Distribution network fault positioning system and method based on traveling wave distance measurement with high anti-interference capability

Abstract

The invention discloses a distribution network fault positioning system and method based on traveling wave distance measurement with high anti-interference capability, and the system comprises an acquisition module which employs a high-speed sampling device disposed at a node of a power distribution network to acquire a sampling frequency and a traveling wave signal, and carries out the preprocessing of the traveling wave signal; the detection module is used for carrying out wavelet transformation on the preprocessed traveling wave signal to obtain a wavelet coefficient, constructing a wavelet coefficient local extreme point model based on a knowledge distillation algorithm, and detecting the traveling wave arrival time by utilizing the wavelet coefficient local extreme point model; the calculation module is used for calculating the traveling wave fault distance in the power distribution network; and the positioning module is used for determining the fault position of the power distribution network by using the traveling wave fault distances of the plurality of traveling waves. According to the invention, the anti-interference capability, the positioning precision and the robustness of fault positioning are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power equipment, and in particular to a distribution network fault location system and method based on traveling wave ranging with high anti-interference capability. Background Art

[0002] Effectively locating faults in distribution networks is of great significance to improving the safe operation level and operation quality of distribution networks. However, the current distribution network fault location methods still have the following shortcomings: 1) Low detection accuracy of the arrival time of traveling waves. Traditional traveling wave detection methods are difficult to accurately capture the arrival time of traveling waves in complex signal environments (such as noise interference, signal attenuation, etc.). This error directly affects the accuracy of fault location. 2) Sensitive to noise. Existing methods are easily disturbed in noisy environments, resulting in false detection or missed detection, which in turn affects the effect of fault location. This problem is more prominent in complex distribution network environments. 3) Lack of adaptability. Traditional methods are usually based on fixed models or rules and are difficult to adapt to dynamic changes in the power grid, such as load fluctuations, topological changes, etc., resulting in reduced positioning accuracy. 4) High computational complexity. When facing large-scale distribution networks or multi-point faults, the computational complexity of traditional positioning methods is high, making it difficult to complete fault location in a short time, affecting real-time performance and efficiency. Summary of the invention

[0003] In order to solve the above problems, the present invention proposes a distribution network fault location system and method based on traveling wave ranging with high anti-interference capability, which can improve the accuracy and speed of traveling wave arrival time detection.

[0004] In order to achieve the above object, the present invention is implemented by the following technical solutions:

[0005] A distribution network fault location system based on traveling wave ranging with high anti-interference capability of the present invention comprises:

[0006] The acquisition module is used to collect the sampling frequency and the traveling wave signal by using the high-speed sampling device arranged at the key nodes of the distribution network, and to pre-process the traveling wave signal;

[0007] The detection module is used to perform wavelet transform on the preprocessed traveling wave signal to obtain wavelet coefficients, and to construct a local extreme point model of the wavelet coefficients based on the knowledge distillation algorithm, and to detect the arrival time of the traveling wave using the local extreme point model of the wavelet coefficients;

[0008] A calculation module for calculating the distance of traveling wave faults in the distribution network;

[0009] The positioning module is used to determine the fault location of the distribution network by using the traveling wave fault distances of multiple traveling waves.

[0010] A further improvement of the present invention is that the traveling wave signal collected by the collection module is:

[0011]

[0012] Among them, x() is the collected traveling wave signal, n is the discrete time index, that is, the time point, T s is the sampling interval, f s is the sampling frequency, used to capture high-speed changing traveling wave signals;

[0013] The expression for preprocessing the traveling wave signal is:

[0014]

[0015] Among them, x filtered (n) is the preprocessed traveling wave signal, h(k) is the impulse response of the filter, and k is the discrete time index.

