Signal denoising method, device, equipment, storage medium and product

Abstract

This invention discloses a signal denoising method, apparatus, device, storage medium, and product. The method includes: acquiring the original signal in a distribution network line and determining noise parameters based on the original signal; acquiring the response signal in the distribution network line; wherein the response signal is the signal in the distribution network line after a high-frequency short-time excitation signal is injected; and using a preset wavelet operator denoising model, determining the target denoised signal corresponding to the response signal based on the noise parameters and the response signal; wherein the preset wavelet operator denoising model includes multiple wavelet integral layers; each wavelet integral layer is a neural network layer containing wavelet operators. This invention can efficiently denoise the response signal corresponding to a high-frequency short-time excitation signal, thereby improving the accuracy and reliability of fault detection in distribution network lines.

Description

Technical Field

[0001] This invention relates to the field of signal processing technology, and in particular to signal denoising methods, apparatus, devices, storage media, and products. Background Technology

[0002] Distribution network cables are a critical component of urban power distribution systems, and their operational status directly impacts power supply safety and reliability. With the expansion of distribution networks, issues such as cable aging and insulation defects are becoming increasingly prominent. High-frequency short-time excitation detection technology, by actively injecting signals and analyzing response characteristics, can achieve accurate fault diagnosis and early warning for cables. However, in complex electromagnetic environments, the acquired response signals are susceptible to white noise, power frequency harmonics, and pulse interference, resulting in a low signal-to-noise ratio and severely reducing the accuracy of defect location and condition assessment.

[0003] Existing denoising methods mostly use fixed filtering parameters or empirical thresholds, which have poor adaptability; denoising models based on neural networks generally focus on learning single signal features, which are not robust enough to complex interference in distribution network cables and cannot meet the high-precision and strong anti-interference signal processing requirements in engineering. Summary of the Invention

[0004] This invention provides a signal denoising method, apparatus, device, storage medium, and product to solve the problem in the prior art of accurately denoising high-frequency signals under strong interference environments.

[0005] According to one aspect of the present invention, a signal denoising method is provided, comprising: Acquire the original signals in the distribution network lines, and determine the noise parameters based on the original signals; Acquire the response signal in the power distribution network line; wherein, the response signal is the signal in the power distribution network line after a high-frequency short-time excitation signal is injected into the power distribution network line; Using a preset wavelet operator denoising model, the target denoised signal corresponding to the response signal is determined based on the noise parameters and the response signal; wherein, the preset wavelet operator denoising model includes multiple wavelet integral layers; each wavelet integral layer is a neural network layer containing wavelet operators.

[0006] According to another aspect of the present invention, a signal denoising apparatus is provided, comprising: The noise parameter determination module is used to acquire the original signal in the distribution network line and determine the noise parameter based on the original signal; A response signal acquisition module is used to acquire the response signal in the power distribution network line; wherein, the response signal is the signal in the power distribution network line after a high-frequency short-time excitation signal is injected into the power distribution network line; The denoising module is used to determine the target denoised signal corresponding to the response signal based on the noise parameters and the response signal using a preset wavelet operator denoising model; wherein the preset wavelet operator denoising model includes multiple wavelet integral layers; each wavelet integral layer is a neural network layer containing wavelet operators.

[0007] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the signal denoising method according to any embodiment of the present invention.

[0008] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the signal denoising method according to any embodiment of the present invention.

[0009] According to another aspect of the present invention, a computer program product is provided, the computer program product comprising a computer program that, when executed by a processor, implements the signal denoising method according to any embodiment of the present invention.

