Power cable fracture prediction method based on sound wave analysis

Abstract

The invention discloses a power cable fracture prediction method based on sound wave analysis. The method comprises the following steps: acquiring sound wave signal data of a power cable based on a sound wave detection system; carrying out preprocessing and feature extraction on the sound wave signal; constructing a fracture prediction model by using a machine learning algorithm; and monitoring the state of the power cable in real time and predicting the potential fracture risk. According to the method, the sound wave analysis technology and the machine learning algorithm are combined, the fracture condition of the power cable can be efficiently and accurately predicted, early warning can be carried out before fracture occurs, and power accidents are avoided. The method has the advantages of high precision, real-time performance, automatic prediction and the like.

Description

Technical Field

[0001] The present invention relates to the fields of machine learning and audio signal processing, and in particular to a method for predicting power cable breakage based on acoustic wave analysis. Background Art

[0002] With the increasing demand for electricity in modern society, the safe operation of transmission lines has become a key factor in ensuring the stable development of social economy. Traditional transmission line inspection mainly relies on manual methods, which is not only labor-intensive and inefficient, but also has certain safety hazards and management difficulties. In order to improve the efficiency and accuracy of transmission line inspection, an automated monitoring method based on acoustic wave analysis and machine learning technology has emerged.

[0003] During operation, power cables may break due to external forces, environmental factors, and service life. Traditional monitoring methods mainly rely on regular manual inspections and simple physical detection methods, which make it difficult to achieve real-time and accurate prediction of breakage. Acoustic wave analysis technology can monitor the health status of power cables in real time by capturing and analyzing acoustic wave signals in power cables. However, acoustic wave signals are complex and variable, and need to be finely analyzed in combination with machine learning algorithms to improve the accuracy of breakage prediction.

[0004] At present, the power cable monitoring method based on acoustic wave analysis has achieved certain research results, but there are still some challenges in practical application, such as complex signal processing and insufficient model training data. In order to overcome these difficulties, a power cable fracture prediction method based on acoustic wave analysis is proposed. By preprocessing and extracting features of acoustic wave signals, a fracture prediction model is constructed using machine learning algorithms, and then verified and optimized in a practical environment, so as to realize real-time monitoring and fracture prediction of power cables. Summary of the invention

[0005] In order to improve the accuracy and real-time performance of power cable break prediction, the present invention provides a power cable break prediction method based on acoustic wave analysis. This method combines acoustic wave sensors and machine learning technology to achieve accurate prediction of power cable breakage through in-depth analysis of acoustic wave signals, thereby ensuring the safe operation of transmission lines.

[0006] To achieve the above object, the present invention discloses a method for predicting power cable breakage based on acoustic wave analysis, comprising the following steps:

[0007] Step S10: Acquiring acoustic wave signal data of the power cable based on the acoustic wave detection system;

[0008] Step S101: According to different types of power cables, Rayleigh wave sensors and optical fiber sensors are installed on power towers or power lines to form an acoustic wave detection system to collect audio signals in real time;

[0009] Step S20: preprocessing and feature extraction of the acoustic wave signal;

[0010] Step S201: First, the collected sound wave signal is filtered and denoised, and the operation formula is as follows:

[0011] y(t)=x(t)*h(t)

[0012] Wherein, * represents convolution calculation, x(t) is the collected original signal, h(t) is the transfer function of the bandpass filter, and the filtered signal is y(t).

[0013] Use wavelet transform to denoise, decompose the signal by wavelet, retain important signal components and remove noise components. Suppose the discrete wavelet transform of the signal is W x (a,b), then the denoised signal for:

[0014]

[0015] in, is the wavelet basis function.

[0016] Step S202: For the pre-processed sound wave signal, first extract its time domain features and frequency domain features. The formula for extracting the time domain features is as follows:

[0017]

[0018]

[0019] Among them, y(t) is the preprocessed signal, N is the signal length, i is the moment in the signal, μ(t) is the mean of the preprocessed sound wave signal, and σ(t) is the variance of the preprocessed sound wave signal.