[0016] A further improvement of the present invention is that the operations performed by the detection module include:

[0017] Perform wavelet transform on the preprocessed traveling wave signal to obtain the wavelet coefficient, which is expressed as:

[0018]

[0019] Among them, W(a,b) is the wavelet coefficient, is the mother wavelet function, a and b are the scale factor and time smoothing factor respectively, x filtered (n) is the preprocessed traveling wave signal, n is the discrete time index, i.e., the time point;

[0020] Based on the knowledge distillation algorithm, a wavelet coefficient local extreme point model is constructed, where the wavelet coefficient local extreme point model includes a teacher model, a student model and a loss function; the teacher model is used to process the wavelet coefficient data and detect local extreme points, and the output is the wavelet transform coefficient W teacher (a, b) The predicted value under different scale factors a and time smoothing factors b; the expression of the loss function is:

[0021] L total =aL distill +βL standard (4)

[0022]

[0023]

[0024] Among them, L distill is the knowledge distillation loss, L standard is the standard loss, L totalis the total loss, W teacher (a, b) are the wavelet transform coefficients output by the teacher model, W student (a, b) are the wavelet transform coefficients output by the student model, N is the number of samples in the training data set, is the predicted value of the student model, indicating whether there is a local extreme value at time point n, where 1 means yes and 0 means no. y(n) is the actual position of the local extreme point. α and β are weight coefficients, which are used to balance the influence of knowledge distillation loss and standard loss respectively.

[0025] Through back propagation and gradient descent method, the student model is trained so that the total loss L total Minimize and use the Adam optimization algorithm to update the parameters of the student model. The expression is:

[0026]

[0027] Among them, θ student are the parameters of the student model, is the gradient descent function, η is the gradient descent coefficient;

[0028] According to the parameters of the updated student model, an extreme point prediction model is constructed to predict local extreme points. That is, predict the arrival time of the traveling wave, the expression is:

[0029]

[0030] in, is a local extreme point, is the extreme point prediction model, x() is the collected traveling wave signal, and n is the discrete time index, that is, the time point.

[0031] A further improvement of the present invention is that the operations performed by the detection module include:

[0032] Perform wavelet transform on the preprocessed traveling wave signal to obtain the wavelet coefficient, which is expressed as:

[0033]

[0034] Among them, W(a,b) is the wavelet coefficient, is the mother wavelet function, a and b are the scale factor and time smoothing factor respectively, x filtered (n) is the preprocessed traveling wave signal, n is the discrete time index, i.e., the time point;

[0035] Based on the knowledge distillation algorithm, a wavelet coefficient local extreme point model is constructed, where the wavelet coefficient local extreme point model includes a teacher model, a student model and a loss function; the teacher model is used to process the wavelet coefficient data and detect local extreme points, and the output is the wavelet transform coefficient Wteacher (a, b) The predicted value under different scale factors a and time smoothing factors b; the expression of the loss function is:

[0036] L total =aL distill +βL standard (4)

[0037]

[0038]

[0039] Among them, L distill is the knowledge distillation loss, L standard is the standard loss, L total is the total loss, W teacher (a, b) are the wavelet transform coefficients output by the teacher model, W student (a, b) are the wavelet transform coefficients output by the student model, N is the number of samples in the training data set, is the predicted value of the student model, indicating whether there is a local extreme value at time point n, where 1 means yes and 0 means no. y(n) is the actual position of the local extreme value point. α and β are weight coefficients, which are used to balance the influence of knowledge distillation loss and standard loss respectively.

[0040] Through back propagation and gradient descent method, the student model is trained so that the total loss L total Minimize and use the Adam optimization algorithm to update the parameters of the student model. The expression is:

[0041]

[0042] Among them, θ student are the parameters of the student model, is the gradient descent function, η is the gradient descent coefficient;

[0043] According to the parameters of the updated student model, an extreme point prediction model is constructed to predict local extreme points. That is, predict the arrival time of the traveling wave, the expression is:

[0044]

[0045] in, is a local extreme point, is the extreme point prediction model, x() is the collected traveling wave signal, and n is the discrete time index, that is, the time point.

[0046] A further improvement of the present invention is that the calculation operation performed by the calculation module is:

[0047]

[0048] Among them, d i is the traveling wave fault distance of the ith traveling wave in the distribution network, v i is the propagation speed of the ith traveling wave, is a local extreme point.