[0010] The technical solution of this invention involves acquiring the original signal in the distribution network line and determining noise parameters based on the original signal; acquiring the response signal in the distribution network line; wherein the response signal is the signal in the distribution network line after a high-frequency short-time excitation signal is injected into the distribution network line; and using a preset wavelet operator denoising model, determining the target denoised signal corresponding to the response signal based on the noise parameters and the response signal; wherein the preset wavelet operator denoising model includes multiple wavelet integral layers; each wavelet integral layer is a neural network layer containing wavelet operators. The above technical solution, by determining noise parameters based on the original signal, ensures that the preset wavelet operator denoising model can learn the complex noise characteristics in the distribution network line, thereby accurately denoising the response signal. This avoids excessive suppression of high-frequency signals using fixed denoising parameters and prevents the filtering out of high-frequency transient features in the response signal. Furthermore, the wavelet integral layer integrates wavelet transform and neural networks, enabling the preset wavelet operator denoising model to combine the advantages of wavelet time-frequency analysis with the efficient feature extraction capabilities of deep learning. This ensures efficient acquisition of noise distribution characteristics in the distribution network line and effective signal features in the response signal, further improving the denoising capability of the response signal and making the results of fault detection of the distribution network line based on high-frequency short-time excitation signals more reliable.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of a signal denoising method provided in Embodiment 1 of the present invention; Figure 2 This is a flowchart of a signal denoising method provided in Embodiment 2 of the present invention; Figure 3 This is a network structure diagram of the preset wavelet operator denoising model applicable to the embodiments of the present invention; Figure 4 This is a schematic diagram of a signal denoising device according to Embodiment 3 of the present invention; Figure 5 This is a schematic diagram of the structure of an electronic device that implements the signal denoising method of Embodiment 4 of the present invention. Detailed Implementation

[0014] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0015] It should be noted that the terms "first," "second," "original," and "target," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0016] Example 1 Figure 1 This is a flowchart illustrating a signal denoising method according to Embodiment 1 of the present invention. This embodiment is applicable to signal processing situations. The method can be executed by a signal denoising device, which can be implemented in hardware and / or software and can be configured in an electronic device. Figure 1 As shown, the method includes: S101. Obtain the original signals in the distribution network lines and determine the noise parameters based on the original signals.

[0017] In this embodiment, the distribution network line can be the cable line in the power distribution system; the original signal can be understood as the signal in the distribution network line during the actual operation of the power distribution system, that is, the signal in the distribution network line before the high-frequency short-time excitation signal is injected into the distribution network line; since the distribution network line is usually under complex electromagnetic conditions during the actual operation of the power distribution system, the original signal generally contains a large amount of white noise, power frequency harmonics, and random pulses; therefore, the noise parameters can be determined based on the original signal; among them, the high-frequency short-time excitation signal can be an excitation signal with high frequency components and short duration, such as a narrow pulse signal.

[0018] For example, raw signals of a preset duration can be collected from the distribution network line, and statistical analysis can be performed on the raw signals to obtain statistical parameters of the raw signals, such as mean and variance, and noise parameters can be determined based on the statistical parameters of the raw signals.

[0019] S102. Obtain the response signal in the distribution network line; wherein, the response signal is the signal in the distribution network line after the high-frequency short-time excitation signal is injected into the distribution network line.

[0020] In this embodiment, after the high-frequency short-time excitation signal is injected into the distribution network line, the high-frequency short-time excitation signal will be reflected and refracted in the distribution network line. The response signal can be understood as a superposition of the reflected signal, refracted signal, and original signal of the high-frequency short-time excitation signal in the distribution network line. The response signal generally includes signals with high-frequency characteristics.

[0021] For example, a high-frequency short-time excitation signal can be injected into the distribution network line, and the signals in the distribution network can be collected, with the collection results used as a response signal.

[0022] S103. Using a preset wavelet operator denoising model, determine the target denoised signal corresponding to the response signal based on the noise parameters and the response signal; wherein, the preset wavelet operator denoising model includes multiple wavelet integral layers; each wavelet integral layer is a neural network layer containing wavelet operators.

[0023] In this embodiment, the preset wavelet operator denoising model can be a model trained using a first noise parameter, a first response signal, and a first denoised signal within a historical period, and its training result already meets the training cutoff condition. The first noise parameter can be determined based on the original signal of a first historical duration within the historical period. The first response signal can be a response signal of a second historical duration collected after injecting a high-frequency short-time excitation signal into the distribution network line within the historical period. The first denoised signal can be a signal filtered based on the first noise parameter using a preset filtering algorithm (e.g., a Gaussian filtering algorithm), which can be used as the true value of the first response signal. The training cutoff condition can be preset, for example, the number of iterations reaching a preset iteration threshold, or the loss value being less than a preset loss threshold.

[0024] The preset wavelet operator denoising model can include multiple wavelet integral layers. Each wavelet integral layer can be a neural network layer containing wavelet operators. During the inference process of the preset wavelet operator denoising model, wavelet operators can be used to perform wavelet processing on noise parameters and response signals, thereby capturing the signal characteristics of noise signals and response signals, so that the denoised signal can retain its high-frequency characteristics. The wavelet operators can be preset, for example, determined according to preset wavelet basis functions.