[0020] The extracted frequency domain features are shown as follows:

[0021]

[0022] Fast Fourier transform is used to extract frequency domain features. Y(f) is the fast Fourier transform result of the signal, y(n) is the sampled signal of the preprocessed signal y(t), and N is the sampling length.

[0023] Step S30: constructing a power cable break prediction model using a machine learning algorithm;

[0024] Step S301: Collect and annotate a large amount of acoustic wave signal data in normal and broken states to construct a power cable break prediction data set.

[0025] Step S302: Use principal component analysis (PCA) to select and reduce the dimension of the extracted features, select the most representative features, reduce the feature dimension, and improve the model training efficiency. The method formula is as follows:

[0026] Z=XW

[0027] Among them, X is the extracted feature matrix, W is the principal component matrix, and Z is the feature after dimensionality reduction.

[0028] Step S303: designing a multi-layer perceptron (MLP) neural network for power cable break prediction, wherein the network structure includes an input layer, multiple hidden layers and an output layer.

[0029] For each layer of neurons, let the input of layer I be a (l) , the weight matrix is ​​W (l) , the bias vector is b (l) , the activation function is σ, then the input of the l+1th is:

[0030] z (l+1) =W (l) a (l) +b (l)

[0031] a (l+1) =σ(z (l+1) )

[0032] The mean square error (MSE) is used as the loss function to calculate the difference between the predicted value and the actual value. Suppose the predicted value is The actual value is y, then the loss function is:

[0033]

[0034] Use the back propagation algorithm to calculate the gradient of the loss function relative to the weights and biases, and update the weights and biases. Assuming the learning rate is α, the update formula for the weights and biases is:

[0035]

[0036]

[0037] Step S303: Use the power cable break prediction data set constructed in S301 to perform model training, and iteratively update weights and biases until the loss function converges or reaches a preset number of training times.

[0038] Step S40: Real-time detection of power cable status and prediction of potential breakage risks;

[0039] Step S401: Acquire the acoustic wave signal data of the power cable through the acoustic wave detection system of step S10, and perform preprocessing and feature extraction on the acoustic wave signal in step S20.

[0040] Step S401: input the extracted signal features into the optimized power cable fracture prediction model to detect the risk of power cable fracture in real time. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 The present invention is a flowchart of the steps of detecting power cable breakage based on acoustic wave analysis.

[0042] Figure 2 This is a flow chart of the algorithm training steps of the power cable break detection neural network of the present invention. DETAILED DESCRIPTION

[0043] The following describes the implementation of the present invention through specific examples and in conjunction with the accompanying drawings. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific examples, and the details in this specification can also be modified and changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention.

[0044] Step S10: Acquiring acoustic wave signal data of the power cable based on the acoustic wave detection system;

[0045] Step S101: According to different types of power cables, Rayleigh wave sensors and optical fiber sensors are installed on power towers or power lines to form an acoustic wave detection system to collect audio signals in real time;

[0046] Step S20: preprocessing and feature extraction of the acoustic wave signal;

[0047] Step S201: First, the collected sound wave signal is filtered and denoised, and the operation formula is as follows:

[0048] y(t)=x(t)*h(t)

[0049] Wherein, * represents convolution calculation, x(t) is the collected original signal, h(t) is the transfer function of the bandpass filter, and the filtered signal is y(t).

[0050] Use wavelet transform to denoise, decompose the signal by wavelet, retain important signal components and remove noise components. Suppose the discrete wavelet transform of the signal is W x (a,b), then the denoised signal for:

[0051]

[0052] in, is the wavelet basis function.