[0049] A further improvement of the present invention is that the operations performed by the positioning module include:

[0050] Adding the traveling wave fault distance to the noise, the expression is:

[0051]

[0052] in, is the traveling wave fault distance of the ith traveling wave after adding noise, d i is the traveling wave fault distance of the ith traveling wave in the distribution network, ∈ i is the added noise;

[0053] The traveling wave fault distance of the i-th traveling wave after adding noise is input into the DDPM model, and the diffusion process is carried out to gradually add noise. The expression is:

[0054]

[0055] in, is the noise data of the ith traveling wave at the tth step, is the noise scaling factor in the diffusion process, ∈ is Gaussian noise;

[0056] The noise data of the i-th traveling wave at the t-th step is calculated by the reverse process of the DDPM model. Stepwise denoising, the expression is:

[0057]

[0058] in, The distance gradually approaches the real distance after noise removal;

[0059] Estimate the uncertainty of each traveling wave and calculate the weight of the traveling wave. The expression is:

[0060]

[0061] Among them, w i is the weight of the ith traveling wave, E t () means taking the expectation of the values ​​calculated at different time steps t, that is, the average value;

[0062] According to the gradual approach to the true distance after noise removal And the corresponding traveling wave weight, calculate the final position of the distribution network fault location, the expression is:

[0063]

[0064] Among them, z0 is the final position of the distribution network fault location.

[0065] A distribution network fault location method based on high anti-interference capability traveling wave ranging of the present invention includes the following operations:

[0066] The sampling frequency and traveling wave signals are collected by using high-speed sampling devices arranged at key nodes of the distribution network, and the traveling wave signals are preprocessed;

[0067] Perform wavelet transform on the preprocessed traveling wave signal to obtain wavelet coefficients, and build a local extreme point model of wavelet coefficients based on the knowledge distillation algorithm. Use the local extreme point model of wavelet coefficients to detect the arrival time of traveling waves.

[0068] According to the detection results of the traveling wave arrival time, the traveling wave fault distance in the distribution network is calculated;

[0069] The fault location of the distribution network is determined using the traveling wave fault distances of multiple traveling waves.

[0070] A further improvement of the present invention is to collect the sampling frequency and the traveling wave signal by using a high-speed sampling device arranged at the key nodes of the distribution network, and to pre-process the traveling wave signal, specifically including:

[0071] The traveling wave signal is collected, and the expression of the collected traveling wave signal is:

[0072]

[0073] Among them, x() is the collected traveling wave signal, n is the discrete time index, that is, the time point, T s is the sampling interval, f s is the sampling frequency, used to capture high-speed changing traveling wave signals;

[0074] The expression for preprocessing the traveling wave signal is:

[0075]

[0076] Among them, x filtered (n) is the preprocessed traveling wave signal, and h(k) is the impulse response of the filter.

[0077] A further improvement of the present invention is to perform wavelet transform on the preprocessed traveling wave signal to obtain wavelet coefficients, and to construct a local extreme point model of the wavelet coefficients based on a knowledge distillation algorithm, and to use the local extreme point model of the wavelet coefficients to detect the arrival time of the traveling wave, specifically including:

[0078] Perform wavelet transform on the preprocessed traveling wave signal to obtain the wavelet coefficient, which is expressed as:

[0079]

[0080] Among them, W(a,b) is the wavelet coefficient, is the mother wavelet function, a and b are the scale factor and time smoothing factor respectively, x filtered (n) is the preprocessed traveling wave signal, n is the discrete time index, i.e., the time point;

[0081] Based on the knowledge distillation algorithm, a wavelet coefficient local extreme point model is constructed, where the wavelet coefficient local extreme point model includes a teacher model, a student model and a loss function; the teacher model is used to process the wavelet coefficient data and detect local extreme points, and the output is the wavelet transform coefficient W teacher (a, b) The predicted value under different scale factors a and time smoothing factors b; the expression of the loss function is:

[0082] L total =aL distill +βL standard (4)

[0083]

[0084]

[0085] Among them, L distill is the knowledge distillation loss, L standard is the standard loss, L total is the total loss, W teacher (a, b) are the wavelet transform coefficients output by the teacher model, W student (a, b) are the wavelet transform coefficients output by the student model, N is the number of samples in the training data set, is the predicted value of the student model, indicating whether there is a local extreme value at time point n, where 1 means yes and 0 means no. y(n) is the actual position of the local extreme point. α and β are weight coefficients, which are used to balance the influence of knowledge distillation loss and standard loss respectively.