[0025] The target denoised signal can be the signal obtained after denoising the response signal using a preset wavelet operator denoising model. For example, noise parameters and the response signal are input into the preset wavelet operator denoising model, and the target denoised signal is output. Further processing can be performed on the target denoised signal to determine whether a fault has occurred in the distribution network line. For instance, the target denoised signal can be matched and compared with the input high-frequency short-time excitation signal to determine whether parameters such as amplitude, time delay, and waveform similarity meet relevant preset conditions, thereby determining whether a fault has occurred in the distribution network line.

[0026] This invention provides a signal denoising method, which involves acquiring the original signal in a power distribution network line and determining noise parameters based on the original signal; acquiring the response signal in the power distribution network line; wherein the response signal is the signal in the power distribution network line after a high-frequency short-time excitation signal is injected into the power distribution network line; and using a preset wavelet operator denoising model, determining the target denoised signal corresponding to the response signal based on the noise parameters and the response signal; wherein the preset wavelet operator denoising model includes multiple wavelet integral layers; each wavelet integral layer is a neural network layer containing wavelet operators. The above technical solution, by determining noise parameters based on the original signal, ensures that the preset wavelet operator denoising model can learn the complex noise characteristics in the distribution network line, thereby accurately denoising the response signal. This avoids excessive suppression of high-frequency signals using fixed denoising parameters and prevents the filtering out of high-frequency transient features in the response signal. Furthermore, the wavelet integral layer integrates wavelet transform and neural networks, enabling the preset wavelet operator denoising model to combine the advantages of wavelet time-frequency analysis with the efficient feature extraction capabilities of deep learning. This ensures efficient acquisition of noise distribution characteristics in the distribution network line and effective signal features in the response signal, further improving the denoising capability of the response signal and making the results of fault detection of the distribution network line based on high-frequency short-time excitation signals more reliable.

[0027] Example 2 Figure 2 This is a flowchart of a signal denoising method provided in Embodiment 2 of the present invention. This embodiment further refines the above embodiment. The preset wavelet operator denoising model in this embodiment may include a feature extraction layer, a wavelet integration layer, and a feature projection layer; wherein, the feature extraction layer and the feature projection layer are both convolutional neural networks. Figure 2 As shown, the method includes: S201. Obtain the original signals in the distribution network lines and determine the noise parameters based on the original signals.

[0028] For example, a preset time window is obtained before injecting a high-frequency short-time excitation signal into the distribution network line. The original signal within the specified time window is used to calculate its mean and variance, which are then used as noise parameters. The timestamp represents the injection of a high-frequency short-time excitation signal into the distribution network line.

[0029] S202. Obtain the response signal in the distribution network line; wherein, the response signal is the signal in the distribution network line after the high-frequency short-time excitation signal is injected into the distribution network line.

[0030] For example, in High-frequency, short-time excitation signals are continuously injected into the distribution network lines, and response signals from the distribution network lines are collected according to a preset acquisition frequency. The response signals are associated with time information, and an input vector can be constructed based on the response signals, time information, and noise parameters to input a preset wavelet operator denoising model. The input vector can be a one-dimensional sequence.

[0031] S203. Use the feature extraction layer to extract features from the noise parameters and response signal to obtain the first processed signal.

[0032] The feature extraction layer can be a convolutional neural network containing a first preset number of layers, and the kernel size can be... The input vector can be input into the feature extraction layer for feature extraction, and the output is the first processed signal.

[0033] S204. The first processed signal is linearly transformed and processed by wavelet integral layer to obtain the second processed signal.

[0034] In this embodiment, the wavelet integral layer can adopt a parallel structure, performing linear transformation and wavelet processing on the input signal respectively to obtain parallel processing results, and fusing the parallel processing results to obtain the output signal corresponding to the wavelet integral layer.

