method has inherent bottleneck in the aspects of feature extraction and pattern recognition, lacks automatic learning capability on deep features of original signals, cannot effectively extract abstract feature combinations highly related to discharge types in signals, and is difficult to establish complex nonlinear classification boundaries, so that the partial discharge on-line monitoring result of the high-voltage cable is inaccurate. Disclosure of Invention Aiming at the defects of the prior art, the invention provides an online monitoring method for the partial discharge of a high-voltage cable, which solves the problem of inaccurate online monitoring result of the partial discharge of the high-voltage cable. In order to achieve the purpose, the invention is realized by the following technical scheme that the method for monitoring the partial discharge of the high-voltage cable on line comprises the following steps: Step S1, embedding a distributed optical fiber sensing unit in a high-voltage cable sheath layer, and converting physical field change data generated when a partial discharge event occurs into an initial partial discharge optical signal sequence through the distributed optical fiber sensing unit; S2, carrying out noise reduction treatment on the initial partial discharge optical signal sequence through a wavelet transformation algorithm to obtain a noise reduction partial discharge optical signal sequence, and carrying out feature extraction on the noise reduction partial discharge optical signal sequence to obtain a partial discharge feature vector; S3, constructing a partial discharge classification model based on a one-dimensional convolutional neural network 1D-CNN model, inputting the partial discharge feature vector into the partial discharge classification model, and outputting to obtain a partial discharge classification result and space-time information of a partial discharge event; s4, based on the partial discharge classification result and the space-time information of the partial discharge event, performing space-time clustering on the partial discharge event by using a spatial clustering DBSCAN algorithm to generate a space-time evolution path of the partial discharge; and S5, constructing a discharge development trend prediction model based on an autoregressive integral moving average model, inputting characteristic parameters of the space-time evolution path into the discharge development trend prediction model, outputting a discharge development trend prediction result, calculating a local discharge risk value according to the discharge development trend prediction result, and carrying out hierarchical early warning on a local discharge event based on the local discharge risk value. Preferably, the high-voltage cable sheath layer is embedded into the distributed optical fiber sensing unit, and the distributed optical fiber sensing unit comprises: In the manufacturing process of the sheath layer of the high-voltage cable, the distributed optical fiber sensing unit is directly embedded, and tight physical contact and signal coupling with the cable metal shielding layer are realized through the conductive adhesive layer; The distributed optical fiber sensing unit adopts a four-layer composite structure design from outside to inside, wherein the outermost layer is a nickel plating fiber woven shielding layer, the middle layer is a polyimide electric erosion resistant coating, a heat conducting silicon rubber layer is arranged between the outermost layer and the middle layer, and the innermost layer is a standard single-mode fib