[0086] Through back propagation and gradient descent method, the student model is trained so that the total loss L total Minimize and use the Adam optimization algorithm to update the parameters of the student model. The expression is:

[0087]

[0088] Among them, θ student are the parameters of the student model, is the gradient descent function, η is the gradient descent coefficient;

[0089] According to the parameters of the updated student model, an extreme point prediction model is constructed to predict local extreme points. That is, predict the arrival time of the traveling wave, the expression is:

[0090]

[0091] in, is a local extreme point, is the extreme point prediction model, x() is the collected traveling wave signal, and n is the discrete time index, that is, the time point.

[0092] A further improvement of the present invention is that the expression for calculating the traveling wave fault distance in the distribution network is:

[0093]

[0094] Among them, d i is the traveling wave fault distance of the ith traveling wave in the distribution network, v i is the propagation speed of the ith traveling wave, is a local extreme point.

[0095] A further improvement of the present invention is to determine the fault location of the distribution network by using the traveling wave fault distances of multiple traveling waves, specifically including:

[0096] Adding the traveling wave fault distance to the noise, the expression is:

[0097]

[0098] in, is the traveling wave fault distance of the ith traveling wave after adding noise, d i is the traveling wave fault distance of the ith traveling wave in the distribution network, ∈ i is the added noise;

[0099] The traveling wave fault distance of the i-th traveling wave after adding noise is input into the DDPM model, and the diffusion process is carried out to gradually add noise. The expression is:

[0100]

[0101] in, is the noise data of the ith traveling wave at the tth step, is the noise scaling factor in the diffusion process, ∈ is Gaussian noise;

[0102] The noise data of the i-th traveling wave at the t-th step is calculated by the reverse process of the DDPM model. Stepwise denoising, the expression is:

[0103]

[0104] in, The distance gradually approaches the real distance after noise removal;

[0105] Estimate the uncertainty of each traveling wave and calculate the weight of the traveling wave. The expression is:

[0106]

[0107] Among them, w i is the weight of the ith traveling wave, E t () means taking the expectation of the values ​​calculated at different time steps t, that is, the average value;

[0108] According to the gradual approach to the true distance after noise removal And the corresponding traveling wave weight, calculate the final position of the distribution network fault location, the expression is:

[0109]

[0110] Among them, z0 is the final position of the distribution network fault location.

[0111] The beneficial effects of the present invention are as follows: the present invention effectively improves the accuracy and speed of the detection of the arrival time of the traveling wave, reduces the consumption of computing resources while maintaining high accuracy, and enhances the real-time performance of the detection. In addition, the present invention still maintains a high fault location accuracy in a strong noise environment and significantly improves the anti-interference ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0112] Figure 1 Schematic diagram of the system structure principle of an embodiment of the present invention;

[0113] Figure 2 It is a schematic diagram of a method flow of an embodiment of the present invention. DETAILED DESCRIPTION

[0114] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0115] like Figure 1 The present embodiment provides a distribution network fault location system based on traveling wave ranging with high anti-interference capability, including:

[0116] The acquisition module is used to collect the sampling frequency and the traveling wave signal by using the high-speed sampling device arranged at the key nodes of the distribution network, and to pre-process the traveling wave signal;

[0117] The detection module is used to perform wavelet transform on the preprocessed traveling wave signal to obtain wavelet coefficients, and to construct a local extreme point model of the wavelet coefficients based on the knowledge distillation algorithm, and to detect the arrival time of the traveling wave using the local extreme point model of the wavelet coefficients;

[0118] A calculation module for calculating the distance of traveling wave faults in the distribution network;

[0119] The positioning module is used to determine the fault location of the distribution network by using the traveling wave fault distances of multiple traveling waves.

[0120] Based on the above-mentioned distribution network fault location system based on high anti-interference capability traveling wave ranging, the following distribution network fault location method based on high anti-interference capability traveling wave ranging is implemented, such as Figure 2 As shown, the method includes:

[0121] Step 1, using a high-speed sampling device arranged at a key node of the distribution network to collect sampling frequency, traveling wave signal, and preprocess the traveling wave signal. The traveling wave signal is the change of voltage or current at different times. The sampling frequency is the sampling frequency of the high-speed sampling device, which is used to capture the traveling wave signal with high-speed changes. This embodiment proposes a traveling wave signal preprocessing method based on filtering, and uses a filtering method to preprocess the signal to eliminate high-frequency noise.