[0035] For example, the wavelet integral layer includes a linear transformation branch and a wavelet integral operator branch; the linear transformation branch is a neural network branch containing a preset parameter matrix; the wavelet integral operator branch is a neural network branch containing a forward wavelet operator and an inverse wavelet operator; the forward wavelet operator and the inverse wavelet operator are determined according to a preset wavelet basis function; wherein, the step of using the wavelet integral layer to perform linear transformation and wavelet processing on the first processed signal to obtain the second processed signal includes the following steps A1 to A3: A1. The first processed signal is linearly transformed based on the preset parameter matrix in the linear transformation branch to obtain a linearly transformed signal; wherein, the preset parameter matrix is ​​a learnable parameter matrix, and the parameters of the preset parameter matrix are determined according to the training results of the preset wavelet operator denoising model.

[0036] In this embodiment, the first processed signal is input into the linear transformation branch, and the first processed signal can be linearly mapped and reconstructed in the time domain using the preset parameter matrix in the linear transformation branch to obtain the linearly transformed signal; the preset parameter matrix is ​​a learnable parameter matrix, and its parameters are determined according to the training results of the preset wavelet operator denoising model.

[0037] A2. Based on the forward wavelet operator and the inverse wavelet operator in the wavelet integral operator branch, perform wavelet processing on the first processed signal to obtain the wavelet processed signal.

[0038] In this embodiment, the forward and inverse wavelet operators can be preset, for example, determined according to a preset wavelet basis function. The preset wavelet basis function can be selected according to the signal characteristics in the actual application scenario. For example, if the response signal in the distribution network line is non-stationary and contains high-frequency transient characteristics, then the Daubechies 6 wavelet (db6) can be selected as the preset wavelet basis function. The first processed signal can be decomposed based on the forward wavelet operator, and the decomposed signal can be reconstructed based on the inverse wavelet operator to obtain the wavelet-processed signal.

[0039] Optionally, the wavelet integral operator branch includes a preset weight matrix; the preset weight matrix is ​​a learnable parameter matrix, and the parameters of the preset weight matrix are determined based on the training results of the preset wavelet operator denoising model; wherein, step A2 specifically includes: performing wavelet decomposition on the first processed signal based on the forward wavelet operator to obtain a wavelet decomposed signal; and performing wavelet reconstruction on the wavelet decomposed signal based on the inverse wavelet operator and the preset weight matrix to obtain a wavelet processed signal. The advantage of this setup is that by performing wavelet decomposition on the first processed signal using the forward wavelet operator, noise signals that are difficult to distinguish in the time domain can be separated from the effective signals in the first processed signal; and by introducing a learnable preset weight matrix, adaptive weighted transformation can be performed on the wavelet coefficients obtained after wavelet decomposition during the wavelet reconstruction process, so that the wavelet reconstruction process can achieve a balance between noise suppression and effective signal enhancement.

[0040] A3. The linearly transformed signal and the wavelet-processed signal are summed, and the summation result is activated using a preset activation function to obtain the second processed signal.

[0041] The system can sum linearly transformed signals and wavelet-processed signals to achieve the fusion of parallel processing results of wavelet integral layers. The summation result can be activated using a preset activation function to obtain a second processed signal. The preset activation function can be pre-defined, for example, a nonlinear activation function can be used to nonlinearly activate the summation result to obtain the second processed signal.

[0042] By setting up a parallel structure in the wavelet integral layer, both the global features of the signal in the time domain and the local time-frequency features in the wavelet domain can be taken into account. The global features of the signal can be mapped by using the linear transform branch, and the time-frequency features of the signal can be accurately extracted by using the forward wavelet operator and the inverse wavelet operator of the wavelet integral operator branch. The output signals of the two branches are fused, which can preserve the overall features of the signal and highlight the high-frequency transient features, thereby effectively improving the denoising accuracy and signal restoration quality of the model.

[0043] S205. The second processed signal is transformed using the feature projection layer to obtain the target denoised signal.

[0044] The feature projection layer may include a convolutional neural network with a second preset number of layers, and the kernel size may be [missing information]. The second processed signal can be input into the feature projection layer for dimensional transformation, and the output can be the target denoised signal.

[0045] The preset wavelet operator denoising model in this embodiment of the invention adopts a three-level structure of feature extraction layer, wavelet integration layer, and feature projection layer. The feature extraction layer is used to perform preliminary feature mining on noise parameters and response signals to obtain a first processed signal. Then, the wavelet integration layer is used to perform linear transformation and wavelet processing on the first processed signal, taking into account both global and local time-frequency features of the signal. This allows the model to efficiently identify high-frequency characteristics in noise signals and response signals, thereby obtaining a second processed signal. Finally, the feature projection layer is used to perform dimensional transformation on the second processed signal to obtain a high-quality target denoised signal. This efficiently achieves accurate denoising of the response signal corresponding to the high-frequency short-time excitation signal in the distribution network line.