[0053] Step S202: For the pre-processed sound wave signal, first extract its time domain features and frequency domain features. The formula for extracting the time domain features is as follows:

[0054]

[0055]

[0056] Among them, y(t) is the preprocessed signal, N is the signal length, i is the moment in the signal, μ(t) is the mean of the preprocessed sound wave signal, and σ(t) is the variance of the preprocessed sound wave signal.

[0057] The extracted frequency domain features are shown as follows:

[0058]

[0059] Fast Fourier transform is used to extract frequency domain features. Y(f) is the fast Fourier transform result of the signal, y(n) is the sampled signal of the preprocessed signal y(t), and N is the sampling length.

[0060] Step S30: constructing a power cable break prediction model using a machine learning algorithm;

[0061] Step S301: Collect and annotate a large amount of acoustic wave signal data in normal and broken states to construct a power cable break prediction data set.

[0062] Step S302: Use principal component analysis (PCA) to select and reduce the dimension of the extracted features, select the most representative features, reduce the feature dimension, and improve the model training efficiency. The method formula is as follows:

[0063] Z=XW

[0064] Among them, X is the extracted feature matrix, W is the principal component matrix, and Z is the feature after dimensionality reduction.

[0065] Step S303: designing a multi-layer perceptron (MLP) neural network for power cable break prediction, wherein the network structure includes an input layer, multiple hidden layers and an output layer.

[0066] For each layer of neurons, let the input of layer I be a (l) , the weight matrix is ​​W (l) , the bias vector is b (l) , the activation function is σ, then the input of the l+1th is:

[0067] z (l+1) =W (l)a (l) +b (l)

[0068] a (l+1) =σ(z (l+1) )

[0069] The mean square error (MSE) is used as the loss function to calculate the difference between the predicted value and the actual value. Suppose the predicted value is The actual value is y, then the loss function is:

[0070]

[0071] Use the back propagation algorithm to calculate the gradient of the loss function relative to the weights and biases, and update the weights and biases. Assuming the learning rate is α, the update formula for the weights and biases is:

[0072]

[0073]

[0074] Step S303: Use the power cable break prediction data set constructed in S301 to perform model training, and iteratively update weights and biases until the loss function converges or reaches a preset number of training times.

[0075] Step S40: monitor the power cable status in real time and predict the potential breakage risk;

[0076] Step S401: Acquire the acoustic wave signal data of the power cable through the acoustic wave detection system of step S10, and perform preprocessing and feature extraction on the acoustic wave signal in step S20.

[0077] Step S401: input the extracted signal features into the optimized power cable fracture prediction model to monitor the risk of power cable fracture in real time.

Claims

1. A method for predicting power cable breakage based on acoustic wave analysis, characterized in that: The following steps are involved: S10: Acquiring acoustic wave signal data of the power cable based on the acoustic wave detection system; S20: preprocessing and feature extraction of acoustic wave signals; S30: Constructing fracture prediction model using machine learning algorithm; S40: Real-time monitoring of power cable status and prediction of potential break risks.

2. The method for obtaining acoustic wave signal data of a power cable based on an acoustic wave sensor according to claim 1, characterized in that: The acoustic wave detection system is composed of Rayleigh wave sensors and optical fiber sensors.

3. The method for preprocessing and feature extraction of acoustic wave signals according to claim 1, characterized in that: The collected sound wave signal is preprocessed using filtering and denoising operations, and then the frequency, amplitude and phase characteristic parameters of the preprocessed sound wave signal are extracted.

4. The method of constructing a fracture prediction model using a machine learning algorithm according to claim 1, wherein: A fracture prediction model is constructed using a neural network algorithm. By collecting and annotating a large amount of acoustic wave signal data in normal and fracture states, the prediction model is trained to learn the relationship between the acoustic wave signal characteristics and the fracture state.

5. The method for predicting power cable fracture based on acoustic wave analysis according to claim 1, characterized in that: The model is tested and optimized in laboratory environments and actual power cable environments to ensure the stability and accuracy of the model under different conditions.

Images