[0122] Step 2, perform wavelet transform on the preprocessed traveling wave signal to obtain wavelet coefficients, and construct a local extreme point model of wavelet coefficients based on the knowledge distillation algorithm, and use the local extreme point model of wavelet coefficients to detect the arrival time of traveling waves. The local extreme point model of wavelet coefficients based on the knowledge distillation algorithm can maintain high accuracy and reduce the consumption of computing resources when detecting local extreme points of wavelet coefficients by transferring the knowledge of a complex, high-performance large model (teacher model) to a smaller, more computationally efficient model (student model). This embodiment trains a high-performance teacher model to detect local extreme points, and transfers the knowledge of the teacher model to a smaller student model, so as to achieve efficient extreme point detection in a resource-constrained environment, thereby improving the accuracy and effectiveness of traveling wave arrival time detection.

[0123] Step 3: Calculate the traveling wave fault distance in the distribution network based on the detection result of the traveling wave arrival time. This embodiment mainly calculates the traveling wave fault distance in the distribution network based on the traveling wave propagation speed and the predicted traveling wave arrival time, providing a data basis for the final fault location.

[0124] Step 4, a distribution network fault high anti-interference and accurate positioning method based on DDPM is proposed, and the fault location of the distribution network is determined by using the traveling wave fault distances of multiple traveling waves. In the distribution network fault location positioning, this embodiment predicts the weight of each traveling wave based on the DDPM model to improve the accuracy of fault location.

[0125] In step 1 of this embodiment, the expression of the collected traveling wave signal for locating the distribution network fault is:

[0126]

[0127] Among them, x() is the collected traveling wave signal, n is the discrete time index, that is, the time point, T s is the sampling interval, f s is the sampling frequency, used to capture high-speed changing traveling wave signals;

[0128] The expression for preprocessing the traveling wave signal is:

[0129]

[0130] Among them, x filtered (n) is the preprocessed traveling wave signal, and h(k) is the impulse response of the filter.

[0131] In step 2, the traveling wave signal preprocessed in step 1 is subjected to wavelet transform using a wavelet transform model to obtain wavelet coefficients. The expression of the wavelet transform model is:

[0132]

[0133] Among them, W(a,b) is the wavelet coefficient, is the mother wavelet function, a and b are the scale factor and time smoothing factor respectively, x filtered (n) is the preprocessed traveling wave signal, and n is the discrete time index, i.e., the time point.

[0134] The wavelet coefficient local extreme point model is constructed based on the knowledge distillation algorithm, where the wavelet coefficient local extreme point model includes a teacher model, a student model and a loss function. The teacher model is based on the Transformer architecture and is specifically used to process high-dimensional wavelet coefficient data and accurately detect local extreme points. The student model is a relatively lightweight model used to detect local extreme points while maintaining high accuracy.

[0135] The teacher model is trained using a high-precision data set so that the teacher model can accurately detect the local extreme point t0 of the wavelet coefficient W(a,b). The output of the teacher model is the wavelet transform coefficient W teacher (a, b) Prediction values ​​under different scaling factors a and time smoothing factors b.

[0136] During the student model training process, the input signal is also transformed by wavelet transform and the corresponding wavelet coefficients W are generated. student(a, b). The goal of training is to make the output of the student model as close as possible to the output of the teacher model, while also considering the actual label y(n) (the true location of the local extreme point).

[0137] Construct a knowledge distillation loss model, as shown in expression (4), to let the student model imitate the output of the teacher model. Commonly used loss functions are based on mean square error (MSE) or cross entropy. Construct a standard loss model, as shown in expression (5), to measure the prediction accuracy of the student model for the true label. The final total loss function is shown in expression (6).

[0138] L total =aL distill +βL standard (4)

[0139]

[0140]

[0141] Among them, L distill is the knowledge distillation loss, L standard is the standard loss, L total is the total loss, W teacher (a, b) are the wavelet transform coefficients output by the teacher model, W student (a, b) are the wavelet transform coefficients output by the student model, N is the number of samples in the training data set, is the predicted value of the student model, indicating whether there is a local extreme value at time point n, where 1 means yes and 0 means no. y(n) is the true position of the local extreme point. α and β are weight coefficients, which are used to balance the influence of knowledge distillation loss and standard loss respectively.