[0046] In some embodiments, the preset wavelet operator denoising model may include multiple wavelet integral layers, each of which can be a neural network layer embedding wavelet operators. These wavelet integral layers can be connected sequentially. The input to the first wavelet integral layer is the first processed signal output from the feature extraction layer; the inputs to the middle and last wavelet integral layers are the outputs of the previous wavelet integral layer; and the output of the last wavelet integral layer is the second processed signal. By setting multiple sequentially connected wavelet integral layers, the first processed signal can be processed in a multi-scale progressive manner, thereby gradually optimizing the denoising effect.

[0047] For example, the network structure of the preset wavelet operator denoising model is as follows: Figure 3 As shown, it includes n wavelet integration layers, each of which includes a linear transformation branch and a wavelet integration operator branch. The linear transformation branch is a neural network branch containing a preset parameter matrix; the wavelet integration operator branch is a neural network branch containing forward and inverse wavelet operators; the forward and inverse wavelet operators are determined according to preset wavelet basis functions; the processing procedure of each wavelet integration layer can be represented by the following expression: ; in, Representing the The output signal of the wavelet integral layer, if the first layer... If the wavelet integration layer is the last wavelet integration layer, then it corresponds to the second processed signal; This represents a preset activation function, such as the Gaussian Error Linear Unit (GELU). Represents a preset parameter matrix; Representing the The input signal of the wavelet integral layer, if the first layer... If the wavelet integral layer is the first layer, then it corresponds to the first processed signal. If the wavelet integral layer is the first layer, then it corresponds to the first processed signal. If the wavelet integral layer is an intermediate layer or the last layer, then it corresponds to the output signal of the wavelet integral layer above it. Represents the positive wavelet operator, Represents the preset weight matrix. This represents the inverse wavelet operator.

[0048] In some embodiments, the loss function of the preset wavelet operator denoising model is set based on preset objectives. These preset objectives include: making the amplitudes of the predicted signal and the label signal at the same sampling time point similar, and making the amplitude changes of the predicted signal and the label signal at the same sampling time interval similar. The predicted signal is the output signal obtained during the training phase by inputting historical response signals into the preset wavelet operator denoising model. The label signal is the ground truth signal corresponding to the historical response signal, used to characterize the noiseless signal features corresponding to the historical response signal. The advantage of this setting is that the loss function is set based on dual preset objectives. On the one hand, by making the amplitudes of the predicted signal and the label signal at the same sampling time point similar, the point-to-point denoising error can be directly reduced, ensuring the accuracy of the noiseless signal amplitude. On the other hand, by making the amplitude changes of the predicted signal and the label signal at the same sampling time interval similar, the high-frequency transient features in the response signal can be effectively preserved, avoiding the loss of signal details due to a single fitted amplitude. Thus, the trained model can accurately denoise the response signal.

[0049] For example, the preset loss function includes a first preset loss function and a second preset loss function; the first preset loss function is expressed by the following expression: ; in, This represents the first preset loss function; Represents the predicted signal; Represents the tag signal; Represents the total number of sampling time points; Representing the The amplitude of the predicted signal corresponding to each sampling time point; Representing the The amplitude of the tag signal corresponding to each sampling time point; The second preset loss function is expressed by the following expression: ; in, This represents the second preset loss function; A first-order differential operator representing amplitude and time; The first-order difference value represents the predicted signal and is used to characterize the magnitude change of the predicted signal within a preset sampling time interval. The first-order difference value representing the tag signal is used to characterize the magnitude change of the tag signal within a preset sampling time interval. Represents the total number of sampling points; Representing the The magnitude change of the predicted signal corresponding to each sampling point; Representing the The amplitude change of the tag signal corresponding to each sampling point.

[0050] In this embodiment, the preset loss function can be determined based on the weighted sum of the first preset loss function and the second preset loss function. During the training phase, the training dataset of the preset wavelet operator denoising model can include multiple signal pairs, each signal pair including a historical response signal and a corresponding label signal, wherein the label signal can be the signal after denoising the historical response signal; in each round of training, the training dataset is input into the preset wavelet operator denoising model for forward operation to obtain the predicted signal for that round, and the loss value corresponding to the predicted signal is calculated based on the preset loss function, and then backpropagation is performed based on the loss value to update the learnable parameters in the preset wavelet operator denoising model until the training cutoff condition is met.

[0051] Example 3 Figure 4 This is a schematic diagram of a signal denoising device provided in Embodiment 3 of the present invention. Figure 4 As shown, the device includes: a noise parameter determination module 401, a response signal acquisition module 402, and a noise reduction module 403.

[0052] The noise parameter determination module is used to acquire the original signal in the distribution network line and determine the noise parameter based on the original signal; A response signal acquisition module is used to acquire the response signal in the power distribution network line; wherein, the response signal is the signal in the power distribution network line after a high-frequency short-time excitation signal is injected into the power distribution network line; The denoising module is used to determine the target denoised signal corresponding to the response signal based on the noise parameters and the response signal using a preset wavelet operator denoising model; wherein the preset wavelet operator denoising model includes multiple wavelet integral layers; each wavelet integral layer is a neural network layer containing wavelet operators.

[0053] This invention provides a signal denoising device that acquires the original signal in a power distribution network line and determines noise parameters based on the original signal; acquires the response signal in the power distribution network line; wherein the response signal is the signal in the power distribution network line after a high-frequency short-time excitation signal is injected into the power distribution network line; and uses a preset wavelet operator denoising model to determine the target denoised signal corresponding to the response signal based on the noise parameters and the response signal; wherein the preset wavelet operator denoising model includes multiple wavelet integral layers; each wavelet integral layer is a neural network layer containing wavelet operators. The above technical solution, by determining noise parameters based on the original signal, ensures that the preset wavelet operator denoising model can learn the complex noise characteristics in the distribution network line, thereby accurately denoising the response signal. This avoids excessive suppression of high-frequency signals using fixed denoising parameters and prevents the filtering out of high-frequency transient features in the response signal. Furthermore, the wavelet integral layer integrates wavelet transform and neural networks, enabling the preset wavelet operator denoising model to combine the advantages of wavelet time-frequency analysis with the efficient feature extraction capabilities of deep learning. This ensures efficient acquisition of noise distribution characteristics in the distribution network line and effective signal features in the response signal, further improving the denoising capability of the response signal and making the results of fault detection of the distribution network line based on high-frequency short-time excitation signals more reliable.

[0054] Optionally, the preset wavelet operator denoising model includes a feature extraction layer, a wavelet integration layer, and a feature projection layer; both the feature extraction layer and the feature projection layer are convolutional neural networks; the denoising module includes: The first processing signal determination unit is used to extract features from the noise parameters and the response signal using the feature extraction layer to obtain a first processing signal; The second processing signal determination unit is used to perform linear transformation and wavelet processing on the first processing signal using the wavelet integration layer to obtain the second processing signal; The target denoising signal determination unit is used to perform dimensional transformation on the second processed signal using the feature projection layer to obtain the target denoising signal.

[0055] Optionally, the wavelet integral layer includes a linear transform branch and a wavelet integral operator branch; the linear transform branch is a neural network branch containing a preset parameter matrix; the wavelet integral operator branch is a neural network branch containing a forward wavelet operator and an inverse wavelet operator; the forward wavelet operator and the inverse wavelet operator are determined according to a preset wavelet basis function; the second processed signal determination unit includes: A linear transformation subunit is used to perform a linear transformation on the first processed signal based on a preset parameter matrix in the linear transformation branch to obtain a linearly transformed signal; wherein, the preset parameter matrix is ​​a learnable parameter matrix, and the parameters of the preset parameter matrix are determined according to the training results of the preset wavelet operator denoising model; The wavelet processing subunit is used to perform wavelet processing on the first processed signal based on the forward wavelet operator and the inverse wavelet operator in the wavelet integral operator branch to obtain the wavelet processed signal. An activation processing subunit is used to sum the linearly transformed signal and the wavelet-processed signal, and to activate the summation result using a preset activation function to obtain a second processed signal.