[0142] Through back propagation and gradient descent method, the student model is trained so that the total loss L total Minimize. The goal of the student model is to optimize the parameters θ of the student model student To accurately predict the local extreme point t0. This embodiment uses the Adam optimization algorithm to update the parameters θ of the student model student , the expression is:

[0143]

[0144] Among them, θ student The parameters of the student model, including weights and biases, are mainly obtained through optimization operations during the training process and determine the final performance of the student model. is the gradient descent function, and η is the gradient descent coefficient.

[0145] According to the updated parameters θ of the student model student, build an extreme point prediction model to predict local extreme points This embodiment uses local extreme points represents the arrival time of the traveling wave, and the expression is:

[0146]

[0147] in, is a local extreme point, It is the extreme point prediction model.

[0148] In step 3, the expression for calculating the traveling wave fault distance in the distribution network is:

[0149]

[0150] Among them, d i is the traveling wave fault distance of the ith traveling wave in the distribution network, v i is the propagation speed of the ith traveling wave, is a local extreme point.

[0151] In step 4, the traveling wave fault distance of the traveling wave may be affected by noise in practice, so it is necessary to pre-process the traveling wave fault distance, that is, add the noise to the traveling wave fault distance, and the expression is:

[0152]

[0153] in, is the traveling wave fault distance of the ith traveling wave after adding noise, d i is the traveling wave fault distance of the ith traveling wave in the distribution network, ∈ i is the added noise.

[0154] The traveling wave fault distance of the i-th traveling wave after adding noise is input into the DDPM model, and the diffusion process is carried out to gradually add noise. The expression is:

[0155]

[0156] in, is the noise data of the ith traveling wave at the tth step, is the noise scaling factor in the diffusion process, used to scale the original signal, and ∈ is Gaussian noise.

[0157] The noise data of the i-th traveling wave at the t-th step is calculated by the reverse process of the DDPM model. Stepwise denoising, the expression is:

[0158]

[0159] in, It is the distance that gradually approaches the real distance after removing noise.

[0160] On the basis of the denoising operation, the uncertainty of each traveling wave is estimated and the weight of the traveling wave is calculated. The expression is:

[0161]

[0162] Among them, w i is the weight of the ith traveling wave, E t () means taking the expectation, that is, the average value, of the values ​​calculated at different time steps t.

[0163] According to the gradual approach to the true distance after noise removal And the corresponding traveling wave weight, calculate the final position of the distribution network fault location, the expression is:

[0164]

[0165] Among them, z0 is the final position of the distribution network fault location.

[0166] After the above steps, the fault in the distribution network can be accurately located, and the anti-interference ability, positioning accuracy and robustness of the fault location can be improved.

[0167] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in common dictionaries should be understood to have the same meaning as in the context of the prior art, and will not be interpreted with an idealized or overly formal meaning unless defined similarly here.

[0168] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A distribution network fault location system based on traveling wave ranging with high anti-interference capability, characterized by: The system comprises: The acquisition module is used to collect the sampling frequency and the traveling wave signal by using the high-speed sampling device arranged at the node of the distribution network, and to pre-process the traveling wave signal; The detection module is used to perform wavelet transform on the preprocessed traveling wave signal to obtain wavelet coefficients, and to construct a local extreme point model of the wavelet coefficients based on the knowledge distillation algorithm, and to detect the arrival time of the traveling wave using the local extreme point model of the wavelet coefficients; A calculation module for calculating the distance of traveling wave faults in the distribution network; The positioning module is used to determine the fault location of the distribution network by using the traveling wave fault distances of multiple traveling waves.

2. A distribution network fault location system and method based on traveling wave ranging with high anti-interference capability according to claim 1, characterized in that: The traveling wave signal collected by the acquisition module is: Among them, x() is the collected traveling wave signal, n is the discrete time index, that is, the time point, T s is the sampling interval, f s is the sampling frequency, used to capture high-speed changing traveling wave signals; The expression for preprocessing the traveling wave signal is: Among them, x filtered (n) is the preprocessed traveling wave signal, h(k) is the impulse response of the filter, and k is the discrete time index.