[0056] Optionally, the wavelet integral operator branch includes a preset weight matrix; the preset weight matrix is ​​a learnable parameter matrix, and the parameters of the preset weight matrix are determined based on the training results of the preset wavelet operator denoising model; the wavelet processing subunit is specifically used for: The first processed signal is decomposed using the forward wavelet operator to obtain the wavelet-decomposed signal; the wavelet-decomposed signal is then reconstructed using the inverse wavelet operator and the preset weight matrix to obtain the wavelet-processed signal.

[0057] Optionally, the loss function of the preset wavelet operator denoising model is set based on a preset objective; the preset objective includes: making the amplitudes of the predicted signal and the label signal corresponding to the same sampling time point similar, and making the amplitude changes of the predicted signal and the label signal corresponding to the same sampling time interval similar; the predicted signal is the output signal obtained by inputting the historical response signal into the preset wavelet operator denoising model during the training phase; the label signal is the ground truth signal corresponding to the historical response signal, used to characterize the noiseless signal features corresponding to the historical response signal.

[0058] Optionally, the preset loss function includes a first preset loss function and a second preset loss function; the first preset loss function is expressed by the following expression: ; in, This represents the first preset loss function; Represents the predicted signal; Represents the tag signal; Represents the total number of sampling time points; Representing the The amplitude of the predicted signal corresponding to each sampling time point; Representing the The amplitude of the tag signal corresponding to each sampling time point; The second preset loss function is expressed by the following expression: ; in, This represents the second preset loss function; A first-order differential operator representing amplitude and time; The first-order difference value represents the predicted signal and is used to characterize the magnitude change of the predicted signal within a preset sampling time interval. The first-order difference value representing the tag signal is used to characterize the magnitude change of the tag signal within a preset sampling time interval. Represents the total number of sampling points; Representing the The magnitude change of the predicted signal corresponding to each sampling point; Representing the The amplitude change of the tag signal corresponding to each sampling point.

[0059] The signal denoising device provided in the embodiments of the present invention can execute the signal denoising method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method.

[0060] Example 4 Figure 5 A schematic diagram of an electronic device 500 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0061] like Figure 5As shown, the electronic device 500 includes at least one processor 501 and a memory, such as a read-only memory (ROM) 502 and a random access memory (RAM) 503, communicatively connected to the at least one processor 501. The memory stores computer programs executable by the at least one processor. The processor 501 can perform various appropriate actions and processes based on the computer program stored in the ROM 502 or loaded into the RAM 503 from storage unit 508. The RAM 503 can also store various programs and data required for the operation of the electronic device 500. The processor 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0062] Multiple components in electronic device 500 are connected to I / O interface 505, including: input unit 506, such as keyboard, mouse, etc.; output unit 507, such as various types of monitors, speakers, etc.; storage unit 508, such as disk, optical disk, etc.; and communication unit 509, such as network card, modem, wireless transceiver, etc. Communication unit 509 allows electronic device 500 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0063] Processor 501 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 501 performs the various methods and processes described above, such as signal denoising methods.

[0064] In some embodiments, the signal denoising method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 508. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 500 via ROM 502 and / or communication unit 509. When the computer program is loaded into RAM 503 and executed by processor 501, one or more steps of the signal denoising method described above may be performed. Alternatively, in other embodiments, processor 501 may be configured to perform the signal denoising method by any other suitable means (e.g., by means of firmware).

[0065] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0066] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0067] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0068] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0069] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0070] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0071] This disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the signal denoising method provided in the above embodiments.

[0072] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0073] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method of signal denoising, the method comprising: include: Acquire the original signals in the distribution network lines, and determine the noise parameters based on the original signals; Acquire the response signal in the power distribution network line; wherein, the response signal is the signal in the power distribution network line after a high-frequency short-time excitation signal is injected into the power distribution network line; Using a preset wavelet operator denoising model, the target denoised signal corresponding to the response signal is determined based on the noise parameters and the response signal; wherein, the preset wavelet operator denoising model includes multiple wavelet integral layers; each wavelet integral layer is a neural network layer containing wavelet operators.

2. The signal denoising method of claim 1, wherein, The preset wavelet operator denoising model includes a feature extraction layer, a wavelet integration layer, and a feature projection layer; both the feature extraction layer and the feature projection layer are convolutional neural networks. The step of using a preset wavelet operator denoising model to determine the target denoised signal corresponding to the response signal based on the noise parameters and the response signal includes: The feature extraction layer is used to extract features from the noise parameters and the response signal to obtain a first processed signal; The first processed signal is linearly transformed and then processed by wavelet integration using the wavelet integral layer to obtain the second processed signal. The second processed signal is transformed using the feature projection layer to obtain the target denoised signal.

3. The signal denoising method of claim 2, wherein, The wavelet integral layer includes a linear transformation branch and a wavelet integral operator branch; the linear transformation branch is a neural network branch containing a preset parameter matrix; the wavelet integral operator branch is a neural network branch containing a forward wavelet operator and an inverse wavelet operator; the forward wavelet operator and the inverse wavelet operator are determined according to a preset wavelet basis function; The step of using the wavelet integral layer to perform linear transformation and wavelet processing on the first processed signal to obtain the second processed signal includes: The first processed signal is linearly transformed based on the preset parameter matrix in the linear transformation branch to obtain a linearly transformed signal; wherein, the preset parameter matrix is ​​a learnable parameter matrix, and the parameters of the preset parameter matrix are determined according to the training results of the preset wavelet operator denoising model; Based on the forward wavelet operator and the inverse wavelet operator in the wavelet integral operator branch, the first processed signal is subjected to wavelet processing to obtain a wavelet processed signal; The linearly transformed signal and the wavelet-processed signal are summed, and the summation result is activated using a preset activation function to obtain the second processed signal.

4. The signal denoising method of claim 3, wherein, The wavelet integral operator branch includes a preset weight matrix; the preset weight matrix is ​​a learnable parameter matrix, and the parameters of the preset weight matrix are determined according to the training results of the preset wavelet operator denoising model. The step of performing wavelet processing on the first processed signal based on the forward and inverse wavelet operators in the wavelet integral operator branch to obtain a wavelet-processed signal includes: Based on the positive wavelet operator, the first processed signal is decomposed into a wavelet decomposed signal; Based on the inverse wavelet operator and the preset weight matrix, the wavelet decomposed signal is reconstructed by wavelet to obtain the wavelet-processed signal.

5. The method of claim 1, wherein, The loss function of the preset wavelet operator denoising model is set based on preset objectives; the preset objectives include: making the amplitudes of the predicted signal and the label signal corresponding to the same sampling time point similar, and making the amplitude changes of the predicted signal and the label signal corresponding to the same sampling time interval similar; the predicted signal is the output signal obtained by inputting the historical response signal into the preset wavelet operator denoising model during the training phase; the label signal is the ground truth signal corresponding to the historical response signal, used to characterize the noiseless signal features corresponding to the historical response signal.

6. The signal denoising method of claim 5, wherein, The preset loss function includes a first preset loss function and a second preset loss function; the first preset loss function is expressed by the following expression: ； in, This represents the first preset loss function; Represents the predicted signal; Represents the tag signal; Represents the total number of sampling time points; Representing the The amplitude of the predicted signal corresponding to each sampling time point; Representing the The amplitude of the tag signal corresponding to each sampling time point; The second preset loss function is expressed by the following expression: ； in, This represents the second preset loss function; A first-order differential operator representing amplitude and time; The first-order difference value represents the predicted signal and is used to characterize the magnitude change of the predicted signal within a preset sampling time interval. The first-order difference value representing the tag signal is used to characterize the magnitude change of the tag signal within a preset sampling time interval. Represents the total number of sampling points; Representing the The magnitude change of the predicted signal corresponding to each sampling point; Representing the The amplitude change of the tag signal corresponding to each sampling point.

7. A signal denoising apparatus characterized by comprising: include: The noise parameter determination module is used to acquire the original signal in the distribution network line and determine the noise parameter based on the original signal; A response signal acquisition module is used to acquire the response signal in the power distribution network line; wherein, the response signal is the signal in the power distribution network line after a high-frequency short-time excitation signal is injected into the power distribution network line; The denoising module is used to determine the target denoised signal corresponding to the response signal based on the noise parameters and the response signal using a preset wavelet operator denoising model; wherein the preset wavelet operator denoising model includes multiple wavelet integral layers; each wavelet integral layer is a neural network layer containing wavelet operators.

8. An electronic device, comprising: The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the signal denoising method according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the signal denoising method according to any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the signal denoising method according to any one of claims 1-6.

Images