3. A distribution network fault location system and method based on traveling wave ranging with high anti-interference capability according to claim 1, characterized in that: The operations performed by the detection module include: Perform wavelet transform on the preprocessed traveling wave signal to obtain the wavelet coefficient, which is expressed as: Among them, W(a,b) is the wavelet coefficient, is the mother wavelet function, a and b are the scale factor and time smoothing factor respectively, x filtered (n) is the preprocessed traveling wave signal, n is the discrete time index, i.e., the time point; Based on the knowledge distillation algorithm, a wavelet coefficient local extreme point model is constructed, where the wavelet coefficient local extreme point model includes a teacher model, a student model and a loss function; the teacher model is used to process the wavelet coefficient data and detect local extreme points, and the output is the wavelet transform coefficient W teacher (a,b) The predicted value under different scale factors a and time smoothing factors b; the expression of the loss function is: L total =aL distill +βL standard (4) Among them, L distill is the knowledge distillation loss, L standard is the standard loss, L total is the total loss, W teacher (a, b) are the wavelet transform coefficients output by the teacher model, W student (a, b) are the wavelet transform coefficients output by the student model, N is the number of samples in the training data set, is the predicted value of the student model, indicating whether there is a local extreme value at time point n, where 1 means yes and 0 means no. y(n) is the actual position of the local extreme value point. α and β are weight coefficients, which are used to balance the influence of knowledge distillation loss and standard loss respectively. Through back propagation and gradient descent method, the student model is trained so that the total loss L total Minimize and use the Adam optimization algorithm to update the parameters of the student model. The expression is: Among them, θ student are the parameters of the student model, is the gradient descent function, η is the gradient descent coefficient; According to the parameters of the updated student model, an extreme point prediction model is constructed to predict local extreme points. That is, predict the arrival time of the traveling wave, the expression is: in, is a local extreme point, is the extreme point prediction model, x() is the collected traveling wave signal, and n is the discrete time index, that is, the time point.

4. A distribution network fault location system and method based on traveling wave ranging with high anti-interference capability according to claim 1, characterized in that: The calculation operation performed by the calculation module is: Among them, d i is the traveling wave fault distance of the ith traveling wave in the distribution network, v i is the propagation speed of the ith traveling wave, is a local extreme point.

5. A distribution network fault location system and method based on traveling wave ranging with high anti-interference capability according to claim 1, characterized in that: The operations performed by the positioning module include: Adding the traveling wave fault distance to the noise, the expression is: in, is the traveling wave fault distance of the ith traveling wave after adding noise, d i is the traveling wave fault distance of the ith traveling wave in the distribution network, ∈ i is the added noise; The traveling wave fault distance of the i-th traveling wave after adding noise is input into the DDPM model, and the diffusion process is carried out to gradually add noise. The expression is: in, is the noise data of the ith traveling wave at the tth step, is the noise scaling factor in the diffusion process, ∈ is Gaussian noise; The noise data of the i-th traveling wave at the t-th step is calculated by the reverse process of the DDPM model. Stepwise denoising, the expression is: in, The distance gradually approaches the real distance after noise removal; Estimate the uncertainty of each traveling wave and calculate the weight of the traveling wave. The expression is: Among them, w i is the weight of the ith traveling wave, E t () means taking the expectation of the values ​​calculated at different time steps t, that is, the average value; According to the gradual approach to the true distance after noise removal And the corresponding traveling wave weight, calculate the final position of the distribution network fault location, the expression is: Among them, z0 is the final position of the distribution network fault location.

6. A distribution network fault location method based on high anti-interference capability traveling wave ranging, characterized in that: The following operations are included: The sampling frequency and traveling wave signals are collected by using high-speed sampling devices arranged at key nodes of the distribution network, and the traveling wave signals are preprocessed; Perform wavelet transform on the preprocessed traveling wave signal to obtain wavelet coefficients, and build a local extreme point model of wavelet coefficients based on the knowledge distillation algorithm. Use the local extreme point model of wavelet coefficients to detect the arrival time of traveling waves. According to the detection results of the traveling wave arrival time, the traveling wave fault distance in the distribution network is calculated; The fault location of the distribution network is determined using the traveling wave fault distances of multiple traveling waves.

7. A distribution network fault location method based on traveling wave ranging with high anti-interference capability according to claim 6, characterized in that: The method uses a high-speed sampling device arranged at a key node of the distribution network to collect sampling frequency and traveling wave signals, and pre-processes the traveling wave signals, specifically including: The traveling wave signal is collected, and the expression of the collected traveling wave signal is: Among them, x() is the collected traveling wave signal, n is the discrete time index, that is, the time point, T s is the sampling interval, f s is the sampling frequency, used to capture high-speed changing traveling wave signals; The expression for preprocessing the traveling wave signal is: Among them, x filtered (n) is the preprocessed traveling wave signal, and h(k) is the impulse response of the filter.

8. A distribution network fault location method based on traveling wave ranging with high anti-interference capability according to claim 6, characterized in that: The preprocessed traveling wave signal is subjected to wavelet transform to obtain wavelet coefficients, and a local extreme point model of the wavelet coefficients is constructed based on a knowledge distillation algorithm, and the arrival time of the traveling wave is detected by using the local extreme point model of the wavelet coefficients, which specifically includes: Perform wavelet transform on the preprocessed traveling wave signal to obtain the wavelet coefficient, which is expressed as: Among them, W(a,b) is the wavelet coefficient, is the mother wavelet function, a and b are the scale factor and time smoothing factor respectively, x filtered (n) is the preprocessed traveling wave signal, n is the discrete time index, i.e., the time point; Based on the knowledge distillation algorithm, a wavelet coefficient local extreme point model is constructed, where the wavelet coefficient local extreme point model includes a teacher model, a student model and a loss function; the teacher model is used to process the wavelet coefficient data and detect local extreme points, and the output is the wavelet transform coefficient W teacher (a,b) The predicted value under different scale factors a and time smoothing factors b; the expression of the loss function is: L total =aL distill +βL standard (4) Among them, L distill is the knowledge distillation loss, L standard is the standard loss, L total is the total loss, W teacher (a, b) are the wavelet transform coefficients output by the teacher model, W student (a, b) are the wavelet transform coefficients output by the student model, N is the number of samples in the training data set, is the predicted value of the student model, indicating whether there is a local extreme value at time point n, where 1 means yes and 0 means no. y(n) is the actual position of the local extreme value point. α and β are weight coefficients, which are used to balance the influence of knowledge distillation loss and standard loss respectively. Through back propagation and gradient descent method, the student model is trained so that the total loss L total Minimize and use the Adam optimization algorithm to update the parameters of the student model. The expression is: Among them, θ student are the parameters of the student model, is the gradient descent function, η is the gradient descent coefficient; According to the parameters of the updated student model, an extreme point prediction model is constructed to predict local extreme points. That is, predict the arrival time of the traveling wave, the expression is: in, is a local extreme point, is the extreme point prediction model, x() is the collected traveling wave signal, and n is the discrete time index, that is, the time point.

9. A distribution network fault location method based on traveling wave ranging with high anti-interference capability according to claim 6, characterized in that: The expression for calculating the traveling wave fault distance in the distribution network is: Among them, d i is the traveling wave fault distance of the ith traveling wave in the distribution network, v i is the propagation speed of the ith traveling wave, is a local extreme point.

10. A distribution network fault location method based on traveling wave ranging with high anti-interference capability according to claim 6, characterized in that: The method of determining the fault location of the distribution network by using the traveling wave fault distances of multiple traveling waves specifically includes: adding noise to the traveling wave fault distance, and the expression is: in, is the traveling wave fault distance of the ith traveling wave after adding noise, d i is the traveling wave fault distance of the ith traveling wave in the distribution network, ∈ i is the added noise; The traveling wave fault distance of the i-th traveling wave after adding noise is input into the DDPM model, and the diffusion process is carried out to gradually add noise. The expression is: in, is the noise data of the ith traveling wave at the tth step, is the noise scaling factor in the diffusion process, ∈ is Gaussian noise; The noise data of the i-th traveling wave at the t-th step is calculated by the reverse process of the DDPM model. Stepwise denoising, the expression is: in, The distance gradually approaches the real distance after noise removal; Estimate the uncertainty of each traveling wave and calculate the weight of the traveling wave. The expression is: Among them, w i is the weight of the ith traveling wave, E t () means taking the expectation of the values ​​calculated at different time steps t, that is, the average value; According to the gradual approach to the true distance after noise removal And the corresponding traveling wave weight, calculate the final position of the distribution network fault location, the expression is: Among them, z0 is the final position of the distribution network fault location.

Images